Gdf15 links adipose tissue lipolysis with anxiety

ABSTRACT Psychological stress changes both behaviour and metabolism to protect organisms. Adrenaline is an important driver of this response. Anxiety correlates with circulating free fatty

acid levels and can be alleviated by a peripherally restricted β-blocker, suggesting a peripheral signal linking metabolism with behaviour. Here we show that adrenaline, the β3 agonist

CL316,243 and acute restraint stress induce growth differentiation factor 15 (GDF15) secretion in white adipose tissue of mice. Genetic inhibition of adipose triglyceride lipase or genetic

deletion of β-adrenergic receptors blocks β-adrenergic-induced increases in GDF15. Increases in circulating GDF15 require lipolysis-induced free fatty acid stimulation of M2-like macrophages

within white adipose tissue. Anxiety-like behaviour elicited by adrenaline or restraint stress is eliminated in mice lacking the GDF15 receptor GFRAL. These data provide molecular insights

into the mechanisms linking metabolism and behaviour and suggest that inhibition of GDF15–GFRAL signalling might reduce acute anxiety. SIMILAR CONTENT BEING VIEWED BY OTHERS STRESS INCREASES

HEPATIC RELEASE OF LIPOCALIN 2 WHICH CONTRIBUTES TO ANXIETY-LIKE BEHAVIOR IN MICE Article Open access 08 April 2024 SIRT1 IN THE BNST MODULATES CHRONIC STRESS-INDUCED ANXIETY OF MALE MICE

VIA FKBP5 AND CORTICOTROPIN-RELEASING FACTOR SIGNALING Article 29 June 2023 EBI2 IS A NEGATIVE MODULATOR OF BROWN ADIPOSE TISSUE ENERGY EXPENDITURE IN MICE AND HUMAN BROWN ADIPOCYTES Article

Open access 29 March 2022 MAIN The incidence of anxiety disorders has rapidly increased over the last two decades, and anxiety is now the most prevalent psychiatric disorder, affecting

nearly 30% of the Western population at some point1. Acute stress-induced anxiety helps maintain the arousal and vigilance required to avoid repeated exposures to dangerous conditions2,3.

Catecholamines, namely noradrenaline and adrenaline (international nonproprietary names: norepinephrine and epinephrine, respectively), are critical to coordinating the central and

peripheral responses to psychological stress3,4,5,6. Central noradrenaline, particularly from the locus coeruleus, is known to control acute anxiety-like behaviour3,4. Interestingly, acute

peripheral administration of adrenaline also produces anxiety5,6,7,8,9, despite adrenaline not crossing the blood–brain barrier, and this effect is abolished by the β-blocker

propranolol10,11,12,13. Central mechanisms could explain the effects of propranolol, which readily crosses the blood–brain barrier; however, treatment with practolol, a β-blocker that does

not cross the blood–brain barrier, elicits similar anxiolytic effects14. These studies suggest that there may be peripheral endocrine signals linking β-adrenergic signalling with behaviour

that have yet to be identified. Growth differentiation factor 15 (GDF15) is a distant member of the transforming growth factor-β superfamily that circulates at low levels under normal

physiological conditions but is elevated during cellular stress and mitochondrial diseases15,16. GDF15 signals through its receptor, glial-cell-derived neurotrophic factor receptor α-like

(GFRAL), which is found exclusively in the brainstem17. GDF15–GFRAL signalling leads to behavioural changes, including reduced food intake18, nausea19 and aversive19,20 and anxiety-like

behaviours20,21,22. These effects may involve activation of the sympathetic and hypothalamic–pituitary–adrenal (HPA) axes23,24,25,26, both of which are classic responses to psychological

stress. In this study, we aimed to identify potential links between peripheral β-adrenergic activity and anxiety. We show that β-adrenergic stimulation activates adipocyte lipolysis, which

promotes the secretion of GDF15 from adipose tissue-resident macrophages, ultimately linking changes in peripheral metabolism with behaviour. RESULTS PSYCHOLOGICAL STRESS INCREASES GDF15

THROUGH Β-ADRENERGIC SIGNALLING Adipose tissue is a major endocrine organ that links metabolism with neuronal circuits and is acutely sensitive to changes in β-adrenergic

signalling5,6,7,8,9,10,13,27. We treated mice with saline or adrenaline injected intraperitoneally (IP) (Fig. 1a). In agreement with previous studies28, adrenaline induced anxiety-like

behaviour without affecting overall activity levels (Fig. 1b). To identify adipose-derived endocrine signals, we collected gonadal white adipose tissue (gWAT) and serum 1 h after treatment

with saline or the same dose of adrenaline from another cohort of mice. As anticipated29, adrenaline produced marked and distinct separation in principal component analysis (Fig. 1c), with

the most upregulated differentially expressed genes (DEGs) associated with the unfolded protein response and thermogenic and adaptive gene signatures (Fig. 1d and Extended Data Fig. 1a).

From this gene set, we then isolated genes encoding secreted protein products (that is, ligands)30 and found that _Gdf15_ was the most highly upregulated (_P_ = 6.59 × 10−29) secreted

factor, showing a >4 log2-fold increase in adrenaline-treated mice compared with saline-treated mice (Fig. 1e and Extended Data Fig. 1b); indeed, _Gdf15_ was one of the most upregulated

genes in gWAT. Corresponding with elevations in gWAT _Gdf15_ mRNA, GDF15 protein in serum was increased nearly five-fold within 1 h and remained elevated up to 4 h after injection (Fig. 1f).

Reverse transcription followed by qPCR confirmed that adrenaline produced an approximately eight-fold increase in _Gdf15_ in gWAT without affecting _Gdf15_ in the liver, kidney,

subcutaneous inguinal WAT (iWAT) or brown adipose tissue (Fig. 1g). _Gdf15_ expression was slightly increased in the lung and heart but was undetectable in skeletal muscle (Fig. 1g).

Consistent with lower gWAT mass, serum GDF15 levels were lower in age-matched female mice than in male mice, but the relative response to adrenaline was similar between sexes (Extended Data

Fig. 1c). Adrenaline also induced a similar increase in GDF15 whether mice were housed at room temperature or thermoneutrality (Extended Data Fig. 1d). These data indicate that adrenaline

leads to marked increases in gWAT and serum GDF15 in mice. Previous studies characterizing GDF15-null mice found that these animals exhibit less anxiety-like behaviour in open-field and

elevated-plus tests21,22. Psychological stress can be acutely induced in mice by changing their cages, whereas chronic single housing for 3 days is a form of long-term psychological

stress31. We found that, consistent with adrenaline injection, acute cage change increased serum GDF15 levels; however, single housing did not (Fig. 1h). Consistent with our observations in

mice that chronic stress and GDF15 are not linked, two-sample Mendelian randomization (2SMR) of GDF15 within the UK Biobank and serum GDF15 in children who were overweight and obese did not

show associations with anxiety (Extended Data Fig. 2a–c). Although coding polymorphisms affect the measurement of circulating GDF15 in humans32, these data suggest that adrenergic-induced

GDF15 may be involved in acute but not chronic psychological stress. An acute psychological stress that reliably increases adrenaline in mice is physical restraint33. We observed increased

circulating adrenaline following acute physical restraint, although the levels were lower than after treatment with IP injected adrenaline (Extended Data Fig. 1e). Adrenaline signals in

adipose tissue through β-adrenergic receptors34. Therefore, to examine whether psychological stress induces GDF15 through β-adrenergic receptors, we physically restrained wild-type (WT) mice

and mice lacking β1, β2 and β3 adrenergic receptors (_Adrb1_, _Adrb2_ and _Adrb3_ triple knockout (BR−/−)) in 50-ml conical tubes (Fig. 1i). Restraint of WT mice increased serum and gWAT

GDF15; however, this response was eliminated in BR−/− mice (Fig. 1j,k). Baseline GDF15 levels were indistinguishable between genotypes and, importantly, lipopolysaccharide (LPS) increased

serum GDF15 levels35 and lowered blood glucose levels similarly in WT and BR−/− mice (Extended Data Fig. 3a,b), demonstrating that these mice have an intact GDF15 response to

non-β-adrenergic stimuli. These results show that psychological stress increases GDF15 and that this is dependent on β-adrenergic signalling. ATGL IS CRITICAL FOR PROMOTING

Β-ADRENERGIC-INDUCED GDF15 Lipolysis is a major regulator of the transcriptional responses within adipose tissue29 and is increased by β-adrenergic-induced activation of adipose triglyceride

lipase (ATGL), the rate-limiting enzyme for the release of free fatty acids from WAT36. To determine whether ATGL is responsible for regulating GDF15 secretion from adipose tissue, we

treated adipocyte-specific ATGL-knockout mice (AdATGL−/−) and their floxed littermate controls (AdATGLflox/flox) with IP injected saline or CL316,243 (CL), a potent and selective

β3-adrenergic agonist. ATGL expression was ~90% lower in gWAT of AdATGL−/− mice (Fig. 2a). CL increased serum nonesterified fatty acids and markers of the unfolded protein response, _Atf4_

and _Chop_, in gWAT of AdATGLflox/flox animals (Fig. 2b–d). These responses were eliminated in AdATGL−/− mice despite β-adrenergic signalling remaining intact, as evidenced by the similar

increase in _Ppargc1a_ expression and phosphorylation of cAMP-dependent protein kinase (PKA) substrates in AdATGL−/− and floxed mice (Fig. 2e,f). Consistent with our observations in BR−/−

mice, basal GDF15 levels were similar between AdATGLflox/flox and AdATGL−/− mice; however, CL-induced increases in circulating and gWAT GDF15 were eliminated in AdATGL−/− mice (Fig. 2g,h).

Treatment with IP injected cilostamide, a phosphodiesterase-3 inhibitor that leads to increased cellular cAMP levels, exaggerated the effects of CL on circulating GDF15 levels (Fig. 2i),

supporting the involvement of the β3 adrenergic receptor–cAMP–ATGL pathway in stimulating GDF15. Physical restraint stress promotes adipose tissue lipolysis37. Serum glycerol was increased

in AdATGLflox/flox but not in AdATGL−/− mice in response to acute tube restraint (Fig. 2j). Similarly, circulating GDF15 levels were increased in AdATGLflox/flox but not in AdATGL−/− mice

(Fig. 2k). We tested the behavioural response to acute restraint stress in these mice, but AdATGL−/− mice became hypoglycaemic following restraint, and this was associated with marked

reductions in overall activity during open-field tests (Extended Data Fig. 4a–c), an effect that was likely secondary to the defect in adipose tissue lipolysis. To evaluate whether

β-adrenergic signalling increases GDF15 in a cell-autonomous manner, we cultured and differentiated mouse white adipocytes before treating them with CL and the ATGL inhibitor ATGListatin38.

As expected, CL increased glycerol release in the medium, and this was blocked by ATGListatin; however, neither CL nor ATGListatin altered GDF15 levels in the medium (Fig. 2l,m). These data

suggest that GDF15 is unlikely to be derived from adipocytes in response to adrenergic stimulation. As adipocyte-specific ablation of ATGL inhibited GDF15 secretion but isolated adipocytes

did not secrete GDF15 in response to CL, we hypothesized that another cell type within adipose tissue might be responsible for GDF15 secretion. To test this, we fractioned gWAT and iWAT into

adipocytes and the stromal vascular fraction (SVF). Consistent with our hypothesis and recent studies39, _Gdf15_ expression was higher in the SVF of both WAT depots but particularly in gWAT

(Fig. 2n). Moreover, in response to IP injected adrenaline, _Gdf15_ was increased in only the SVF (Fig. 2o), whereas the related cytokine, _Fgf21_, was increased in only the adipocyte

fraction (Extended Data Fig. 5a). Appropriate separation was confirmed as leptin was found only in the adipocyte fraction (Extended Data Fig. 5b). These results point towards ATGL-dependent

intercellular communication mediating β-adrenergic-induced adipose tissue GDF15 secretion. MACROPHAGES ARE THE PRIMARY SOURCE OF SERUM GDF15 To determine the probable cell type secreting

GDF15 in adipose tissue, we analysed publicly available single-cell RNA-sequencing (scRNA-seq) data of cells from gWAT of CL-treated mice40. Cells were divided into either lineage marker

positive (Lin+) (including monocytes, macrophages and B cells) or Lin− (including fibroblasts and adipocytes). _Gdf15_ was primarily expressed in Lin+ cell types, especially in macrophages

(Fig. 3a,b), with negligible expression in any Lin− cell type, including adipocytes (Extended Data Fig. 5c). In response to CL treatment, there was increased _Gdf15_ expression only in the

macrophage population (Fig. 3a). Because macrophage populations in adipose tissue are highly heterogeneous41, we used established gene signatures to identify three major clusters of

macrophage populations: classically activated M1-like (_Adgre1_/_F480_ and _iNos_/_Cd86_/_Cd80_), alternatively activated M2-like (_Adgre1_/_F480_ and _Cd163_/_Arg1_) and macrophages that

express _Adgre1_/_F480_ but no other markers for M1- or M2-like activation (Fig. 3c). We observed that CL led to the accumulation of both M1- and M2-like macrophages in gWAT, as noted in the

original report40. CL increased _Gdf15_ in M2-like _Adgre1_/_F480_- and _Cd163_/_Arg1_-positive macrophages to a greater degree than observed in M1-like macrophages (Fig. 3c,d). We

subsequently examined whether increases in _Gdf15_ also occurred after acute treatment with adrenaline by isolating F4/80+ cells from the SVF of gWAT (Extended Data Fig. 5d). Consistent with

the scRNA-seq results, we found that _Gdf15_ expression was higher in F4/80+ cells than in other cells within the SVF and that adrenaline induced increases in _Gdf15_ in F4/80+ cells

without any change in F4/80− cells of the SVF (Fig. 3e). Taken together, these data demonstrate that macrophages are the primary source of GDF15 in response to β-adrenergic-induced

lipolysis. Fatty acids are the primary mediators of the transcriptional response to adipocyte lipolysis29. To determine whether fatty acids are responsible for GDF15 secretion from

macrophages, we turned to bone marrow-derived macrophages (BMDMs) polarized with LPS + interferon-γ (IFNγ) (M1-like activation) or interleukin-4 (IL-4) (M2-like) or left untreated (M0) (Fig.

3f)42. We then treated BMDMs with two of the most abundant fatty acid species that are increased in the circulation by β-adrenergic agonists43, namely oleate, an unsaturated fatty acid, and

palmitate, a saturated fatty acid. We also included tunicamycin, a compound that activates the unfolded protein response and is known to increase GDF15 (refs. 44,45). In line with our

scRNA-seq results, palmitate and tunicamycin stimulated the secretion of GDF15 only in M0 and M2-like macrophages but not in M1-like macrophages (Fig. 3g). Similar results were seen at the

transcriptional level, with _Gdf15_ expression being increased by palmitate only in M0 and M2-like macrophages (Fig. 3h). These results align with earlier work showing that adipocytes

preferentially secrete palmitate, which is a critical regulator of macrophage metabolism46. Macrophage polarization was confirmed by the expression levels of _Arg1_ and _iNos_, which were

elevated in M2- and M1-like macrophages, respectively (Extended Data Fig. 5e). Under basal conditions, GDF15 secretion into the medium was similar between M1-like and M2-like polarized

cells, both of which showed greater GDF15 levels than unpolarized M0 cells (Extended Data Fig. 5f). Importantly, treatment of macrophages with adrenaline did not affect GDF15 secretion

(Extended Data Fig. 5g), indicating that adrenergic signalling does not directly stimulate macrophage GDF15 secretion. We then used these scRNA-seq data to compare the M1- and M2-like

macrophage populations to identify possible mechanisms behind the M2-like-specific palmitate-induced GDF15 secretion. M2-like macrophages are known to use fatty acid metabolism to a greater

degree than M1-like macrophages; consistent with this, our analysis of the scRNA-seq data found that fatty acid metabolic processes (Fig. 3i) and fatty acid transporters (Extended Data Fig.

5h) were upregulated in M2- compared with M1-like macrophages. Many fatty acid transporters are upregulated by the transcription factor peroxisome proliferator-activated receptor-γ (PPARγ),

which is enriched in M2-like macrophages41 and was recently shown to increase the secretion of GDF15 from hepatocytes when activated47. To determine whether PPARγ is involved in controlling

the secretion of GDF15, we treated M2-like BMDMs with rosiglitazone, a PPARγ agonist, and T0070907, a PPARγ antagonist. We observed that rosiglitazone increased GDF15 levels in the medium

and that this effect was blocked by T0070907 (Fig. 3j). T0070907 also blocked palmitate-induced GDF15 secretion from M2-like macrophages (Fig. 3k). Together, these data show that palmitate

drives the secretion of GDF15 specifically from macrophages polarized to an alternative M2-like phenotype through a mechanism that probably involves PPARγ. Feeding mice a diet high in

saturated fatty acids can promote anxiety-like behaviour (reviewed in ref. 48). Exogenous delivery of fatty acids such as palm oil by acute oral gavage increases the expression of GDF15 in

the kidney and other tissues, and this leads to elevations in serum GDF15, an effect that peaks after 4 h (ref. 49). We tested whether an oral gavage of palm oil can produce anxiety-like

behaviour after 4 h and, if so, whether this effect is blunted in mice lacking the receptor for GDF15, GFRAL (_Gfral_−/− mice). In contrast to adrenaline injection or tube restraint,

treatment with palm oil did not increase anxiety-like behaviour in either genotype (Extended Data Fig. 6a). We next tested the effects of a high-fat diet predominantly composed of palm oil

for 4 weeks and found similarly increased anxiety-like behaviour in both WT and _Gfral_−/− mice (Extended Data Fig. 6b,c). These results suggest that GDF15–GFRAL signalling is not involved

in mediating anxiety-like behaviour in response to acute or chronic exogenous fatty acids. GDF15 PROMOTES ANXIETY-LIKE BEHAVIOUR THROUGH GFRAL GDF15 regulates food intake and energy

expenditure through its receptor, GFRAL15,18,26, the expression of which is isolated to the area postrema and nucleus tractus solitarius of the hindbrain17. To date, no other ligands besides

GDF15 have been shown to signal through GFRAL. To establish whether GFRAL signalling is important for mediating anxiety-like behaviour, we treated WT or GFRAL-knockout (_Gfral_−/−) mice

with IP injected adrenaline (as in Fig. 1a) 1 h before open-field tests. Adrenaline again induced anxiety-like behaviour in WT mice, but this response was lost in _Gfral_−/− mice (Fig. 4a)

without affecting total activity (Fig. 4b). Anxiety is also reflected by nonambulatory movements, such as grooming50. Further supporting that GFRAL is necessary for the behavioural responses

to acute stress, nonambulatory movements were increased in WT but not in _Gfral_−/− mice in response to IP injected adrenaline (Extended Data Fig. 7a); again, total physical activity was

not altered by adrenaline (Extended Data Fig. 7b). Other responses to adrenaline, including food intake, energy expenditure and substrate use, were similar between genotypes (Extended Data

Fig. 7c–e). We next tested whether GFRAL is important for the behavioural responses to psychological stress by applying acute physical restraint before the open-field test. As expected3,

restraint stress reduced the time spent in the centre, but, remarkably, this response was lost in _Gfral_−/− mice (Fig. 4c). Importantly, this effect was independent of genotypic differences

in the total activity levels during the open-field test, which were similarly reduced regardless of genotype (Fig. 4d). Anxiety-like behaviour was distinct from food intake or nausea, both

of which are altered by stress and can be controlled by GDF15 (ref. 19), given that the same restraint protocol did not affect chow or kaolin clay consumption in WT or _Gfral_−/− mice

(Extended Data Fig. 7f,g); stress-induced hypophagia and nausea have both been attributed to glucagon-like peptide 1 (refs. 51,52,53). In support of the anxiogenic effects of GDF15, a single

intraperitoneal injection of a pharmacological dose of GDF15 (ref. 26) produced anxiety-like behaviour in the open-field and light–dark box tests without affecting overall physical activity

(Extended Data Fig. 7h,i). Critically, the anxiogenic effects of GDF15 persisted following repeated daily exposure, suggesting that tachyphylaxis did not occur (Extended Data Fig. 7j–l),

similar to what we and others have shown for food intake and body weight loss with repeated GDF15 treatment26,54. Several brain regions are involved in the acute stress response2,3,55,56. To

determine the neurobiological substrates of GDF15-induced anxiogenesis, we measured _Fos_, an immediate early gene associated with neuronal activation, in several structures known to be

involved in stress-related behaviours. The locus coeruleus is the main source of noradrenaline in the mammalian brain and is one of the first structures recruited following stressful

stimuli3,4,57. Given that GDF15’s metabolic effects were recently shown to be dependent on β-adrenergic activity26, the locus coeruleus represented a particularly relevant target. However,

high-dose GDF15 did not increase c-Fos expression in locus coeruleus noradrenergic neurons (Extended Data Fig. 8a) nor in the medial prefrontal cortex and basolateral amygdala, both

downstream projections of the locus coeruleus known to drive anxiogenesis (Extended Data Fig. 8b)4,58. In contrast, GDF15 showed increased activity in the central amygdala and the bed

nucleus of the stria terminalis (Fig. 4e), each of which has well-established roles in anxiogenesis59,60,61,62,63. Consistent with other studies, we also observed activation within the

paraventricular nucleus of the hypothalamus (Fig. 4e)23,64. High levels of circulating GDF15 can activate corticotropin-releasing hormone (CRH) neurons within the paraventricular nucleus of

the hypothalamus and activate the HPA axis, leading to the secretion of corticosterone in mice23,24,64. We explored whether GDF15–GFRAL signalling is important for activating the HPA axis

under more physiological conditions by applying physical restraint stress to WT and _Gfral_−/− mice. Expectedly, restraint stress increased circulating corticosterone in both genotypes,

reflecting activation of the HPA axis, but significantly less so in _Gfral_−/− mice (Fig. 4f), despite circulating GDF15 levels being modestly elevated by restraint stress in both genotypes

(Extended Data Fig. 8c). CRH is critical in driving stress-induced behavioural responses, such as anxiety and aversion3,50. Intriguingly, exogenous CRH led to a marked increase in

circulating GDF15 levels (Fig. 4g), further demonstrating the involvement of GDF15 and GFRAL in the stress response. Importantly, exogenous CRH increased adrenocorticotropic hormone (ACTH)

and corticosterone levels, whereas dexamethasone decreased corticosterone levels similarly in WT and _Gfral_−/− mice, suggesting that there was no impairment in HPA axis signalling (Extended

Data Fig. 8d–f). Together, these data support the role of GDF15 in regulating acute stress-induced anxiety-like behaviour (Fig. 4h), possibly through the recruitment of multiple anxiogenic

brain regions56, and emphasize that GDF15–GFRAL signalling is transmitted broadly throughout the brain despite the brainstem-localized receptor expression. DISCUSSION Mammals have evolved

many mechanisms for detecting and avoiding noxious stimuli such as foodborne toxins, bacterial infections and allergens16,65,66,67,68. GDF15 has been linked to all of these by inducing

nausea, emesis (or pica) and aversive behaviour19,64,69. However, in addition to physical stimuli, psychological stress produces its own set of physiological and behavioural responses

intended to protect the organism. The fight-or-flight response is a classic example of psychological stress leading to metabolic responses that mobilize endogenous energy stores from adipose

tissue13,70 and elicit behavioural responses such as anxiety5,6,7,8,9. For decades, there has been evidence that adipose tissue lipolysis and peripheral adrenergic activity are associated

with feelings of anxiety10,11,12,13, but the link between these responses has remained elusive. Here, we identified a mechanism whereby adrenergic activation of adipocyte lipolysis promotes

the secretion of GDF15 from M2-like alternatively activated macrophages within adipose tissue, which, through its receptor, GFRAL, is critical for appropriate behavioural and physiological

responses to acute psychological stress. These results are consistent with previous studies in which chronically supraphysiological levels of GDF15 promoted anxiety-like behaviour whereas

lifelong deletion of GDF15 reduced anxiety21,22. We also identified various anxiogenic brain regions, including the central amygdala, bed nucleus of the stria terminalis and paraventricular

nucleus, that could be involved in mediating these responses. GDF15, at least at high circulating levels, has been shown to reduce physical activity71,72, produce anxiety-like

behaviour20,21,22 and activate the β-adrenergic25,26,68 and HPA axes23,24. Indeed, optogenetic activation of GFRAL-expressing neurons is sufficient to activate the HPA axis and increase the

circulating cortisol levels24. Clinical trials with GDF15 analogues did not report increases in anxiety73,74. However, while our paper was in review, a study was published in which plasma

proteomics conducted in individuals from the UK Biobank demonstrated that GDF15 is the circulating factor with the strongest association with anxiety in humans compared with more than 3,000

other measured proteins75. Herein, we show that GDF15–GFRAL signalling is a critical component of responses to acute psychological stress, including anxiety-like behaviour and HPA axis

activation. Some important limitations to the current data should be considered. First, the M1- and M2-like macrophage distinction is an oversimplification of the highly heterogeneous

macrophage populations. Other publications have identified distinct macrophage subpopulations that secrete GDF15, at least in skeletal muscle76, so it will be interesting to determine

whether a particular subpopulation is responsible for the fatty acid-dependent secretion of GDF15 demonstrated here. Moreover, it will need to be determined how palmitate drives GDF15

secretion from macrophages and whether endogenous adipocyte-derived palmitate is sufficient to elicit this response. It is appealing to speculate that fatty acid-sensing PPARs, which have

been shown to regulate GDF15 (ref. 47), could be involved, but this will need to be confirmed in vivo. Despite the rapid increase in the rates of anxiety, there is a dearth of new targets

for its treatment. The interactions between the brain and peripheral immune cells in the development of stress have been a growing area of interest77,78,79, but previous studies have focused

largely on classic inflammatory cytokines. Here, we provide additional insights into these interactions by identifying a new pathway for immune–brain crosstalk mediated by

adrenergic-induced ATGL-dependent lipolysis and M2-like macrophage-secreted GDF15. These results raise the possibility that blocking GDF15–GFRAL signalling could mitigate acute

stress-induced anxiety. METHODS MICE Germline GFRAL-knockout mice (_Gfral_−/−) were generated as previously described, with breeding pairs provided by R. Seeley26. Mice lacking β1, β2 and β3

adrenergic receptors (BR−/−) were generated as previously described26,80. Adipocyte-specific ATGL-knockout animals were generated by crossing _Adipoq_-Cre mice with _Atgl_/_Pnpla2_-floxed

mice81; _Atgl_/_Pnpla2_flox/flox littermates were used as controls. Animal studies were carried out at McMaster University (210104), the University of Toronto (21-467 and 24-0362H) or

Washington University in St. Louis (20-0139). All animals used in the study were housed and cared for in accordance with the Guidelines for Animal Use at McMaster University and were

approved by the McMaster University Animal Ethics Research Board or the ethics boards of their respective facilities. All mice were group-housed on a 12-h light–dark cycle with ad libitum

access to food and water. The animals were housed either in a HEPA-filtered unit at room temperature or in specific pathogen-free microisolators at 29 °C and 40–60% relative humidity.

Experiments were performed on mice at ages between 16 and 24 weeks. All mice were male unless stated otherwise, including for cell isolations. Experiments were performed, and mice were

killed in a fed state between 09:00 and 12:00. Terminal blood was collected by cardiac puncture, and blood from live animals (for example, time-course GDF15 levels) was collected from a tail

vein. Blood samples were centrifuged at 8,000_g_ for 10 min at 4 °C after clotting at room temperature for 30 min, and the supernatant was collected. Tissues were collected following

anaesthetization with ketamine (75 mg kg−1) and xylazine (10 mg kg−1). Tissues were frozen at −80 °C until future analyses. Male WT and _Gfral_−/− mice were placed on a high-fat,

high-fructose diet (40 kcal% fat (mostly palm oil), 20 kcal% fructose with 0.02% cholesterol; Research Diets, D19101102) starting at 12–20 weeks of age for 4 weeks. Palm oil (Sigma-Aldrich,

70905) was gavaged at ~09:00 as described previously49. MOUSE DRUG TREATMENTS Adrenaline hydrochloride (Sigma-Aldrich, E4642) was diluted in saline and injected IP (~09:00) at a dose of 0.5 

mg per kg body weight based on previous studies82,83. CL (Sigma-Aldrich, C5976) was diluted in saline and injected IP (~09:00) at a dose of 1 mg per kg body weight84. Cilostamide

(Sigma-Aldrich, C7971; 10 mg per kg body weight, injected IP) was prepared in 10% DMSO and 5% Kolliphor85. All control mice received saline IP injected at the same volume. Recombinant human

GDF15 (5 nM kg−1; Novo Nordisk) or vehicle was injected IP at ~09:00 (ref. 26). To test for tachyphylaxis, we injected mice with either vehicle or GDF15 (5 nM kg−1) for 9 days at ~09:00.

Chow diet was measured daily. On the 10th day, mice were injected with either vehicle or GDF15 (5 nM kg−1), and open-field tests were performed 1 h later (described below). LPS

(Sigma-Aldrich, 0111:B4, L2630) was injected IP at a dose of 2 mg kg−1 starting at 09:00, and control mice were injected with saline. CRH (Sigma-Aldrich, C3042) was injected IP at ~09:00 at

a dose of 90 μg kg−1, and blood was collected from a tail vein at 0.5 and 3 h after injection86. Dexamethasone (Sigma-Aldrich, D4902) was injected IP at ~09:00 at a dose of 100 μg kg−1, and

blood was collected from a tail vein 6 h following injection87. HUMAN PAEDIATRIC ANXIETY SAMPLES Children with obesity who were enrolled in the Canadian Pediatric Weight Management Registry

at the McMaster site were included in this study. Study participants with an available fasting blood sample at their initial visit and a clinical diagnosis of anxiety but without

antipsychotic or antidepressant medications (_n_ = 23) were compared with those without any diagnosis of anxiety and also free from the use of medications (_n_ = 24). Control participants

and those with anxiety were matched for age (12.50 ± 2.97 and 12.44 ± 2.86 years), sex (12/11 and 12/12 male-to-female ratio), body weight (87 ± 31.88 and 88 ± 35.32 kg) and body mass index

(33.18 ± 6.41 and 34.08 ± 8.07 kg m−2). Serum GDF15 was assessed from previously frozen samples in duplicate, as described below. SEPARATION OF ADIPOSE TISSUE FRACTIONS Each gWAT or iWAT

sample was combined from two mice. WAT was collected, quickly rinsed in warm PBS, minced and incubated in collagenase (10 mM HEPES–Krebs–Ringer buffer, 4% BSA, 1.5 mg ml−1 collagenase type I

(Gibco, 17100017)) at 37 °C with gentle agitation for 30 min (gWAT) or 45 min (iWAT). Tissues were filtered (100 μm), and fractions were separated by repeated centrifugation at 500_g_ for 5

 min. F4/80+ cells were isolated using a commercially available kit as per the manufacturer’s instructions (EasySep Mouse F4/80 Positive Selection Kit; STEMCELL Technologies, cat. no.

100-0659) along with an EasySep magnet (cat. no. 18000). ADIPOCYTE DIFFERENTIATION Immortalized mouse white adipocytes were generated as previously described88. Preadipocytes were grown to

confluence in DMEM supplemented with 1% GlutaMAX, 1% penicillin–streptomycin, 10% FBS, insulin (20 nM) and T3 (1 nM; differentiation medium). Confluent cells were incubated for 48 h in

differentiation medium further supplemented with isobutylmethylxanthine (0.5 mM), dexamethasone (0.5 μM) and indomethacin (0.125 mM; induction medium). All chemicals were from Sigma, and

experiments were performed within 20 passages following immortalization. Adipocytes were then treated with either CL (10 μM; Sigma-Aldrich, C5976), ATGListatin (40 μM; Sigma-Aldrich,

SML1075), neither of the compounds (control) or a combination of the two compounds for 24 h in serum-free DMEM supplemented with 1% BSA (Sigma-Aldrich, cat. no. A8806). Medium was collected,

and excess lipids and cellular debris were removed by centrifugation. Cells were washed with PBS and snap-frozen either in TRIzol for RNA extraction or in cell lysis buffer for protein

content quantification using BCA. Lipolysis was assessed by measuring glycerol levels in the medium using a commercially available kit (Cayman Chemical, cat. no. 10010755). BMDM ISOLATION,

DIFFERENTIATION AND POLARIZATION Femurs and tibias were collected from mice and cleaned of remaining muscle and soft tissues, and both ends of each bone were removed89. The bone marrow was

extracted by centrifugation at 1,900_g_ for 5 min. The marrow was resuspended in 1 ml DMEM (Wisent, cat. no. 319-005-CL, 25 mM glucose) and strained through a 40-μm cell strainer into a

50-ml Falcon tube. Cells were cultured for 4 h at 37 °C and 5% CO2. After incubation, nonadherent cells were collected and combined with M-CSF (20 ng ml−1; Peprotech, 315-02) to stimulate

macrophage differentiation and plated for 7 days in DMEM (Wisent, cat. no. 319-005-CL) supplemented with FBS (10%, Wisent, cat. no. 098150) and penicillin–streptomycin (1%, Invitrogen).

After differentiation, cells were washed, collected and plated onto 12-well plates (1 × 106 cells per ml) and left to adhere overnight. Fully differentiated BMDMs were then polarized either

with 100 ng ml−1 LPS (Sigma-Aldrich, L2630) + 20 ng ml−1 IFNγ (Peprotech, 315-05) to an M1-like phenotype or with 20 ng ml−1 IL-4 (Peprotech, 214-14) to an M2-like phenotype for 24 h (ref.

42). Following polarization, cells were treated with either sodium palmitate (500 μM, Sigma-Aldrich, P9767), sodium oleate (500 μM, Sigma-Aldrich, O7501), tunicamycin (5 ng ml−1, Cayman

Chemical, 11089-65-9), rosiglitazone (1 μM; Sigma-Aldrich, R2408) or T0070907 (1 μM; Sigma-Aldrich, T8703) for another 24 h in DMEM (Wisent, 319-005-CL) supplemented with 1% fatty acid and

endotoxin-free BSA (Sigma-Aldrich, cat. no. A8806). Medium was collected after 24 h and frozen at −80 °C until analysis. Cells were rinsed in PBS and collected either in TRIzol for RNA

extraction (see below) or in cell lysis buffer for protein quantification using BCA. RNA ISOLATION, CDNA SYNTHESIS AND QPCR Tissues were homogenized in TRIzol reagent. An RNeasy kit (Qiagen,

cat. no. 74106) was used for subsequent total RNA extraction and purification according to the manufacturer’s instructions. cDNA synthesis was performed using the SuperScript IV Reverse

Transcriptase kit (Invitrogen, cat. no. 18090010) following the manufacturer’s instructions. cDNA expression for specific genes was detected by qPCR using the AmpliTaq Gold DNA Polymerase

kit (Applied Biosystems, cat. no. N8080241). Relative mRNA levels were quantified with the Δ_C_T method using mouse _Ppia_ (Mm02342430_g1) as an endogenous control, except for adipocyte and

SVF comparisons for which _Polr2a_ was used (Mm00839502_m1). The gene-specific primers used were as follows: _Gdf15_ (Mm00442228_m1), _Atf4_ (Mm00515324_m1), _Chop_/_Ddit3_ (Mm01135937_g1),

_iNos_ (Mm00440502_m1), _Arg1_ (Mm00475988_m1), _Ppargc1a_ (Mm01208835_m1), _Emr1_/_F480_/_Adgre1_ (Mm00802529_m1), _Fgf21_ (Mm07297622_g1), _Atgl_/_Pnpla2_ (Mm00503040_m1) and _Lep_

(Mm00434759_m1). All probes were purchased from Thermo Fisher. RNA-SEQ AND ANALYSES For RNA-seq, mice were injected with adrenaline (0.5 mg kg−1 IP at 09:00) and gWAT was collected after 1 

h. RNA was extracted as described above. Raw RNA-seq FASTQ data were imported into Galaxy for quality control and processing steps. The FastQC tool was used to check read quality, and the

Cutadapt tool was used to trim adaptor sequences and remove low-quality reads. The remaining reads were aligned to the mm10 _Mus musculus_ reference genome with the HISAT2 tool and

quantified with the featureCounts tool. These counts data were imported into R for differential expression analysis with the DESeq2 package to detect DEGs (adjusted _P_ value (_P_-adj) < 

0.05). Principal component analysis was performed in R using variance stabilizing transformation (VST)-normalized data from the DESeq2 analysis. Pathway analysis was performed with the GSEA

software in combination with MSigDB gene sets. A list of human genes encoding secreted proteins (ligands) was obtained from the HUGO Gene Nomenclature Committee website. Mouse genes

orthologous to these human ligand-encoding genes were obtained using the g:Profiler tool. A heat map was created with DEGs belonging to this group of ligand-encoding genes in R with the

pheatmap package, using _z_-scored VST-normalized data. These genes were also highlighted in a volcano plot, which was created with the Enhanced Volcano package in R, using fold-change and

_P_-adj values from the DESeq2 analysis. SINGLE-CELL RNA-SEQ The scRNA-seq data used for the analyses described in this article were obtained from the National Center for Biotechnology

Information Sequence Read Archive under reference number SRP145475 (ref. 40). Cell Ranger was used to perform sample qualification, alignment, filtering, counting and aggregation using the

Linux system or R and RStudio software. Clustering and gene expression were visualized with the 10X Genomics Loupe Browser (v.6.5.0). 2SMR USING GWAS SUMMARY DATA 2SMR was performed using

the exposure and outcome from two nonoverlapping and independent datasets to conduct the summary-level instrument–exposure analysis and the instrument–outcome association analysis. 2SMR was

performed using the R package TwoSampleMR (v0.5.6)90. To verify the causal effect of GDF15 on nerves, anxiety, tension or depression in humans, we performed 2SMR using the exposure dataset

(GDF15, GWAS ID: ebi-a-GCST90011998, sample size: 21,758)91,92 and outcome dataset (nerves, anxiety, tension or depression, GWAS ID: ebi-a-GCST90013910, sample size: 407,746)93. We

identified genetic variants (single-nucleotide polymorphisms) associated with blood GDF15 protein levels in the GWAS catalogue dataset based on cis-protein quantitative trait loci (within

500 kb of the _Gdf15_ gene) and further selected proxy single-nucleotide polymorphisms by linkage disequilibrium clumping (_P_1 = 5 × 10−8, clump = TRUE, _P_2 = 1 × 10−7, _r_2 = 0.01, kb = 

10,000). After dropping duplicate exposure–outcome summary sets, we further performed sensitivity analyses, including heterogeneity statistics, horizontal pleiotropy and leave-one-out

analysis. After confirming that there was no heterogeneity or horizontal pleiotropy, we next performed Mendelian randomization analysis and visualized the results using the scatter plot and

forest plot functions in the R package TwoSampleMR. The inverse-variance weighted method was used to assess the significance of the causal effect of the exposure on the outcome. 2SMR was

performed using R and RStudio. SERUM AND MEDIUM ANALYSES Serum and cell-based GDF15 levels were measured by ELISA (R&D Systems, DY6385) as per the manufacturer’s instructions18,45. Serum

corticosterone (Thermo Fisher, EIACORT), adrenocorticotropic hormone (Abcam, ab263880) and adrenaline (Abnova, KA1882) were measured following the manufacturer’s instructions. C-FOS

IMMUNOFLUORESCENCE AND QUANTIFICATION C57BL/6J mice were injected with GDF15 (5 nM kg−1) or vehicle; 90 min later, the mice were anaesthetized with a mixture of ketamine (69.57 mg ml−1),

xylazine (4.35 mg ml−1) and acepromazine (0.87 mg ml−1) IP and perfused with 10 ml of 0.1 M phosphate buffer (PB) followed by 25 ml of 4% paraformaldehyde (PFA) in 0.1 M PB. The brains were

removed, postfixed overnight (4% PFA in 0.1 M PB) and kept at 4 °C in 30% sucrose solution until cutting. Coronal sections (40 µm) were obtained using a microtome (Leica SM 2000R) and

serially collected in PBS. Sections were washed in PBS (3 × 10 min) and preincubated in PBS containing Triton X-100 (0.3%) and normal goat serum (5%) for 1 h. Sections were then incubated

overnight at room temperature in PBS containing Triton X-100 (0.3%), normal goat serum (1%) and rabbit anti-c-Fos (1:1,000, Cell Signaling Technology). Sections were then washed in PBS (3 × 

10 min), incubated with Alexa Fluor 488 goat anti-rabbit secondary antibody (1:400, Invitrogen) in PBS containing Triton X-100 (0.3%) and normal goat serum (1%) for 2 h, and washed in PBS (3

 × 10 min). Sections were finally serially mounted in Vectashield medium with DAPI (Vector Laboratories). Image acquisition was performed using Zeiss Axio Scan 7. WESTERN BLOTTING Proteins

were extracted using lysis buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl, 100 mM NaF, 10 sodium pyrophosphate, 5 mM EDTA, 250 mM sucrose, 1 mM DTT, 1 mM sodium orthovanadate, 1% Triton

X-100 and cOmplete Protease Inhibitor Cocktail (Roche). The total protein concentration in the samples was measured by the BCA protein assay. Protein concentrations were adjusted and

diluted in 4× SDS sample buffer. Proteins were separated using SDS–PAGE gels and transferred to poly(vinylidene difluoride) membranes. After blocking for 1 h with 5% BSA in TBST at room

temperature, membranes were incubated with the primary antibodies overnight at 4 °C (phosphorylated PKA substrates (1:1,000, Cell Signalling #9624) or tubulin (1:1,000, Abcam #ab4074)).

After washing, membranes were incubated with a secondary antibody at room temperature for 1 h. Protein bands were visualized with the Fusion FX7 system (MBI) and quantified using ImageJ

software. TUBE RESTRAINT AND STRESS TESTS Mice were put into 50-ml Falcon tubes (beginning at ~09:00) with holes at both ends so the tail was exposed and there was sufficient air flow3.

Control mice remained in their home cage with food and water removed at the initiation of tube restraint. Immediately following tube restraint, mice were either anaesthetized for tissue

collection or submitted to open-field test protocols (see below). Alternatively, social isolation was induced by individually housing previously group-housed mice for 3 days and comparing

them with their continually group-housed littermates31. In the cage-switch model, mice were removed from their home cage and placed in an identical cage with the dirty bedding of

nonlittermate males31. Nondisturbed socially housed littermates were applied as the controls. To assess stress-induced food intake and nausea, we provided mice with kaolin clay (Research

Diets, K50001) in addition to their regular diet for 3 days before testing. Mice were restrained (starting at ~09:00) for 4 h and returned to their cages (group-housed) with access to

standard chow and kaolin clay pellets, which were weighed before and after 18 h (~09:00 the following day). OPEN-FIELD AND LIGHT–DARK BOX TESTS Mice were moved to an isolated room the

evening before open-field testing to minimize stress on the day of testing. Immediately following 4-h tube restraint or 1 h following exogenous GDF15 injection (5 nM kg−1 IP) or 1 h

following exogenous adrenaline treatment (0.5 mg kg−1 IP), mice were placed in the centre of a box (40 × 40 × 40 cm, black walls and white plastic bottom). Mice remained in the box in the

same isolated room, and videos were recorded using a camera (GoPro, 1080p resolution and a sample rate of 3 frames per second) mounted to the ceiling. Videos were recorded for 20 min. The

two-dimensional mouse pose was analysed using DeepLabCut94 to extract the locations of arena boundaries and anatomical landmarks (nose, left ear, right ear, tail base). The DeepLabCut model

(ResNet50 architecture) was trained using 20 distinct frames (_k_-means clustering) from 20 input videos for a total of 200,000 epochs, resulting in a mean testing error of 3.74 pixels

across all tracked positional landmarks. Weighted spatial means of all four anatomical points taken to represent the location of the centre of the head were then analysed using custom MATLAB

(v2022b) scripts for the time spent in the centre (>8 cm from the arena boundaries), time spent in the periphery (<8 cm from one arena boundary), time spent in the corners (<8 cm

from two arena boundaries) and total distance moved during the test. Afterwards, the mice were returned to their home cages, and the boxes were thoroughly cleaned and left to dry before

subsequent rounds. The boxes in which groups (that is, WT/GFRAL, control/restraint, control/GDF15, etc.) were placed were systematically rotated to control for any possible differences in

box positioning during recording. Alternatively, mice were placed into an automated open-field test system (Opto-Varimex-5 Auto-Track, Columbus Instruments) for 20 min, and the time spent in

the centre and periphery and the total movement were analysed using equipment software. For light–dark box testing, mice were placed into the ‘light’ area and videos were recorded using a

camera (GoPro, 1080p resolution and a sample rate of 3 frames per second) for 10 min. The videos were analysed manually by a blinded assessor for the time spent by the mice in the light

area. BEHAVIOURAL AND METABOLIC ACTIVITY Metabolic and behavioural monitoring was conducted using the Promethion system (Sable Systems International). Data were collected following

acclimation to the system for 24 h. After acclimation, mice were injected with adrenaline (0.5 mg kg−1 IP) at 09:00, after which the mice remained in the system for another 24 h. Data on

food intake, physical activity and nonambulatory movements (beam breaks), oxygen consumption (VO2), carbon dioxide production (VCO2), respiratory exchange ratio (VCO2/VO2) and energy

expenditure (kcal h−1) were collected at 15-min intervals. STATISTICS Statistical analyses were performed using GraphPad Prism (version 10) or R software (RNA-seq). Data were analysed using

an unpaired _t_ test, a one-way analysis of variance (ANOVA) or a two-way ANOVA. When significant interactions were detected through ANOVA, subsequent post hoc analyses were reported with

appropriate corrections; where no post hoc correction is stated, there was no interaction detected. Correlational analyses were performed with Pearson’s correlation. No statistical methods

were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications18,26. Normality was tested using the Shapiro–Wilk test. Outliers were

defined as points outside 2 s.d. away from the group mean. All data are presented as mean ± s.e.m. with individual data points. Differences were considered significant when _P_ < 0.05.

REPORTING SUMMARY Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. DATA AVAILABILITY Data that support the findings of

this study are available from the corresponding author on request. Bulk RNA-seq data have been deposited under accession code GSE267183. scRNA-seq data were obtained from the National Center

for Biotechnology Information Sequence Read Archive reference number SRP145475. Source data are provided with this paper. CODE AVAILABILITY The code for RNA-seq analysis and 2SMR is

available on request. REFERENCES * Kessler, R. C., Chiu, W. T., Demler, O., Merikangas, K. R. & Walters, E. E. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the

National Comorbidity Survey Replication. _Arch. Gen. Psychiatry_ 62, 617–627 (2005). Article  PubMed  PubMed Central  Google Scholar  * Ulrich-Lai, Y. M. & Herman, J. P. Neural

regulation of endocrine and autonomic stress responses. _Nat. Rev. Neurosci._ 10, 397–409 (2009). Article  CAS  PubMed  PubMed Central  Google Scholar  * McCall, J. G. et al. CRH engagement

of the locus coeruleus noradrenergic system mediates stress-induced anxiety. _Neuron_ 87, 605–620 (2015). Article  CAS  PubMed  PubMed Central  Google Scholar  * McCall, J. G. et al. Locus

coeruleus to basolateral amygdala noradrenergic projections promote anxiety-like behavior. _eLife_ 6, e18247 (2017). Article  PubMed  PubMed Central  Google Scholar  * Burns, T. W., Mohs, J.

M., Langley, P. E., Yawn, R. & Chase, G. R. Regulation of human lipolysis. In vivo observations on the role of adrenergic receptors. _J. Clin. Invest._ 53, 338–341 (1974). Article  CAS

  PubMed  PubMed Central  Google Scholar  * Elmadjian, F., Hope, J. M. & Lamson, E. T. Excretion of epinephrine and norepinephrine in various emotional states. _J. Clin. Endocrinol.

Metab._ 17, 608–620 (1957). Article  CAS  PubMed  Google Scholar  * Pollin, W. & Goldin, S. The physiological and psychological effects of intravenously administered epinephrine and its

metabolism in normal and schizophrenic men. II. Psychiatric observations. _J. Psychiatr. Res._ 1, 50–67 (1961). Article  CAS  PubMed  Google Scholar  * Clayton, E. C. & Williams, C. L.

Noradrenergic receptor blockade of the NTS attenuates the mnemonic effects of epinephrine in an appetitive light–dark discrimination learning task. _Neurobiol. Learn. Mem._ 74, 135–145

(2000). Article  CAS  PubMed  Google Scholar  * King, S. O. 2nd & Williams, C. L. Novelty-induced arousal enhances memory for cued classical fear conditioning: interactions between

peripheral adrenergic and brainstem glutamatergic systems. _Learn. Mem._ 16, 625–634 (2009). Article  CAS  PubMed  Google Scholar  * Gottschalk, L. A., Stone, W. N. & Gleser, G. C.

Peripheral versus central mechanisms accounting for antianxiety effects of propranolol. _Psychosom. Med._ 36, 47–56 (1974). Article  CAS  PubMed  Google Scholar  * Stone, W. N., Gleser, G.

C. & Gottschalk, L. A. Anxiety and β-adrenergic blockade. _Arch. Gen. Psychiatry_ 29, 620–622 (1973). Article  CAS  PubMed  Google Scholar  * Flemenbaum, A. Cyclic AMP, alpha and beta

receptors as a further explanation of the propranolol–FFA activity. _Psychosom. Med._ 37, 74 (1975). Article  CAS  PubMed  Google Scholar  * Gottschalk, L. A., Stone, W. N., Gleser, G. C.

& Iacono, J. M. Anxiety levels in dreams: relation to changes in plasma free fatty acids. _Science_ 153, 654–657 (1966). Article  CAS  PubMed  Google Scholar  * Bonn, J. A., Turner, P.

& Hicks, D. C. Beta-adrenergic-receptor blockade with practolol in treatment of anxiety. _Lancet_ 1, 814–815 (1972). Article  CAS  PubMed  Google Scholar  * Wang, D. et al. GDF15:

emerging biology and therapeutic applications for obesity and cardiometabolic disease. _Nat. Rev. Endocrinol._ 17, 592–607 (2021). Article  CAS  PubMed  Google Scholar  * Lockhart, S. M.,

Saudek, V. & O’Rahilly, S. GDF15: a hormone conveying somatic distress to the brain. _Endocr. Rev._ 41, bnaa007 (2020). Article  PubMed  PubMed Central  Google Scholar  * Yang, L. et al.

GFRAL is the receptor for GDF15 and is required for the anti-obesity effects of the ligand. _Nat. Med._ 23, 1158–1166 (2017). Article  CAS  PubMed  Google Scholar  * Day, E. A. et al.

Metformin-induced increases in GDF15 are important for suppressing appetite and promoting weight loss. _Nat. Metab._ 1, 1202–1208 (2019). Article  CAS  PubMed  Google Scholar  * Borner, T.

et al. GDF15 induces an aversive visceral malaise state that drives anorexia and weight loss. _Cell Rep._ 31, 107543 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Worth, A.

A. et al. The cytokine GDF15 signals through a population of brainstem cholecystokinin neurons to mediate anorectic signalling. _eLife_ 9, e55164 (2020). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Gil, C. I. et al. Mitochondrial stress-induced GFRAL signaling controls diurnal food intake and anxiety-like behavior. _Life Sci. Alliance_ 5, e202201495 (2022).

Article  CAS  Google Scholar  * Low, J. K. et al. First behavioural characterisation of a knockout mouse model for the transforming growth factor (TGF)-β superfamily cytokine, MIC-1/GDF15.

_PLoS ONE_ 12, e0168416 (2017). Article  PubMed  PubMed Central  Google Scholar  * Cimino, I. et al. Activation of the hypothalamic–pituitary–adrenal axis by exogenous and endogenous GDF15.

_Proc. Natl Acad. Sci. USA_ 118, e2106868118 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  * Ruud, L. E. et al. Activation of GFRAL+ neurons induces hypothermia and

glucoregulatory responses associated with nausea and torpor. _Cell Rep._ 43, 113960 (2024). Article  PubMed  Google Scholar  * Sjøberg, K. A. et al. GDF15 increases insulin action in the

liver and adipose tissue via a β-adrenergic receptor-mediated mechanism. _Cell Metab._ 35, 1327–1340 (2023). Article  PubMed  Google Scholar  * Wang, D. et al. GDF15 promotes weight loss by

enhancing energy expenditure in muscle. _Nature_ 619, 143–150 (2023). Article  CAS  PubMed  PubMed Central  Google Scholar  * Stone, W. N., Gleser, G. C., Gottschalk, L. A. & Iacono, J.

M. Stimulus, affect, and plasma free fatty acid. _Psychosom. Med._ 31, 331–341 (1969). Article  CAS  PubMed  Google Scholar  * Williams, T. & Williams, C. L. Peripheral arousal-related

hormones modulate norepinephrine release in the hippocampus via influences on brainstem nuclei. _Behav. Brain Res._ 153, 87–95 (2004). Article  PubMed  Google Scholar  * Markussen, L. K. et

al. Lipolysis regulates major transcriptional programs in brown adipocytes. _Nat. Commun._ 13, 3956 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  * Bruford, E. A. et al.

Guidelines for human gene nomenclature. _Nat. Genet._ 52, 754–758 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Qing, H. et al. Origin and function of stress-induced IL-6

in murine models. _Cell_ 182, 1660 (2020). Article  CAS  PubMed  Google Scholar  * Karusheva, Y. et al. The common H202D variant in GDF-15 does not affect its bioactivity but can

significantly interfere with measurement of its circulating levels. _J. Appl. Lab. Med._ 7, 1388–1400 (2022). Article  PubMed  Google Scholar  * Jeong, K. H. et al. Impaired basal and

restraint-induced epinephrine secretion in corticotropin-releasing hormone-deficient mice. _Endocrinology_ 141, 1142–1150 (2000). Article  CAS  PubMed  Google Scholar  * Lafontan, M. &

Berlan, M. Fat cell adrenergic receptors and the control of white and brown fat cell function. _J. Lipid Res._ 34, 1057–1091 (1993). Article  CAS  PubMed  Google Scholar  * Patel, A. R. et

al. LPS induces rapid increase in GDF15 levels in mice, rats, and humans but is not required for anorexia in mice. _Am. J. Physiol. Gastrointest. Liver Physiol._ 322, G247–G255 (2022).

Article  CAS  PubMed  Google Scholar  * Haemmerle, G. et al. Defective lipolysis and altered energy metabolism in mice lacking adipose triglyceride lipase. _Science_ 312, 734–737 (2006).

Article  CAS  PubMed  Google Scholar  * Zhou, C. et al. Sympathetic overdrive and unrestrained adipose lipolysis drive alcohol-induced hepatic steatosis in rodents. _Mol. Metab._ 78, 101813

(2023). Article  CAS  PubMed  PubMed Central  Google Scholar  * Mayer, N. et al. Development of small-molecule inhibitors targeting adipose triglyceride lipase. _Nat. Chem. Biol._ 9, 785–787

(2013). Article  CAS  PubMed  Google Scholar  * L’Homme, L. et al. Adipose tissue macrophage infiltration and hepatocyte stress increase GDF-15 throughout development of obesity to MASH.

_Nat. Commun._ 15, 7173 (2024). Article  PubMed  PubMed Central  Google Scholar  * Burl, R. B. et al. Deconstructing adipogenesis induced by β3-adrenergic receptor activation with

single-cell expression profiling. _Cell Metab._ 28, 300–309 (2018). Article  CAS  PubMed  PubMed Central  Google Scholar  * Caputa, G., Castoldi, A. & Pearce, E. J. Metabolic adaptations

of tissue-resident immune cells. _Nat. Immunol._ 20, 793–801 (2019). Article  CAS  PubMed  Google Scholar  * Morgan, P. K. et al. Macrophage polarization state affects lipid composition and

the channeling of exogenous fatty acids into endogenous lipid pools. _J. Biol. Chem._ 297, 101341 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  * Abe, I. et al.

Lipolysis-derived linoleic acid drives beige fat progenitor cell proliferation. _Dev. Cell_ 57, 2623–2637 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  * Townsend, L. K. et

al. High-saturated-fat diet-induced obesity causes hepatic interleukin-6 resistance via endoplasmic reticulum stress. _J. Lipid Res._ 60, 1236–1249 (2019). Article  CAS  PubMed  PubMed

Central  Google Scholar  * Townsend, L. K. et al. AMPK mediates energetic stress-induced liver GDF15. _FASEB J._ 35, e21218 (2021). Article  CAS  PubMed  Google Scholar  * Kratz, M. et al.

Metabolic dysfunction drives a mechanistically distinct proinflammatory phenotype in adipose tissue macrophages. _Cell Metab._ 20, 614–625 (2014). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Lu, J. F. et al. GDF15 is a major determinant of ketogenic diet-induced weight loss. _Cell Metab._ 35, 2165–2182 (2023). Article  CAS  PubMed  Google Scholar  * Fulton, S.,

Décarie-Spain, L., Fioramonti, X., Guiard, B. & Nakajima, S. The menace of obesity to depression and anxiety prevalence. _Trends Endocrinol. Metab._ 33, 18–35 (2022). Article  CAS 

PubMed  Google Scholar  * Wang, D. et al. Fatty acids increase GDF15 and reduce food intake through a GFRAL signaling axis. _Diabetes_ 73, 51–56 (2024). Article  CAS  PubMed  Google Scholar

  * Füzesi, T. et al. Hypothalamic CRH neurons orchestrate complex behaviours after stress. _Nat. Commun._ 7, 11937 (2016). Article  PubMed  PubMed Central  Google Scholar  * Bales, M. B. et

al. High fat diet blunts stress-induced hypophagia and activation of Glp1r dorsal lateral septum neurons in male but not in female mice. _Mol. Metab._ 64, 101571 (2022). Article  CAS 

PubMed  PubMed Central  Google Scholar  * Holt, M. K. et al. Preproglucagon neurons in the nucleus of the solitary tract are the main source of brain GLP-1, mediate stress-induced

hypophagia, and limit unusually large intakes of food. _Diabetes_ 68, 21–33 (2019). Article  CAS  PubMed  Google Scholar  * Fortin, S. M. et al. The locus coeruleus contributes to the

anorectic, nausea, and autonomic physiological effects of glucagon-like peptide-1. _Sci. Adv._ 9, eadh0980 (2023). Article  CAS  PubMed  PubMed Central  Google Scholar  * Hsu, J.-Y. et al.

Non-homeostatic body weight regulation through a brainstem-restricted receptor for GDF15. _Nature_ 550, 255–259 (2017). Article  PubMed  Google Scholar  * Niu, M. et al. Claustrum mediates

bidirectional and reversible control of stress-induced anxiety responses. _Sci. Adv._ 8, eabi6375 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  * Calhood, G. G. & Tye, K.

M. Resolving the neural circuits of anxiety. _Nat. Neurosci._ 18, 1394–1404 (2015). Article  Google Scholar  * Chandler, D. J. et al. Redefining noradrenergic neuromodulation of behavior:

impacts of a modular locus coeruleus architecture. _J. Neurosci._ 39, 8239–8249 (2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Hirschberg, S., Li, Y., Randall, A., Kremer,

E. J. & Pickering, A. E. Functional dichotomy in spinal- vs prefrontal-projecting locus coeruleus modules splits descending noradrenergic analgesia from ascending aversion and anxiety in

rats. _eLife_ 6, e29808 (2017). Article  PubMed  PubMed Central  Google Scholar  * Kim, S.-Y. et al. Diverging neural pathways assemble a behavioural state from separable features in

anxiety. _Nature_ 496, 219–223 (2013). Article  CAS  PubMed  PubMed Central  Google Scholar  * Cowley, N. A. et al. Dynorphin controls the gain of an amygdalar anxiety circuit. _Cell Rep._

14, 2774–2783 (2016). Article  Google Scholar  * Giplin, N. W., Herman, M. A. & Roberto, M. The central amygdala as an integrative hub for anxiety and alcohol use disorders. _Biol.

Psychiatry_ 77, 859–869 (2015). Article  Google Scholar  * Borodovitsyna, O., Duffy, B. C., Pickering, A. E. & Chandler, D. J. Anatomically and functionally distinct locus coeruleus

efferents mediate opposing effects on anxiety-like behavior. _Neurobiol. Stress_ 13, 100284 (2020). Article  PubMed  PubMed Central  Google Scholar  * Sun, Y., Qian, L., Xu, L., Hunt, S.

& Sah, P. Somatostatin neurons in the central amygdala mediate anxiety by disinhibition of the central sublenticular extended amygdala. _Mol. Psychiatry_ 28, 4163–4174 (2023). Article 

PubMed  Google Scholar  * Sabatini, P. V. et al. GFRAL-expressing neurons suppress food intake via aversive pathways. _Proc. Natl Acad. Sci. USA_ 118, e2021357118 (2021). Article  CAS 

PubMed  PubMed Central  Google Scholar  * O’Rahilly, S. GDF15—from biomarker to allostatic hormone. _Cell Metab._ 26, 807–808 (2017). Article  PubMed  Google Scholar  * Patel, S. et al.

GDF15 provides an endocrine signal of nutritional stress in mice and humans. _Cell Metab._ 29, 707–718 (2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Florsheim, E. B. et al.

Immune sensing of food allergens promotes avoidance behaviour. _Nature_ 620, 643–650 (2023). Article  CAS  PubMed  PubMed Central  Google Scholar  * Luan, H. H. et al. GDF15 is an

inflammation-induced central mediator of tissue tolerance. _Cell_ 178, 1231–1244 (2019). Article  CAS  PubMed  PubMed Central  Google Scholar  * Borner, T. et al. GDF15 induces anorexia

through nausea and emesis. _Cell Metab._ 31, 351–362 (2020). Article  CAS  PubMed  PubMed Central  Google Scholar  * Lönnqvist, F., Wennlund, A., Wahrenberg, H. & Arner, P. Effects of

mental stress on lipolysis in humans. _Metabolism_ 41, 622–630 (1992). Article  PubMed  Google Scholar  * Klein, A. B. et al. Pharmacological but not physiological GDF15 suppresses feeding

and the motivation to exercise. _Nat. Commun._ 12, 1041 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  * Macia, L. et al. Macrophage inhibitory cytokine 1 (MIC-1/GDF15)

decreases food intake, body weight and improves glucose tolerance in mice on normal & obesogenic diets. _PLoS ONE_ 7, e34868 (2012). Article  CAS  PubMed  PubMed Central  Google Scholar

  * Benichou, O. et al. Discovery, development, and clinical proof of mechanism of LY3463251, a long-acting GDF15 receptor agonist. _Cell Metab._ 35, 274–286 (2023). Article  CAS  PubMed 

Google Scholar  * Zhang, Y. et al. Activity-balanced GLP-1/GDF15 dual agonist reduces body weight and metabolic disorder in mice and non-human primates. _Cell Metab._ 35, 287–298 (2023).

Article  PubMed  Google Scholar  * Deng, Y.-T. et al. Atlas of the plasma proteome in health and disease in 53,026 adults. _Cell_ 188, 253–271 (2025). Article  CAS  PubMed  Google Scholar  *

Patsalos, A. et al. A growth factor-expressing macrophage subpopulation orchestrates regenerative inflammation via GDF-15. _J. Exp. Med._ 219, e20210420 (2022). Article  CAS  PubMed  Google

Scholar  * Irwin, M. R. & Cole, S. W. Reciprocal regulation of the neural and innate immune systems. _Nat. Rev. Immunol._ 11, 625–632 (2011). Article  CAS  PubMed  PubMed Central 

Google Scholar  * Glaser, R. & Kiecolt-Glaser, J. K. Stress-induced immune dysfunction: implications for health. _Nat. Rev. Immunol._ 5, 243–251 (2005). Article  CAS  PubMed  Google

Scholar  * Le Thuc, O. & García-Cáceres, C. Obesity-induced inflammation: connecting the periphery to the brain. _Nat. Metab._ 6, 1237–1252 (2024). Article  PubMed  Google Scholar  *

Bachman, E. S. et al. βAR signaling required for diet-induced thermogenesis and obesity resistance. _Science_ 297, 843–845 (2002). Article  CAS  PubMed  Google Scholar  * Kaur, S. et al.

Adipose-specific ATGL ablation reduces burn injury-induced metabolic derangements in mice. _Clin. Transl. Med._ 11, e417 (2021). Article  CAS  PubMed  PubMed Central  Google Scholar  * Dibe,

H. A., Townsend, L. K., McKie, G. L. & Wright, D. C. Epinephrine responsiveness is reduced in livers from trained mice. _Physiol. Rep._ 8, e14370 (2020). Article  PubMed  PubMed Central

  Google Scholar  * Pedersen, L. et al. Voluntary running suppresses tumor growth through epinephrine- and IL-6-dependent NK cell mobilization and redistribution. _Cell Metab._ 23, 554–562

(2016). Article  CAS  PubMed  Google Scholar  * Townsend, L. K. et al. Loss of glucagon signaling alters white adipose tissue browning. _FASEB J._ 33, 4824–4835 (2019). Article  CAS  PubMed

  Google Scholar  * Medak, K. D., McKie, G. L., Shamshoumn, H., Seguin, I. & Wright, D. C. The glucose lowering effects of CL 316,243 dissipate with repeated use and are rescued

bycilostamide. _Physiol. Rep._ 10, e15187 (2022). Article  CAS  PubMed  PubMed Central  Google Scholar  * Muglia, L. J. et al. Corticotropin-releasing hormone links pituitary

adrenocorticotropin gene expression and release during adrenal insufficiency. _J. Clin. Invest._ 105, 1269–1277 (2000). Article  CAS  PubMed  PubMed Central  Google Scholar  * Laryea, G.,

Schütz, G. & Muglia, L. J. Disrupting hypothalamic glucocorticoid receptors causes HPA axis hyperactivity and excess adiposity. _Mol. Endocrinol._ 27, 1655–1665 (2013). Article  CAS 

PubMed  PubMed Central  Google Scholar  * Mottillo, E. P. et al. Lack of adipocyte AMPK exacerbates insulin resistance and hepatic steatosis through brown and beige adipose tissue function.

_Cell Metab._ 24, 118–129 (2016). Article  CAS  PubMed  PubMed Central  Google Scholar  * Galic, S. et al. Hematopoietic AMPK β1 reduces mouse adipose tissue macrophage inflammation and

insulin resistance in obesity. _J. Clin. Invest._ 121, 4903–4915 (2011). Article  CAS  PubMed  PubMed Central  Google Scholar  * Hemani, G. et al. The MR-Base platform supports systematic

causal inference across the human phenome. _eLife_ 7, e34408 (2018). Article  PubMed  PubMed Central  Google Scholar  * Taliun, S. A. G. & Evans, D. M. Ten simple rules for conducting a

Mendelian randomization study. _PLoS Comput. Biol._ 17, e1009238 (2021). Article  Google Scholar  * Sanderson, E. et al. Mendelian randomization. _Nat. Rev. Methods Primers_ 2, 6 (2022).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Liu, Y. et al. Genetic architecture of 11 organ traits derived from abdominal MRI using deep learning. _eLife_ 10, e65554 (2021).

Article  CAS  PubMed  PubMed Central  Google Scholar  * Nath, T. et al. Using DeepLabCut for 3D markerless pose estimation across species and behaviors. _Nat. Protoc._ 14, 2152–2176 (2019).

Article  CAS  PubMed  Google Scholar  Download references ACKNOWLEDGEMENTS G.R.S. acknowledges the support of a Diabetes Canada Investigator Award (OG-3-22-5645-GS), a Canadian Institutes of

Health Research (CIHR) Foundation Grant (201709FDN-CEBA-116200), a Tier 1 Canada Research Chair in Metabolic Diseases and a J. Bruce Duncan Endowed Chair in Metabolic Diseases. L.K.T.

acknowledges the support of a CIHR Post-doctoral Fellowship Award, a Michael DeGroote Fellowship Award in Basic Biomedical Science and a MITACS fellowship supported by Novo Nordisk. D.W.

acknowledges the support of fellowship grants from the McMaster Institute for Research on Aging (MIRA) at McMaster University. E.M.D. acknowledges the support of a Vanier Canada Graduate

Scholarship. A.M. was supported by the Natural Sciences and Engineering Research Council of Canada. D.P.B. acknowledges the support of the GSK Chair in Diabetes of Université de Sherbrooke

and an FRQS J1 salary award. A.C.C. acknowledges the support of a Tier 1 Canada Research Chair in Molecular Imaging of Diabetes and a CIHR Grant (no. 299962). J.G.M. acknowledges the support

of a National Institutes of Health award (R01NS117899). AUTHOR INFORMATION AUTHORS AND AFFILIATIONS * Centre for Metabolism, Obesity and Diabetes Research, McMaster University, Hamilton,

Ontario, Canada Logan K. Townsend, Dongdong Wang, Russta Fayyazi, Evangelia E. Tsakiridis, Eric M. Desjardins, Zeel Patel, Celina M. Valvano, Junfeng Lu, Alice E. Payne, Ofure Itua, Daniel

M. Marko, Jonathan D. Schertzer, Katherine M. Morrison & Gregory R. Steinberg * Division of Endocrinology and Metabolism, Department of Medicine, McMaster University, Hamilton, Ontario,

Canada Logan K. Townsend, Dongdong Wang, Russta Fayyazi, Evangelia E. Tsakiridis, Eric M. Desjardins, Zeel Patel, Celina M. Valvano, Junfeng Lu, Alice E. Payne & Gregory R. Steinberg *

Sunnybrook Research Institute, University of Toronto, Toronto, Ontario, Canada Carly M. Knuth * Department of Health Science, Brock University, St. Catherines, Ontario, Canada Ahmad Mohammad

 & Rebecca E. K. MacPherson * Department of Anesthesiology, Washington University in St. Louis, St. Louis, MO, USA Léa J. Becker & Jordan G. McCall * Department of Biochemistry and

Biomedical Sciences, McMaster University, Hamilton, Ontario, Canada Ofure Itua, Daniel M. Marko, Jonathan D. Schertzer & Gregory R. Steinberg * Human Health and Nutritional Science,

University of Guelph, Guelph, Ontario, Canada Kyle D. Medak * Farncombe Family Digestive Health Research Institute, McMaster University, Hamilton, Ontario, Canada Daniel M. Marko & 

Jonathan D. Schertzer * School of Kinesiology, University of British Columbia, Vancouver, British Columbia, Canada David C. Wright * British Columbia Children’s Hospital Research Institute,

Vancouver, British Columbia, Canada David C. Wright * Faculty of Land and Food Systems, University of British Columbia, Vancouver, British Columbia, Canada David C. Wright * Department of

Kinesiology, Brock University, St. Catherines, Ontario, Canada Shawn M. Beaudette * Department of Pediatrics, McMaster University, Hamilton, Ontario, Canada Katherine M. Morrison *

Department of Medicine, Faculty of Medicine and Health Sciences, Université de Sherbrooke, Centre de recherche du Centre hospitalier universitaire de Sherbrooke, Sherbrooke, Quebec, Canada

André C. Carpentier & Denis P. Blondin * David Braley Cardiac, Vascular and Stroke Research Institute, Hamilton, Ontario, Canada Marc G. Jeschke * Hamilton General Hospital, Hamilton

Health Sciences, Hamilton, Ontario, Canada Marc G. Jeschke * Department of Surgery, McMaster University, Hamilton, Ontario, Canada Marc G. Jeschke Authors * Logan K. Townsend View author

publications You can also search for this author inPubMed Google Scholar * Dongdong Wang View author publications You can also search for this author inPubMed Google Scholar * Carly M. Knuth

View author publications You can also search for this author inPubMed Google Scholar * Russta Fayyazi View author publications You can also search for this author inPubMed Google Scholar *

Ahmad Mohammad View author publications You can also search for this author inPubMed Google Scholar * Léa J. Becker View author publications You can also search for this author inPubMed 

Google Scholar * Evangelia E. Tsakiridis View author publications You can also search for this author inPubMed Google Scholar * Eric M. Desjardins View author publications You can also

search for this author inPubMed Google Scholar * Zeel Patel View author publications You can also search for this author inPubMed Google Scholar * Celina M. Valvano View author publications

You can also search for this author inPubMed Google Scholar * Junfeng Lu View author publications You can also search for this author inPubMed Google Scholar * Alice E. Payne View author

publications You can also search for this author inPubMed Google Scholar * Ofure Itua View author publications You can also search for this author inPubMed Google Scholar * Kyle D. Medak

View author publications You can also search for this author inPubMed Google Scholar * Daniel M. Marko View author publications You can also search for this author inPubMed Google Scholar *

Jonathan D. Schertzer View author publications You can also search for this author inPubMed Google Scholar * David C. Wright View author publications You can also search for this author

inPubMed Google Scholar * Shawn M. Beaudette View author publications You can also search for this author inPubMed Google Scholar * Katherine M. Morrison View author publications You can

also search for this author inPubMed Google Scholar * André C. Carpentier View author publications You can also search for this author inPubMed Google Scholar * Denis P. Blondin View author

publications You can also search for this author inPubMed Google Scholar * Rebecca E. K. MacPherson View author publications You can also search for this author inPubMed Google Scholar *

Jordan G. McCall View author publications You can also search for this author inPubMed Google Scholar * Marc G. Jeschke View author publications You can also search for this author inPubMed 

Google Scholar * Gregory R. Steinberg View author publications You can also search for this author inPubMed Google Scholar CONTRIBUTIONS The overall conceptualization of studies included in

this work was done by L.K.T. and G.R.S. Mouse experiments were designed, managed, performed and analysed by L.K.T., D.W., C.M.K., K.D.M., D.M.M., E.M.D., Z.P., E.E.T., J.L., A.E.P., O.I. and

C.M.V. RNA-seq analyses were performed by D.W. and R.F. Mouse behaviour studies were done with the assistance of A.M., S.M.B. and R.E.K.M. c-Fos staining experiments were designed, managed,

performed and analysed by L.J.B. and J.G.M. Adipocyte and macrophage experiments were completed by L.K.T., E.E.T., E.M.D., Z.P. and C.M.V. L.K.T. and G.R.S. wrote the paper, with the

assistance of J.G.M., M.G.J., A.C.C., D.P.B., K.M.M., J.D.S. and D.C.W. The paper was reviewed, edited and approved by all the authors. CORRESPONDING AUTHOR Correspondence to Gregory R.

Steinberg. ETHICS DECLARATIONS COMPETING INTERESTS G.R.S. is a cofounder and shareholder of Espervita Therapeutics, a company developing new medications for fibrosis and cancer. McMaster

University has received funding from Cambrian Biosciences, Catalym, Espervita Therapeutics, Esperion Therapeutics, Merck, Nestle, Novo Nordisk and Poxel Pharmaceuticals for research

conducted in the laboratory of G.R.S. G.R.S. has received consulting and speaking fees from AstraZeneca, CurieBio, Eli Lilly, Esperion Therapeutics, Korro Bio, Keros Therapeutics, Merck,

Novo Nordisk, Versant Ventures and Poxel Pharmaceuticals. A.C.C. received consulting and speaking fees from Eli Lilly, HLS Therapeutics, Janssen, Novartis Pharmaceuticals Canada and Novo

Nordisk Canada. The other authors declare no competing interests. PEER REVIEW PEER REVIEW INFORMATION _Nature Metabolism_ thanks Mónica Moreira-Rodrigues and the other, anonymous,

reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Christoph Schmitt, in collaboration with the _Nature Metabolism_ team. ADDITIONAL INFORMATION

PUBLISHER’S NOTE Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. EXTENDED DATA EXTENDED DATA FIG. 1 SEX- AND HOUSING

TEMPERATURE-INDEPENDENT EFFECTS OF EXOGENOUS EPINEPHRINE ON GDF15. A) Pathway analysis from RNA-sequencing of gWAT following 1-hr epinephrine treatment. _n_ = _6_ per group. B) Heat map

showing ligand-encoded genes from RNA-sequencing of gWAT following 1-hr epinephrine treatment. _n_ = _6_ per group. C) Circulating GDF15 from age-matched male (saline _n_ = _9_, epinephrine

_n_ = _9_) and female mice (saline _n_ = _10_, epinephrine _n_ = _11_) following 1-hr epinephrine with fold-change relative to baseline levels (inset). Data presented as mean ± s.e.m. with

p-values calculated using 2-way ANOVA with post-hoc test and Tukey’s correction and unpaired two-tail t-test, respectively. D) Time-course of circulating GDF15 post-saline (RT _n_ = _5_, TN

_n_ = _4_) or epinephrine (RT _n_ = _5_, TN _n_ = _5_) in mice housed at room temperature (RT ~ 22 °C) or thermoneutrality (TN ~ 29 °C) for 4 weeks. n = 4-5/group. Data presented as mean ±

s.e.m. with p-values calculated using 2-way ANOVA at each time point. E) Serum epinephrine in control (_n_ = _9_ mice), 1-hr post IP epinephrine (_n_ = _7_), and 4-hr physical restraint (_n_

 = _8_). Data presented as mean ± s.e.m. with p-values calculated using 1-way ANOVA with post-hoc test and Tukey’s correction. Source data EXTENDED DATA FIG. 2 CHRONIC PSYCHOLOGICAL STRESS

NOT ASSOCIATED WITH ELEVATED CIRCULATING GDF15 LEVELS IN HUMANS. A) Serum GDF15 levels in children with overweight and obesity with (_n_ = _23_) or without (_n_ = _24_) diagnosis of anxiety.

Data presented as mean ± s.e.m. with p-values calculated using unpaired two-tail t-test. B) Scatter plot of the SNP-effect on GDF15 and single nucleotide polymorphism (SNP)-effect on

nervous anxiety tension or depression in humans by using two sample Mendelian Randomization (2SMR). Data presented as mean ± error bars indicate 95% CI, n = 407,746 participants in UK

Biobank. MR analysis was performed by using Simple median method, MR weighted mode estimator, Weighted median method, MR Egger regression, Inverse variance weighted methods. C) Single SNP

analysis of GDF15 on anxiety and depression in humans, Data presented as mean ± error bars indicate 95% CI, n = 407,746 participants in UK Biobank. Source data EXTENDED DATA FIG. 3 EFFECTS

OF LPS IN BETA RECEPTOR KNOCKOUT MICE. A) Serum GDF15 from WT and β-adrenergic receptor knockout (BR-/-) mice following 2-hr LPS treatment (WT _n_ = _6_, BR-/- _n_ = _6_) or control (WT _n_ 

= _6_, BR-/- _n_ = _8_). Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. B) Blood glucose from WT and BR-/- mice following 2-hr LPS treatment (WT _n_ = _6_, BR-/-

_n_ = _6_) or control (WT _n_ = _6_, BR-/- _n_ = _8_). Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. Source data EXTENDED DATA FIG. 4 RESTRAINT STRESS IN

ADATGLFLOX/FLOX AND ADATGL-/- MICE. A) Blood glucose levels in AdATGLflox/flox mcie (control _n_ = _6_, restraint _n_ = _6_) and AdATGL-/- (control _n_ = _13_, restraint _n_ = _11_). Data

presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA with post-hoc test and Tukey’s correction. B) Total distance for AdATGLflox/flox mice (control _n_ = _6_, restraint _n_ 

= _6_) and AdATGL-/- (control _n_ = _13_, restraint _n_ = _11_) during open-field test. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA with post-hoc test and

Tukey’s correction. C) Time in the centre for AdATGLflox/flox mice (control _n_ = _6_, restraint _n_ = _6_) and AdATGL-/- (control _n_ = _13_, restraint _n_ = _11_) during open-field test.

Data presented as mean ± s.e.m. with no statistical test performed. Source data EXTENDED DATA FIG. 5 MACROPHAGES ARE THE MOST LIKELY SOURCE OF GDF15 WITHIN ADIPOSE TISSUE. A) _Fgf21_

expression in mouse adipocyte and SVF from gWAT post-saline (_n_ = _4_ per group) or epinephrine (_n_ = _5_ per group). Data presented as mean ± s.e.m. with p-values calculated using 2-way

ANOVA with post-hoc test and Tukey’s correction. B) _Leptin_ expression in mouse adipocyte and SVF from gWAT post-saline (_n_ = _3_ per group) or epinephrine (_n_ = _4_ per group). Data

presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. C) t-SNE plot of Lin- stromal vascular cells from gWAT of control mice and mice treated with CL for 3 days. Clustering

identified 6 major cell types/states. Clusters are highlighted in different colors. The data set was queried for cells expressing _Gdf15_. Heatmap shows expression of _Gdf15_ and other

cell-identifying factors in the various identified cell populations. D) _F480/Adgre1_ expression in mouse F480+ and F480- fractions of SVF from gWAT 1-hr post-epinephrine treatment (0.5 

mg/kg). Data presented as mean ± s.e.m. with _n_ = _3_ per group. p-values calculated using 2-way ANOVA. E) _Arg1_ and _Nos2_ expression in bone marrow-derived macrophages (BMDMs) polarized

to either M1-like or M2-like. _n_ = _6_ per group. Individual data points represent duplicates from 3 independent experiments. Data presented as mean ± s.e.m. with p-values calculated using

1-way ANOVA with post-hoc test and Tukey’s correction. F) GDF15 levels in media from BMDMs polarized to either M1-like or M2-like for 24-hrs. _n_ = _3_ per group. Individual data points

represent triplicates from 3 independent experiments. Data presented as mean ± s.e.m. with p-values calculated using 1-way ANOVA with post-hoc test and Tukey’s correction. G) GDF15 levels in

media from M2-like BMDMs treated with epinephrine (1 μM) or tunicamycin (5 ng/mL) for 24-hrs. _n_ = _3_ per group. Individual data points represent triplicates from 3 independent

experiments. Data presented as mean ± s.e.m. with p-values calculated using 1-way ANOVA with post-hoc test and Tukey’s correction. H) Volcano plot showing fatty acid transporters identified

between M1- and M2-like macrophage populations from scRNA-seq data. Source data EXTENDED DATA FIG. 6 LIPIDS AND ANXIETY. A) Time in the centre during open-field test (% total) following 4-hr

palm oil gavage (10 mL/kg) gavage in WT and _Gfral_ -/- mice. _n_ = _8_ per group. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. B) Time in the centre during

open-field test for chow (WT _n_ = _6_ mice, _Gfral_-/- _n_ = _6) or_ 4-week high-fat diet fed mice (WT _n_ = _7_ mice, _Gfral_-/- _n_ = _6)_. Data presented as mean ± s.e.m. with p-values

calculated using 2-way ANOVA. C) Total distance traveled during open-field test for chow (WT _n_ = _6_ mice, _Gfral_-/- _n_ = _6) or_ 4-week high-fat diet fed mice (WT _n_ = _7_ mice,

_Gfral_-/- _n_ = _6)_. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. Source data EXTENDED DATA FIG. 7 GDF15-GFRAL SIGNALLING IS IMPORTANT FOR BEHAVIOURAL

RESPONSES TO EPINEPHRINE. A) Non-ambulatory movements, B) Total physical activity, C) Total food intake, D) respiratory exchange ratio (RER), E) Energy expenditure in WT (saline _n_ = _4_,

epinephrine _n_ = _4_) and _Gfral_-/- mice (saline _n_ = _4_, epinephrine _n_ = _4_). Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA with post-hoc test and

Tukey’s correction. F) Chow intake (18-hrs) following tube restraint in WT (control _n_ = _7_, restraint _n_ = _7_) and _Gfral_ -/- mice (control _n_ = _10_, restraint _n_ = _9_). Data

presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. G) Kaolin clay intake (18-hrs) following tube restraint in WT (control _n_ = _7_, restraint _n_ = _7_) and _Gfral_ -/-

mice (control _n_ = _10_, restraint _n_ = _9_). Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. H) Time in center and total distance during open-field test

following GDF15 treatment with representative images showing movement of individual mice. _n_ = _9_ per group. Data presented as mean ± s.e.m. with p-values calculated using unpaired

two-tail t-test. I) Time in light during light-dark box test following vehicle (_n_ = _8_) or GDF15 treatment (_n_ = _9_). Data presented as mean ± s.e.m. with p-values calculated using

unpaired two-tail t-test. J) Daily chow food intake throughout repeated GDF15 (5 nM/kg IP). _n_ = _12_ per group. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA

with post-hoc test and Tukey’s correction. K) Time in the center during open-field test following GDF15 treatment (5 nM/kg IP) with representative images of movement of individual mice. _n_ 

= _6_ per group. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. Chr. GDF15: Chronic GDF15. L) Total distance traveled during open-field test following GDF15

treatment (5 nM/kg IP) with representative images showing representative movement of individual mice. _n_ = _6_ per group. Data presented as mean ± s.e.m. with p-values calculated using

2-way ANOVA. Chr. GDF15: Chronic GDF15. Source data EXTENDED DATA FIG. 8 NEITHER THE LOCUS COERULEUS, NOR ITS ANGIOGENIC DOWNSTREAM BRAIN REGIONS, ARE ACTIVATED BY GDF15. A) Quantification

of c-Fos positive cells in the locus coeruleus 90-min following GDF15 treatment (5 nm/kg IP), with representative pictures of the staining for TH and c-Fos. _n_ = _4_ per group. Data

presented as mean ± s.e.m. with p-values calculated using unpaired two-tail t-test. Scale bar=200 µm. 4 V: 4th ventricle, LC: locus coeruleus, TH: tyrosine hydroxylase. B) Quantification of

c-Fos positive cells in the PFC and BLA 90-min following GDF15 treatment (5 nm/kg IP), with representative pictures of the staining. _n_ = _4_ per group. Data presented as mean ± s.e.m. with

p-values calculated using unpaired two-tail t-test. Scale bar=200 µm. BLA: basolateral amygdala, cc: corpus callosum, mPFC: medial prefrontal cortex. C) Serum GDF15 in WT (control _n_ = 

_9_, restraint _n_ = _9_) and _Gfral_ -/- mice (control _n_ = _8_, restraint _n_ = _8_). Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA with Tukey’s post-hot

correction. D) Serum corticosterone in WT (control _n_ = 4, CRH _n_ = _6_) and _Gfral_ -/- mice (control _n_ = _4_, CRH _n_ = _6_). Data presented as mean ± s.e.m. with p-values calculated

using 2-way ANOVA at each timepoint. E) Serum adrenocorticotropic hormone (ACTH) in WT (control _n_ = 4, CRH _n_ = _6_) and _Gfral_ -/- mice (control _n_ = _4_, CRH _n_ = _6_). Data

presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA at each timepoint. F) Serum corticosterone 6-hr post treatment with dexamethasone (Dexa, 100 µg/kg IP). _n_ = _4_ per

group. Data presented as mean ± s.e.m. with p-values calculated using 2-way ANOVA. Source data SUPPLEMENTARY INFORMATION REPORTING SUMMARY SOURCE DATA SOURCE DATA FIG. 1 Source data. SOURCE

DATA FIG. 2 Source data. SOURCE DATA FIG. 2 Unprocessed western blots and gels. SOURCE DATA FIG. 3 Source data. SOURCE DATA FIG. 4 Source data. SOURCE DATA EXTENDED DATA FIG. 1 Source data.

SOURCE DATA EXTENDED DATA FIG. 2 Source data. SOURCE DATA EXTENDED DATA FIG. 3 Source data. SOURCE DATA EXTENDED DATA FIG. 4 Source data. SOURCE DATA EXTENDED DATA FIG. 5 Source data. SOURCE

DATA EXTENDED DATA FIG. 6 Source data. SOURCE DATA EXTENDED DATA FIG. 7 Source data. SOURCE DATA EXTENDED DATA FIG. 8 Source data. RIGHTS AND PERMISSIONS OPEN ACCESS This article is

licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any

medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed

material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are

included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and

your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this

licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/. Reprints and permissions ABOUT THIS ARTICLE CITE THIS ARTICLE Townsend, L.K., Wang, D., Knuth, C.M. _et al._ GDF15 links

adipose tissue lipolysis with anxiety. _Nat Metab_ 7, 1004–1017 (2025). https://doi.org/10.1038/s42255-025-01264-3 Download citation * Received: 27 August 2024 * Accepted: 06 March 2025 *

Published: 15 April 2025 * Issue Date: May 2025 * DOI: https://doi.org/10.1038/s42255-025-01264-3 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this

content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative