Roles of fibroblast growth factor 21 in the control of depression‐like behaviours after social defeat stress in male rodents

Abstract Fibroblast growth factor 21 (FGF21) modulates energy metabolism and neuroendocrine stress responses. FGF21 synthesis is increased after environmental or metabolic challenges. Detailed roles of FGF21 in the control of behavioural disturbances under stressful conditions remain to be clarified. Here, we examined the roles of FGF21 in the control of behavioural changes after social defeat stress in male rodents. Central administration of FGF21 increased the number of tyrosine hydroxylase‐positive catecholaminergic cells expressing c‐Fos protein, an activity marker of neurones, in the nucleus tractus solitarius and area postrema. Double in situ hybridisation showed that some catecholaminergic neurones in the dorsal medulla oblongata expressed β‐Klotho, an essential co‐receptor for FGF21, in male mice. Social defeat stress increased FGF21 concentrations in the plasma of male mice. FGF21‐deficient male mice showed social avoidance in a social avoidance test with C57BL/6J mice (background strain of FGF21‐deficient mice) and augmented immobility behaviour in a forced swimming test after social defeat stress. On the other hand, overexpression of FGF21 by adeno‐associated virus vectors did not significantly change behaviours either in wild‐type male mice or FGF21‐deficient male mice. The present data are consistent with the view that endogenous FGF21, possibly during the developmental period, has an inhibitory action on stress‐induced depression‐like behaviour in male rodents.


| INTRODUC TI ON
Fibroblast growth factor 21 (FGF21) is synthesised in several organs including the liver, pancreas, muscle and brain. [1][2][3][4][5][6][7] or environmental challenges, including food deprivation, overfeeding, ketogenic or high-carbohydrate diet, and physical exercise, have been shown to increase the expression of FGF21 in peripheral organs and/or in the brain. 8 FGF21 acts on receptor complexes consisting of the FGF receptor (FGFR1c, FGFR2c or FGFR3c) and β-Klotho. [9][10][11] FGF21 has been shown to induce a shift to catabolic metabolism including glucose uptake and fatty acid oxidation in white adipose tissue, fatty acid β-oxidation and reduction of triglyceride accumulation in the liver, and an increase in energy expenditure. 3 Within the brain, β-Klotho, which is essential for high affinity binding of FGF receptors to FGF21, has been shown to reside in the hypothalamus, area postrema and nucleus tractus solitarius, whereas FGF receptors are broadly expressed in the brain. 12 Brain regions where β-Klotho resides play an important role for maintenance of homeostasis under stressful environmental challenges by controlling behavioural and autonomic responses. Indeed, FGF21 has been shown not only to control energy homeostasis, but also to modulate stress responses and emotional behaviours. FGF21 activates the sympathetic nervous system and the hypothalamicpituitary adrenal axis. 4,12 High serum levels of FGF21 were reported in bipolar manic patients. 13 FGF21 levels in depressed bipolar disorder patients have been shown to be affected by mood stabiliser treatments. 13,14 A negative correlation between cerebrospinal fluid FGF21 level and depression scores has also been reported. 15 Furthermore, administration of FGF21 has been reported to reduce lipopolysaccharide-induced depression-like behaviours in mice, 16 although FGF21 has also been reported to induce anxiety-related behaviour. 17,18 It remains unclear whether endogenous or exogenous FGF21 modulates psychological stressinduced depression-like behaviours.
Social defeat stress, which is an ethologically relevant social stress, has been used as an animal model for induction of depressionlike behaviours in mice. 19,20 Depression has been shown to be associated with low activity of noradrenergic neurones in the brain, 21 whereas activation of noradrenergic neurones has been shown to attenuate depression-like behaviours observed after social defeat stress. 22 On the other hand, selective destruction of noradrenergic neurones in the nucleus tractus solitarius has been reported to augment depression-like behaviour, 23 suggesting anti-depressive action of noradrenergic neurones in the nucleus tractus solitarius. β-Klotho is expressed in the nucleus tractus solitarius. Our previous study showed that i.c.v. injection of FGF21 activated cells in the nucleus tractus solitarius in rats. 24 Using these brains, we first determined whether central administration of FGF21 activated catecholaminergic neurones in the nucleus tractus solitarius by examining the expression of c-Fos protein in tyrosine hydroxylase-positive neurones in the medulla oblongata. We next determined whether tyrosine hydroxylase-positive neurones in the nucleus tractus solitarius expressed β-Klotho mRNA using double in situ hybridisation. We also determined whether social defeat stress increased FGF21 concentrations in plasma and activated catecholaminergic neurones in the nucleus tractus solitarius. We then examined whether endogenous FGF21 modulated depression-like behaviours by analysing social defeat stress-induced depression-like behaviours in FGF21-deficient mice. We further examined the effects of overexpression of FGF21 using adeno-associated virus (AAV) vectors on depression-like behaviours in wild-type and FGF21-deficient mice.

| AAV vectors
AAV8 vectors expressing mouse FGF21 or humanised Renilla reniformis green fluorescent protein (hrGFP) under the control of the CAG promoter (chicken β-actin promoter with the cytomegalovirus immediateearly enhancer) were prepared as described previously. 24,27 In brief, 293 cells were transfected simultaneously with adenovirus helper plasmid, AAV type 8 helper plasmid, and the hrGFP or FGF21 plasmid by calcium phosphate transfection and were maintained for 3 days.
The AAV vectors were purified by cesium chloride density gradient centrifugation and the titres of the AAV vectors were measured by a quantitative real-time polymerase chain reaction. Each mouse was i.p.

| Social defeat stress
Social defeat stress was applied as described previously. 20 In brief, aggressive CD1 mice, which were selected as reported previously, 19 were kept singly in their home cages (each 91 × 260 × 128 mm in size) for more than 2 days. The test mouse was introduced as the intruder into the cage of the aggressive resident male CD1 mouse and kept there for 10 minutes. This procedure of social defeat stress was applied once per day for 5 consecutive days. Behaviours of the intruder mouse during exposure to the resident aggressor mouse were videotaped. The duration of defeat posture (an upright posture with the belly exposed toward the resident), duration of immobility behaviour (no movements except for respiration with four paws on the ground) and duration of escape (rapid escape movements away from the aggressor) were measured in the first and fifth stress sessions of the experiments with FGF21-deficient mice. A nonstressed control mouse was placed in an empty cage and was kept there for 10 minutes. This procedure was repeated once per day for 5 days.
Ten animals in each group were used for measurements of plasma concentrations of FGF21. In experiments with FGF 21-deficient mice and wild-type mice, the number of wild-type mice was 6 and the number of FGF21-deficient mice was 8. In experiments for immunocytochemical detection of c-Fos protein and tyrosine hydroxylase, social defeat stress was applied once and the number of animals for nonstress and social defeat stress was 8 and 6, respectively. The same brains were used in a previous study showing immunocytochemical detection of c-Fos protein and oxytocin. 20

| Behavioural analysis
Behavioural tests were performed before and after social defeat stress in experiments with FGF21-deficient mice and in experiments with mice injected with AAV-FGF21. In the experiments with FGF21-deficient mice, an elevated plus maze test, social avoidance test, tube test, social interaction test and forced swimming test were performed sequentially before and after social defeat stress. In the experiments using mice injected with AAV-FGF21, an elevated plus maze test, social avoidance test and forced swimming test were performed sequentially before and after social defeat stress. In the experiments using FGF21-deficient mice injected with AAV-FGF21, a social avoidance test and a forced swimming test were performed sequentially before and after social defeat stress. The intervals between the behavioural tests were at least one day for each animal.
Amounts of water intake per day and food intake per day were recorded by measuring weights of water bottles and remaining food pellets in individual cages in the morning of the day of AAV injection, in the morning of the day of the first session of social defeat stress before the application of social defeat stress and in the next morning after the day of the fifth session of social defeat stress.

| Elevated plus maze test
The maze consisted of two enclosed arms (

| Social interaction test
In the experiments with FGF21-deficient mice, a social interaction test was performed to assess social contact in a familiar environment. In the test, social contact between two mice in the test cage was measured with a system that automatically analyses behaviour in cages (each 29 × 18 × 12 cm in size) (O'Hara & Co., Ltd), as described previously. 28 A wild-type or FGF21-deficient mouse was placed in pair with a C57BL/6J mouse in the test cage. Images from each cage were captured at a rate of two frames per second and the number of particles in each frame was counted automatically for 30 hours using behaviour analysing software (Time HC8_Multi; O'Hara & Co., Ltd); two particles indicated that the mice were not in contact with each other and one particle indicated contact between the two mice. Reduction of social contact has been detected in various animal models of psychiatric disorders with impaired social behaviour. 28

| Single in situ hybridisation and double in situ hybridisation for detection of mRNA of tyrosine hydroxylase and β-Klotho
Brains of C57BL/6J mice were freshly obtained by decapitation and immediately frozen in powdered dry ice. Fresh frozen sections were cut coronally at intervals of 20 μm with a cryostat. Sections were mounted on silane-coated glass slides and then processed for mRNA detection as described previously. 31 Fluorescein-or digoxigenin (DIG)-labelled cRNA probes were prepared to detect single mRNA and multiple mRNAs simultaneously. cDNA fragments of mouse tyrosine hydroxylase (1-1025 bp; GenBank, NM_009377) and mouse β-Klotho (110-646 bp; GenBank, AF165170 32 ) were subcloned into the Bluescript II plasmid vector. In vitro transcription was performed using T3 or T7 RNA polymerase (Roche Diagnostics GmbH, Mannheim, Germany).

| Immunocytochemical detection of c-Fos protein, tyrosine hydroxylase and prolactin-releasing peptide (PrRP)
One hundred minutes after injection of FGF21 or a vehicle, the animals were deeply anesthetised with Avertin (tribromoethanol, 200 mg kg -1 body weight) and perfused with a fixative solution containing 2% paraformaldehyde and 3.75% acrolein in 0.1 mol L -1 phosphate buffer (pH 7.4). In the experiments with mice after social defeat stress, one hundred minutes after the end of 10-minute social defeat stress, mice were deeply anaesthetised with Avertin (tribromoethanol, 200 mg kg -1 body weight) and perfused with 4% paraformaldehyde followed by heparinised saline.
Brain sections were cut coronally at intervals of 30 μm with a freezing sledge microtome and then processed for immunohistochemical detection of c-Fos and tyrosine hydroxylase as described previously. 20,33 In brief, sections were incubated with a rabbit polyclonal anti-   Figure 2B). These findings suggest that significant amounts of β-Klotho were expressed in catecholaminergic neurones.

| Effects of social defeat stress on plasma FGF21 concentrations
Effects of social defeat stress on plasma FGF21 concentrations were examined. Male C57BL/6J mice were exposed to aggressive ICR mice as social defeat stress once per day. Plasma concentra-

| Effects of social defeat stress on expression of c-Fos protein in tyrosine hydroxylase-positive cells of the nucleus tractus solitarius and area postrema
Effects of social defeat stress on activation of catecholaminergic neurones in the medulla oblongata were investigated in mice by F I G U R E 3 Plasma concentrations of FGF21 and corticosterone after social defeat stress. Plasma FGF21 concentrations after social defeat stress and plasma corticosterone concentrations after social defeat stress are shown. Mice received 10 minutes of social defeat stress once per day. Plasma concentrations 24 hour after the fourth social defeat stress (0) and 30 min (30), 60 min (60) and 120 min (120) after initiation of the fifth social defeat stress were determined. Non-stress rats were placed in a test cage but received no social defeat stress. Plasma FGF21 concentrations were significantly higher in the social defeat stress group compared to the non-stress group. Plasma corticosterone concentrations were significantly higher after social defeat stress compared to the concentrations in the non-stress group and those before social defeat stress. The number of animals in each group was 10. †P < 0.05 (left), social defeat stress vs non-stress, twoway factorial ANOVA. *P < 0.05, ****P < 0.0001, stress group vs non-stress group at corresponding time points; ##P < 0.01 vs after the fifth stress (30 min), ####P < 0.0001 vs before the fifth stress (0 min) in the stress group, two-way factorial ANOVA followed by Bonferroni's test

| Effects of FGF21 deficiency on water intake, food intake, body weight and behaviours after social defeat stress
The roles of endogenous FGF21 in the control of water intake, body weight, food intake and anxiety-related or depression-like behaviour in response to social defeat stress were then examined using FGF21deficient mice. These parameters were assessed before and after application of social defeat stress.
Water intake was recorded because FGF21 has been shown to be involved in water intake. Administration of FGF21 has been reported to increase water intake 40 and FGF21 deficiency blocked metabolic stress-induced water intake. 41 Water intake was increased after social defeat stress compared to before social defeat stress, and no significant difference was found between wild-type and Body weight was significantly increased after social defeat stress compared to before social defeat stress, and no significant difference was found between wild-type and FGF21-deficient mice (no significant effect of group F 1,12 = 0.273, P = 0.611; significant effect of time

| Behaviours during exposure to aggressor mice
Wild-type or FGF21-deficient mice were exposed to aggressive ICR mice for 10 minutes per day for 5 days. Durations of aggressive behaviours received, defeat posture, escape behaviour and immobility behaviour were measured during the first and fifth sessions of social defeat stress.
The duration of attacks received during exposure to aggressors was not significantly different between wild-type and FGF21-

| Elevated plus maze test
Anxiety-related behaviour was assessed by an elevated plus maze test.
Locomotor activity during an elevated plus maze test was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found between wild-type and FGF21-deficient mice (no significant effect of group These findings suggest that social defeat stress increased anxietyrelated behaviours and that FGF21 is not indispensable for increased anxiety-related behaviour observed after social defeat stress.

| Social avoidance test
Social avoidance from mice of the ICR strain, comprising the strain of mice from which the test mice received attacks, was assessed by examining time spent for social investigation towards a cylinder containing a novel ICR male mouse. Social avoidance from mice of the C57BL/6J strain, the same strain as that of the test mice, was assessed by examining time spent for social investigation towards a cylinder containing a novel C57BL/6J male mouse. The cumulative duration that test mice spent for sniffing towards perforated parts of the cylinders was determined.
Locomotor activity during a social avoidance test with an ICR mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference Amount of water intake, amount of food intake and body weight before and after social defeat stress in wild-type mice and FGF21-deficient mice. Amount of water intake per day and body weight were increased after social defeat stress compared to before social defeat stress. No significant difference was found between wild-type mice and FGF21-deficient mice. The number of wild-type mice and FGF21-deficient mice was 6 and 8, respectively. † † † †P < 0.0001, before stress vs after stress, two-way repeated measures ANOVA  Investigation time towards the cylinder containing an ICR mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found be- Locomotor activity during a social avoidance test with a C57BL/6J mouse after social defeat stress compared to before social defeat stress was not significantly reduced in wild-type mice but was reduced in FGF21-deficient mice, and it was significantly smaller in FGF21-deficient mice compared to wild-type mice (no significant effect of group F 1,12 = 0.911, P = 0.359; significant effect of time F 1,12 = 28.164, P = 0.000186; significant interaction, F 1,12 = 6.640, P = 0.0242; repeated measures two-way ANOVA; before stress in wild-type mice vs after stress in wild-type mice, P = 0.0961; before stress in FGF21-deficient mice vs after stress in FGF21-deficient mice, P = 6.02 × 10 −5 ; before stress in wild-type mice vs before stress in FGF21-deficient mice, P = 0.335; after stress in wild-type mice vs after stress in FGF21-deficient mice, P = 0.0148; post-hoc Bonferroni's multiple comparison test) ( Figure 7C, left).
Investigation time towards the cylinder containing a C57BL/6J mouse was significantly reduced after social defeat stress compared to before social defeat stress, and it was significantly reduced in FGF21-deficient mice compared to wild-type mice (significant effect of group F 1,12 = 4.889, P = 0.0472; significant effect of time The findings suggest that social defeat stress induced social avoidance from ICR mice in both wild-type and FGF21-deficient mice and that avoidance from C57BL/6J mice appeared to be greater in FGF21-deficient mice compared to wild-type-mice, being consistent with the idea that FGF21 attenuates social defeat stress-induced social avoidance. The findings suggest that social defeat stress decreased locomotor activity during the dark phase but did not significantly affect contact time with C57BL/6J mice both in wild-type and FGF21-deficient mice.

F I G U R E 6
Duration of attacks received, social defeat posture, escape behaviour and immobility behaviour during exposure to aggressor ICR mice. No significant difference was found in durations for attacks received, defeat posture, immobility behaviour and escape behaviour between wild-type mice and FGF21-deficient mice. The time spent for immobility behaviour was increased during the fifth social defeat stress compared to during the first social defeat stress. The number of wild-type mice and FGF21-deficient mice was 6 and 8, respectively. † †P < 0.01, day 1 vs day 5, two-way repeated measures ANOVA The values of these parameters were decreased after social defeat stress compared to the values before social defeat stress. No significant difference was found between wild-type mice and FGF21deficient mice. B, Total distance of locomotion and time spent for social contact with ICR mice in a social avoidance test. Total locomotor distance and time spent for investigating a cylinder containing an ICR mouse were reduced after social defeat stress, and no significant difference was found between wild-type mice and FGF21-deficent mice. C, Total distance of locomotion and time spent for social contact with a C57BL/6J mouse in a social avoidance test. Total locomotion activity was reduced after social defeat stress in FGF21-deficient mice but not in wild-type mice, and locomotor activity after stress was significantly lower in FGF21-deficient mice compared to wild-type mice. The time spent for investigating a cylinder containing a C57BL/6J mouse was significantly reduced after social defeat stress, and it was significantly shorter in FGF21-deficient mice compared to wild-type mice. The number of wild-type mice and FGF21-deficient mice was 6 and 8, respectively. † †P < 0.01, † † † †P < 0.0001, before stress vs after stress; †P < 0.05, wild-type mice vs FGF21-deficient mice, two-way repeated measures ANOVA. #### P < 0.0001, FGF21-deficient mice before stress vs FGF21-deficient mice after stress; *P < 0.05, wild-type mice after stress vs FGF21-deficient mice after stress, two-way repeated measures ANOVA followed by post-hoc Bonferroni's test in FGF21-deficient mice, P = 1.42 × 10 −6 ; before stress in wild-type mice vs before stress in FGF21-deficient mice, P = 0.986; after stress in wild-type mice vs after stress in FGF21-deficient mice, P = 0.0315; post-hoc Bonferroni's test) (Figure 9), being consistent with the idea that FGF21 has anti-depressant actions.

| Effects of FGF21 overexpression on water intake, food intake, body weight and behaviours after social defeat stress in wild-type mice
Effects of exogenous FGF21 were examined by use of an intraperitoneal injection of AAV vectors.
Plasma Water intake was increased in mice injected with 1 × 10 9 vg compared to mice injected with AAV-hrGFP and it was also increased after social defeat stress (significant effect of group, F 2,34 = 4.599,  Locomotor activity (unit) † Wild-type FGF21-deficient F I G U R E 9 Duration of immobility behaviour of wild-type mice and FGF21-deficient mice in a forced swimming test. The time spent for immobility behaviour in a forced swimming test was increased after social defeat stress compared to that before social defeat stress. Immobility behaviour after social defeat stress was augmented in FGF21-defecient mice compared to that in wild-type mice. The number of wild-type mice and FGF21-deficient mice was 6 and 8, respectively. *P < 0.05, wild-type mice after stress vs FGF21-deficient mice after stress; ## P < 0.01, wild-type mice before stress vs wild-type mice after stress; ####P < 0.0001, FGF21-deficient mice before stress vs FGF21-deficient mice after stress, two-way repeated measures ANOVA followed by Bonferroni's test

| Elevated plus maze test
Locomotor activity during an elevated plus maze test was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21 (no

The time spent staying in open arms was significantly reduced
after social defeat stress compared to before social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21 (no significant effect of group F 2,34 = 2.070, P = 0.142; significant effect F I G U R E 1 0 Plasma FGF21 concentrations, water intake, body weight and food intake in mice injected with AAV-FGF21. Plasma concentrations of FGF21 in mice injected with AAV-hrGFP, 5 × 10 8 vg AAV-FGF21 and 1 × 10 9 vg AAV-FGF21 at 33 days after AAV injection are shown (upper). The plasma concentrations of FGF21 were significantly higher in the mice injected with 1 × 10 9 vg AAV-FGF21 and in the mice injected with 5 × 10 8 vg AAV-FGF21 compared to the mice injected with AAV-hrGFP. Amount of water intake per day, amount of food intake per day and body weight before AAV injection, before social defeat stress and after social defeat stress (lower left) and increase in water intake per day, increase in food intake per day and increase in body weight during social defeat stress for 5 days (lower right) are shown. Water intake was greater in mice injected with AAV-FGF21 1 × 10 9 vg compared to mice injected with AAV-hrGFP. Body weight after social defeat stress was significantly lower in mice injected with AAV-FGF21 1 × 10 9 vg. Increase in body weight during social defeat stress was significantly smaller in mice injected with AAV-FGF21 5 × 10 8 vg compared to mice injected with AAV-hrGFP. The number of mice injected with AAV-hrGFP, 5 × 10 8 vg AAV-FGF21 and 1 × 10 9 vg AAV-FGF21 was 12, 13 and 12, respectively. **P < 0.01, ****P < 0.0001 vs AAV-hrGFP-injected mice, Kluskal-Wallis test followed by Dunn's test. †P < 0.05, AAV-FGF21 1 × 10 9 vg-injected mice vs AAV-hrGFP-injected mice, two-way repeated measures ANOVA. #P < 0.05, before stress in AAV-FGF21 5 × 10 8 vg-injected mice vs after stress in AAV-FGF21 5 × 10 8 vg-injected mice; $P < 0.05, before stress in AAV-FGF21 1 × 10 9 vg-injected mice vs before stress in AAV-FGF21 5 × 10 8 vg-injected mice; ¥P < 0.05, after stress in AAV-FGF21 1 × 10 9 vg-injected mice vs after stress in AAV-hrGFP-injected mice; two-way repeated measures ANOVA followed by Bonferroni's test. Absolute values of water intake, food intake and body weight were analysed by two-way repeated measures ANOVA followed by Bonferroni's test. Plasma FGF21 concentrations and increases in water intake and body weight after stress were analysed by Kluskal-Wallis test followed by Dunn's test. Increase in food intake was analysed by one-way factorial ANOVA. SDS, social defeat stress  These findings suggest that overexpression of FGF21 did not affect anxiety-related behaviour observed after social defeat stress.

| Social avoidance test
Locomotor activity during a social avoidance test with an ICR mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-  Locomotor activity during a social avoidance test with a C57BL/6J mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice in-  The findings suggest that overexpression of FGF21 did not significantly change social avoidance.

| Forced swimming test
The time spent for immobility behaviour in a forced swimming test was increased after social defeat stress compared to before social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21 (no significant effect of group F 2,34 = 1.587, P = 0.219; significant effect of time F 1,34 = 264.292, P = 1.33 × 10 −17 ; no significant interaction, F 2,34 = 0.050, P = 0.951; repeated measures two-way ANOVA) ( Figure 11D), suggesting that overexpression of FGF21 did not significantly change immobility behaviour in a forced swimming test.

| Effects of FGF21 overexpression on behaviours after social defeat stress in FGF21deficient mice
Effects of exogenous FGF21 in FGF21-deficient mice were examined by use of an intraperitoneal injection of AAV vectors.

| Social avoidance test
Locomotor activity during a social avoidance test with an ICR mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found between mice injected with AAV-hrGFP and mice injected with Locomotor activity during a social avoidance test with a C57BL/6J mouse was significantly reduced after social defeat stress compared to before social defeat stress, and no significant difference was found between mice injected with AAV-hrGFP and mice in- These findings suggest that overexpression of FGF21 did not significantly change social avoidance in FGF21-deficient mice.

| Forced swimming test
The time spent for immobility behaviour in a forced swimming test was increased after social defeat stress compared to before social defeat stress, and no significant difference was found between mice injected with AAV-hrGFP and mice injected with AAV-FGF21 (no

| D ISCUSS I ON
The present study showed that considerable percentage of catecholaminergic neurones in the medulla oblongata and area postrema to modulate various stress responses in behavioural, autonomic and metabolic systems. FGF21 has been shown to activate the hypothalamic CRH-peripheral sympathetic nervous system, resulting in increased energy expenditure. 44 It is tempting to speculate that activation of noradrenergic afferent projections derived from the nucleus tractus solitarius by FGF21 modulates neuroendocrine, behavioural and metabolic responses induced by social defeat stress.
Anti-depressants have been shown to reduce the duration of immobile behaviour during a forced swimming test, whereas stressful stimuli including social defeat stress have been shown to increase the duration of immobile behaviour. Thus, the forced swimming test has been shown to enable detection of anti-depressant actions of various compounds. 45 On the other hand, immobile behaviour can also be interpreted as passive coping behaviour or acquired immobility rather than depression-like behaviour during inescapable stressful stimuli 46 and may be modulated by previous experience of a forced swimming test. Considering the present consistent data of increased immobile behaviour in a forced swimming test and augmented social avoidance in FGF21-deficient mice, it is likely that FGF21-deficient F I G U R E 11 Anxiety-related behaviours in an elevated plus maze test and social avoidance behaviours and immobility behaviours in a forced swimming test in mice injected with AAV-hrGFP or AAV-FGF21. A, Total distance of locomotion (Total distance), time spent for staying in the open arms (Open arm duration) and number of entries into the open arms (Number of open arm entries) in an elevated plus maze test. The values of these parameters were decreased after social defeat stress compared to the values before social defeat stress. No significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21. B, Total distance of locomotion and time spent for social contact with ICR mice in a social avoidance test. Total locomotor distance and time spent for investigating a cylinder containing an ICR mouse were reduced after social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21. C, Total distance of locomotion and time spent for social contact with a C57BL/6J mouse in a social avoidance test. Total locomotor distance and time spent for investigating a cylinder containing a C57BL/6J mouse were reduced after social defeat stress, and no significant difference was found among mice injected with AAV-hrGFP and mice injected with AAV-FGF21. The number of mice injected with AAV-hrGFP, 5 × 10 8 vg AAV-FGF21 and 1 × 10 9 vg AAV-FGF21 was 12, 13 and 12, respectively. The number of wild-type mice and FGF21-deficient mice was 6 and 8, respectively. † † † †P < 0.0001, before stress vs after stress, two-way repeated measures ANOVA mice showed augmented behavioural depression-like behaviours after social defeat stress.
A previous study showed that FGF21 administered twice per day for 3 days attenuated lipopolysaccharide-induced depressionlike behaviour in mice. 16 In the present study, FGF21-deficient animals showed augmented depression-like behaviour induced by social defeat stress, suggesting anti-depressant actions of endogenous FGF21. On the other hand, chronic overexpression of FGF21 via peripherally administered AAV vectors in the present study had no significant effects on anxiety or depression-like behaviours in wild-type mice and FGF21-deficient mice. Noradrenaline has been shown to play an important role in the development of emotional behaviours. 47 The present study showed that noradrenergic neurones expressed β-Klotho and that FGF21 administration activated catecholaminergic neurones. It is thus possible that FGF21 released during the developmental period has an inhibitory action on stressinduced depression-like behaviours via noradrenergic neurones in the nucleus tractus solitarius. Further study is necessary to determine the effects of a lack of FGF21 signalling in medullary catecholaminergic neurones during the developmental period on stress-related behaviour in adulthood. On the other hand, FGF21 might have differential actions on behaviours dependent upon experimental conditions. Plasma FGF21 concentrations in AAV-FGF21-injected mice in the present study were increased to approximately 6.4 ng mL -1 , which is much higher than FGF21 levels after social defeat stress, although the level is comparable to or lower than FGF21 levels during starvation or ketogenic diet feeding. 12 The effective dose of FGF21 for induction of anti-depressant action might be low and in a narrow dose range, and overdose of FGF21 might not have anti-depressant F I G U R E 1 2 Plasma FGF21 concentrations, social avoidance behaviours and immobility behaviours in a forced swimming test in FGF21-deficient mice injected with AAV-hrGFP or AAV-FGF21. Plasma concentrations of FGF21 in mice injected with AAV-hrGFP and in mice injected with 1 × 10 9 vg AAV-FGF21 are shown (top). AAV-hrGFP or AAV-FGF21 was injected 26 days before blood sampling. Plasma FGF21 concentrations were significantly higher after injection of AAV-FGF21 compared to after injection of AAV-hrGFP. Total distance of locomotion and time spent for social investigation towards ICR mice in a social avoidance test are shown (second row). Total locomotor distance and time spent for investigating a cylinder containing an ICR mouse were reduced after social defeat stress, and no significant difference was found between FGF21-deficient mice injected with AAV-hrGFP and FGF21-deficient mice injected with AAV-FGF21. Total distance of locomotion and time spent for social contact with a C57BL/6J mouse in a social avoidance test are shown (third row). Total locomotor distance and time spent for investigating a cylinder containing a C57BL/6J mouse were reduced after social defeat stress, and no significant difference was found between FGF21-deficient mice injected with AAV-hrGFP and FGF21deficient mice injected with AAV-FGF21. Duration of immobility behaviour of FGF21-deficient mice injected with AAV-hrGFP or AAV-FGF21 in a forced swimming test and differences between immobility behaviour time before social defeat stress and that after social defeat stress are shown (lowest row). Time spent for immobility behaviour in a forced swimming test was reduced after social defeat stress, although no significant difference was found between mice injected with AAV-hrGFP and mice injected with AAV-FGF21. The number of AAV-hrGFP-injected and AAV-FGF21injected mice was 10 and 11. ****P < 0.0001, vs AAV-hrGFP, Mann Whitney test. †P < 0.05, † † † †P < 0.0001, vs AAV-hrGFP, two-way repeated measures ANOVA

Before
After Social investigation (sec) Before After Before After Total distance (cm) Social investigation (sec)

Before After
Before After Plasma FGF21 concentration (pg/ml) * * * * actions. It is also possible that exposure to AAV-FGF21-induced high concentrations of FGF21 for a relatively long period affected the results. Overexpression of FGF21 has been shown to induce disturbed circadian rhythm and high serum corticosterone levels. 12 Chronic disturbed circadian rhythm and high glucocorticoid levels have been shown to exert deteriorating effects on depression status, 48,49 which might have obscured possible anti-depressant actions of FGF21 in the present study.
The results of a previous study showed that FGF21 administration increased water intake. 41 Consistent with those results, overexpression of FGF21 induced water intake in the present study. Endogenous FGF21 has also been shown to mediate water intake induced by alcohol ingestion or by a ketogenic diet, whereas FGF21 does not mediate basal water intake or water intake induced by dehydration. 41 In the present study, social defeat stress-induced water intake was not affected by FGF21 deficiency or by FGF21 overexpression, suggesting that involvement of FGF21 in the control of water intake is dependent on causes of water intake and that FGF21 does not play an important role in water intake induced by stress as in water intake induced by dehydration. In the present study, body weight was decreased after overexpression of FGF21 under the condition of social defeat stress but not under the basal condition.
Both FGF21 and stress activate the hypothalamic CRH-peripheral sympathetic system and increase energy expenditure. Considering that food intake was not changed after FGF21 overexpression or stress in the present study, it is possible that the combination of exogenous FGF21 and social defeat stress synergistically or additively increased energy expenditure, resulting in a significant decrease in body weight.
Social defeat stress has been shown to perturb lipid and glucose metabolism including reduction of fatty acid utilisation and induction of insulin resistance. 50 Imbalanced activities of sympathetic and parasympathetic systems and increased activity of the hypothalamic-pituitary adrenal axis observed after chronic stress may be involved in these metabolic changes. 51 In the present study, social defeat stress transiently increased plasma concentrations of FGF21. The functions of a transient increase in FGF21 in energy metabolism remain unclear. However, FGF21 increased after stress may attenuate disturbances of lipid metabolism after chronic stress because FGF21 has been shown to increase lipid utilisation by acting on adipose tissue and on the hypothalamus 1,2 and to suppress sucrose preference, 52 resulting in a metabolic shift from using carbohydrates to using fat.
Plasma FGF21 has been assumed to originate mainly from the liver. Peripheral FGF21 acts on peripheral organs and on the central nervous system after penetrating the blood-brain barrier to affect energy metabolism and stress responses. Recently, production of FGF21 has also been suggested in several brain regions including the hypothalamus and hippocampus. 8,16,[53][54][55] Under certain conditions, the brain has been suggested to be the source of an increase in plasma FGF21. 55 It remains to be determined whether social defeat stress changes FGF21 production in the brain and whether central FGF21 is responsible for anti-depressant actions. The present study suggests that centrally administered FGF21 activates catecholaminergic neurones in the nucleus tractus solitarius, where β-Klotho is expressed. It is tempting to speculate that these catecholaminergic neurones modulate emotion-related behaviours. 43 Further study is necessary to clarify whether activation of catecholaminergic neurones by centrally administered FGF21 is dependent on β-Klotho in the nucleus tractus solitarius, or whether endogenous FGF21 activates medullary catecholaminergic neurones after social defeat stress.
In conclusion, the present study provides evidence for an anti-

DATA AVA I L A B I L I T Y
The data that support the findings of this study are available from the corresponding author upon reasonable request.