Impaired brain homeostasis and neurogenesis in diet-induced overweight zebrafish: a preventive role from A. borbonica extract

Overweight and obesity are worldwide health concerns leading to many physiological disorders. Recent data highlighted their deleterious effects on brain homeostasis and plasticity, but the mechanisms underlying such disruptions are still not well understood. In this study, we developed and characterized a fast and reliable diet-induced overweight (DIO) model in zebrafish, for (1) studying the effects of overfeeding on brain homeostasis and for (2) testing different preventive and/or therapeutic strategies. By overfeeding zebrafish for 4 weeks, we report the disruption of many metabolic parameters reproducing human overweight features including increased body weight, body mass index, fasting blood glucose levels and liver steatosis. Furthermore, DIO fish displayed blood–brain barrier leakage, cerebral oxidative stress, neuroinflammation and decreased neurogenesis. Finally, we investigated the preventive beneficial effects of A. borbonica, an endogenous plant from Reunion Island. Overnight treatment with A. borbonica aqueous extract during the 4 weeks of overfeeding limited some detrimental central effects of DIO. In conclusion, we established a relevant DIO model in zebrafish demonstrating that overfeeding impairs peripheral and central homeostasis. This work also highlights the preventive protective effects of A. borbonica aqueous extracts in DIO, and opens a way to easily screen drugs aiming at limiting overweight and associated neurological disorders.


Results
Dio models induces phenotypic and metabolic changes. In order to set up a fast and reliable DIO model in zebrafish, a 4-week overfeeding treatment was performed by providing dry food and freshly hatched artemia during the day (dry food: 15 mg (CTRL) vs. 52,5 mg (DIO)/fish/day; artemia: 6 mg (CTRL) vs 30 mg (DIO)/fish/day; see Suppl Fig. 1). The effects of such a diet were subsequently investigated for body weight, body length and BMI (Body Mass Index). After the first week of diet, a significant increase in body weight was observed and was maintained until week 4 in DIO fish compared to controls (CTRL) in both male and female groups (Fig. 1A,B). DIO-treated fish were markedly bigger than CTRL at week 4, corresponding to a 144.5% and 240% increase in body weight for males and females, respectively.
In addition, at the end of the experimental procedure, the body length of male and female fish was significantly higher in DIO-treated fish (108% and 122% increase in males and females, respectively) (Fig. 1C,D). The body mass index (BMI) was significantly increased in the DIO fish compared to CTRL (125% and 162% increase in males and females, respectively), suggesting that the gain in body weight is not only due to an increase in body length (Fig. 1E,F). Blood cholesterol and triglycerides levels were also measured and no significant changes were observed between CTRL and DIO fish (data not shown). Taken together, these data indicate that a 4-week overfeeding was sufficient to markedly increase body weight and BMI. Given that egg production could lead to a bias in body weight measurements in female and that sex hormones (i.e. estrogens) are known to impact brain plasticity including neurogenesis in both mammals and fish 34,35 , the next experiments were performed only in males.
Numerous studies have previously shown that obesity and overweight result in metabolic disorders such as dysregulation in glucose and lipid metabolisms including the development of liver steatosis 5 . In the model developed in this study, fasting blood glucose was significantly increased in DIO-treated fish compared to CTRL ( Fig. 2A). In addition, the liver from DIO fish was phenotypically bigger and more yellowish than those from CTRL, suggesting liver lipid accumulation (Fig. 2B). Oil Red O staining was consequently performed on liver cryo-sections to test this hypothesis. The liver of DIO fish exhibited an obvious red staining compared to controls ( Fig. 2C-F), demonstrating hepatic lipid accumulation in overfed fish. Of note, the intensity of the liver red oil staining was heterogeneous among the DIO fish analyzed (Fig. 2D,F). Together, these data indicate that overfed fish exhibit dysregulations in glucose and hepatic lipid metabolism at week 4.
As a first conclusion, the phenotypic and metabolic data obtained from the DIO model strongly support that the developed protocol efficiently led to overweight and dysregulation in lipid and glucose metabolism.

DIO induces BBB leakage, neuroinflammation and oxidative stress.
Metabolic disorders such as diabetes and obesity are known to disrupt BBB 36,37 , an important interface corresponding to a highly selective barrier, that separates the blood flow from the fluids in the CNS. BBB disruption induced by obesity results in the leakage of substances into the brain and can consequently disrupt central homeostasis leading to increased brain inflammation and oxidative stress 37 . In the present study, DIO fish displayed an increased body weight and BMI and showed metabolism dysfunctions. Consequently, the impact of overfeeding was next investigated for brain homeostasis focusing on blood-brain barrier (BBB) physiology, neuroinflammation, oxidative stress and neurogenesis. Given the impact of sex steroids on brain homeostasis and peculiarly in zebrafish neurogenesis 38 , we decided to perform the following investigations in males.
To investigate the impact of DIO on the BBB physiology, intraperitoneal injection of Evans Blue was performed, this dye quickly reaching the bloodstream. It results that overfeeding led to BBB leakage as revealed by the blue staining of the DIO brains ( Fig. 3A; 5 brains out of 6 were blue) compared to CTRL, that mostly remained white ( Fig. 3A; 5 brains out of 6 were white). The effect of overfeeding was next investigated on neuroinflammation by performing qPCR analyses. A consistent trend towards increased il1β, il6 and tnfα gene expression was observed in the brains of DIO fish compared to those of controls ( Fig. 3C-E). Furthermore, the pro-inflammatory nfkb transcription factor was slightly but significantly up-regulated (Fig. 3B). The morphology of microglial cells was also studied as a reflect of the cerebral neuroinflammatory state. In homeostatic conditions, microglial cells displayed a ramified morphology, while under activation, they became amoeboid (rounded morphology without ramifications) and exhibit phagocytic properties. By performing L-plastin immunohistochemistry, a switch in microglia morphology was observed (Fig. 3F,G) www.nature.com/scientificreports/ and display stronger L-plastin staining along the neurogenic niches compared to controls (Fig. 3G). As shown in Fig. 3F, the number of ramified microglia was lower in DIO fish than in CTRL in the ventral telencephalon (Vv Vd: p-value = 0.0027) and in the anterior part of the hypothalamic region (Hv: p-value = 0.2762). In addition, the number of amoeboid microglial cells was significantly higher in DIO fish compared to controls in both regions (Fig. 3F). Taken together, these qPCR and IHC data demonstrate that DIO promotes neuroinflammation. The dysregulations in lipid and glucose metabolisms observed in overfed fish could potentially affect redox homeostasis. The enzymatic antioxidant system associated with proteasome participates in the maintenance of this redox homeostasis. Consequently, some antioxidant enzymes activities (catalase, superoxide dismutase and peroxidase) and the chymotrypsin like activity of the proteasome were investigated in brain (Fig. 4). While there is no significant difference between DIO and control fish for SOD activity, a significant enhanced peroxidase activity was measured in brain lysates from DIO fish (+ 25%, p < 0.05 vs. CTRL) and a slight increase in catalase activity was also observed in DIO but failed to reach statistical significance (+ 24%, p = 0.09 vs. CTRL). These enhanced peroxidase and catalase activities could reflect the antioxidant response of the main detoxifying enzymes following a redox status imbalance in fish subjected to an overfeeding. In addition, a significant reduction of the chymotrypsin-like activity of the proteasome was measured in brain from DIO fish (− 48%, p < 0.001 vs. CTRL). This reduced proteasome activity may contribute to the altered redox status in overfed zebrafish Graphs illustrating the body weight measurements during 4 weeks for both CTRL and DIO-treated zebrafish in male and female, respectively. The zebrafish pictures highlight the morphological differences at week 4. (C), (D) Body length measurements at week 4 in male and female zebrafish, respectively. (E), (F) Body mass index (BMI; grams per square centimeter) calculated at week 4. n = number of fish. One-way ANOVA (A, B) and Student's t-test (C-F): **p < 0.01; ****p < 0.0001. Error bars correspond to standard error of the mean (SEM). Scale bar: 7 mm.

Dio impairs adult neurogenesis and locomotor activity. BBB disruption and neuroinflammation
are well-established to be potential disruptors of adult neurogenesis. Brain cell proliferation in neurogenic niches was consequently studied by performing immunohistochemistry against the proliferative marker PCNA (Proliferating Cell Nuclear Antigen) in key neurogenic regions including the ventral (Vv-Vd) and dorsal (Dm) telencephalon, the anterior part of the preoptic area (Ppa), the periventricular pretectal nucleus (PPv) and two caudal hypothalamic regions (Hv LR and LR PR) (Fig. 5). Overfed fish displayed a general blunted neurogenesis in all the regions studied. Although, it did not reach statistical significance, a consistent trend was observed towards lower proliferation in the dorsomedian telencephalon (Dm SY) and around the lateral (LR) and posterior (PR) recess of the hypothalamic nucleus (Fig. 5A). Furthermore, a significant decrease in proliferative cells was observed along the neurogenic niches from the ventral (Vv) and dorsal (Vd) nuclei of the ventral telencephalic area, the anterior part of the parvocellular preoptic nucleus (PPa), the periventricular pretectal nucleus (PPv), the ventral zone of the periventricular hypothalamus as well as along the lateral recess of the diencephalic nucleus (Hv LR). Importantly, such decreased brain cell proliferation was observed in independent experiments, reinforcing the fact that overfeeding strongly resulted in blunted neurogenesis. www.nature.com/scientificreports/ We next addressed the potential effects of overfeeding on fish behavior and thus monitored the locomotor activity. By recording the locomotion of individual fish during a 10 min period, a significant increase in inactive state was observed in DIO fish compared to their respective controls, while the total distance traveled remained unchanged (Fig. 6). These data reveal the existence of different locomotion patterns between CTRL and DIO fish.
A. borbonica herbal tea treatment prevents from deleterious effects induced by DIO. Overfed zebrafish appear as an appropriate model to test the effects of some compounds on weight gain and its www.nature.com/scientificreports/ complications 15,39,40 . A. borbonica was selected as an interesting candidate given its potential or demonstrated anti-oxidant, anti-inflammatory, and anti-diabetic according to its traditional use and to in vitro studies 29,32,33 . First, the total phenolic acids content of the A. borbonica mother infusion (4 g/L), evaluated by using Folin-Ciocalteu assay, showed a concentration of 17.2 ± 0.9 mg GAE/g of plant dry powder. By using an aluminium chloride colorimetric method, the total flavonoids content determined was of 8.9 ± 0.5 mg EE/g of plant dry powder, twice less compared to phenolic acids content (Suppl. Figure 2). These results confirmed that A. Borbonica extract contains a substantial significant level of polyphenols and antioxidant properties as other medicinal plants as Rosmarinus officinalis L (16.67 ± 0.40 mg GAE/g) 41 . Then, a high-resolution mass spectrometry (HR-MS) analysis showed that a 4 g/L A. Borbonica infusion contained a variety of polyphenols including phenolic acids such as caffeic acid, caffeoylquinic acid, dicaffeoylquinic acid and some flavonoids such as Kaempferol hexoside and quercetin hexoside (Suppl. Figure 3). Caffeic acid derivatives including caffeoylquinic acid and dicaffeoylquinic acid were the most concentrated polyphenols identified in A. Borbonica infusion, at 1,278 ± 75 ng/mL and 531 ± 83 ng/mL, respectively.
Next, to study the effects of A. Borbonica on weight gain and its consequences, DIO fish were treated overnight (6 pm to 8 am; 5 days a week) with water containing A. borbonica infusion (final concentration: 0.5 g/L water) during the 4-week period. We confirmed that overfed fish without plant extract displayed a significant increase in body weight from week 1 to week 4 compared to controls (Fig. 7A). Interestingly, DIO + A. borbonica did not result in any significant change in body weight during the first two weeks, compared to the DIO, but significantly exhibited a decrease in body weight gain during the third and fourth week compared to the DIO fish ( Fig. 7A). At week 4 ( Fig. 7B) DIO and DIO + A. borbonica fish displayed a significant BMI increase compared to CTRL but no significant difference was observed between DIO groups. The size of DIO and DIO + A. borbonica fish was also significantly higher compared to controls, but not significant differences were observed between the groups (data not shown). As well, A. Borbonica did not exhibit striking preventive effects on lipid deposits in the liver (data not shown) and did not strikingly prevent from increased fasting blood glucose levels induced by DIO (DIO vs CTRL: + 144.4%, DIO + A. borbonica vs CTRL: + 129%; n = 19-22, data not shown). www.nature.com/scientificreports/

A. borbonica aqueous extracts prevent BBB leakage and oxidative stress but has mitigated effects on neuroinflammation and neurogenesis. Given the anti-oxidant properties of A. borbonica
infusion, its effect on the central disruptions (BBB leakage, oxidative stress and neurogenesis) induced by DIO was investigated. We confirmed that DIO resulted in BBB leakage and we showed that A. Borbonica extracts limited BBB disruption (Fig. 8A). Cerebral oxidative stress was next investigated by performing dot-blot against 4-hydroxynonenal (4-HNE), a well-established marker of oxidative stress. 4-HNE levels were increased in the brain of DIO-treated fish (Fig. 8B), and remained at basal levels in DIO + A. Borbonica treated fish showing that A. borbonica prevented the increase in oxidative stress induced by DIO (Fig. 8B). However, neuroinflammation  www.nature.com/scientificreports/ in the brain of DIO + A. Borbonica appeared mitigated. Indeed, some pro-inflammatory cytokines were slightly decreased (i.e. tnfa fold induction is 1,36 vs. 4,47; data not shown) and microglial cells tended to be less amoeboid in the ventral telencephalon (Vv Vd) than in DIO fish, but remained activated in the hypothalamus (Hv), suggesting the persistence of a local neuroinflammatory state (data not shown). Finally, the potential effect of A. Borbonica on impaired neurogenesis induced by overfeeding was also studied. It did not induce any striking change in brain cell proliferation in the neurogenic niches, except in the preoptic area and the caudal hypothalamus (Hv LR) for which proliferation was slightly up-regulated and was maintained at basal level (Fig. 8C). It suggests that A. Borbonica aqueous extract could partially rescue the impaired neurogenesis induced by overfeeding in a regional dependent manner.

Discussion
We developed a reliable overfeeding zebrafish model resulting in many metabolic disorders including increased body weight, BMI and fasting blood glucose levels as well as the development of liver steatosis (Figs. 1, 2). Such increase in body weight and BMI appears more pronounced in female than male, probably in links with oogenesis and egg storage. In zebrafish, several DIO and/or HFD models have been proposed and lead to similar metabolic impairments as observed in our study, reinforcing the data obtained in this work [13][14][15][16][17]42,43 . The higher body weight and BMI observed in overfed fish should reflect the development of adipose tissue, as overfeeding protocols using artemia have been previously shown to promote subcutaneous and/or visceral adipocyte expansion in zebrafish 14,15,17 . In our work, we decided to use a mix of dry food and artemia in order to provide a diversity in the feeding as well as enrichment for zebrafish.
We also showed that DIO fish display heterogenous levels of liver steatosis. This pathological process was also previously observed in fish by Nakayama et al. 43 in a 6-week DIO model using artemia. In contrast, Landgraf et al. 42 observed hepatic steatosis only in fish overfed with both artemia and egg powder, but not with artemia alone, and this even after 8 weeks of treatment. Such differences could be explained by the quantity of food www.nature.com/scientificreports/ supplied to the overfed fish between these respective different protocols, and put our model as an intermediate one concerning liver steatosis. Interestingly, overweight and obesity are known to be associated with insulin resistance and increased glycemia 44 . Lipid steatosis is also associated with fasting hyperglycemia and type 2 diabetes 45,46 . In our work, fasting blood glucose levels of DIO-treated fish were significantly higher than in controls ( Fig. 2; ~ 40 mg/dl in CTRL vs ~ 60 mg/dl in DIO). Such an increase was previously obtained in fish overfed with artemia (without liver steatosis) but did not reach significant levels, while fish overfed with artemia and egg powder (with liver steatosis) exhibited a significant hyperglycemia 42 . Among the possible explanations for such differences, we could argue (1) for the power of the statistical analysis (26-29 fish/group in our study versus 10 for Landgraf et al.), (2) that the food quantity supplied in our study is higher than those from Landgraf 's one (137% increase in feeding), and (3) last but not least, it is tempting to speculate that the degree of liver steatosis could be correlated with hyperglycemia levels in accordance with the literature 47 , as it is known to promote chronic inflammation and insulin-resistance.
Taken together, these data highlight that the model developed in this study is relevant to other overfeeding models in the field. In addition, our data demonstrate that overfed zebrafish share common features with overweight/obese pathologies in human 1 . It consequently reinforces its use for studying the deleterious impact of overweight on several physiological processes such as brain homeostasis and plasticity.
Numerous works have shown that overweight and obesity were associated with cognitive impairments and that adult neurogenesis is involved in memory 11,[48][49][50] , raising the question of the effects of DIO on brain homeostasis. In this context, the integrity of the blood-brain barrier and brain plasticity (i.e. neurogenesis) are two key parameters to investigate as they are linked to cognitive impairments. In our work, we report an increased BBB leakage, neuroinflammation and oxidative stress in the brain of overweight fish compared to controls (Figs. 3,4,8). These results are of peculiar interest given that oxidative stress and inflammation are known to promote BBB disruption 51,52 . The neuroinflammatory state observed in the brain of overfed fish through the up-regulation of pro-inflammatory cytokines and nfkb gene expression was reinforced by the switch from ramified to amoeboid microglia (Fig. 3). In a similar way, the impact of a high-glucose/high-cholesterol diet in zebrafish was recently associated with the increase in pro-inflammatory cytokines and cd11b gene expression, a microglial marker 53 , reinforcing the results obtained in our work. As well, these results should be also paralleled with the mammalian situation for which overweight and/or high fat diets in rodents induce neuroinflammation and microglial reactivity, mainly in the hypothalamus [54][55][56] .
Interestingly, in our experimental conditions, DIO fish display an increase in cerebral oxidative stress (Figs. 4, 8; increased 4-HNE levels and higher catalase and peroxidase activity), with a reduced proteasome activity that could contribute to impair redox balance in overfed fish and could result in increased oxidized protein or lipid peroxidation product accumulation. Indeed, oxidized proteins that are not degraded by the altered proteasome system may also contribute to the enhanced reactive oxygen species generation in DIO zebrafish tissues. For instance, increased 4-HNE accumulation observed in the brain of DIO fish may induce adduct formation in proteasome subunit leading to the inhibition of proteolytic activity. Very interestingly, it has been previously shown that oxidative stress homeostasis could affect the differentiation and proliferation of progenitor cells 57,58 , bringing evidence that redox imbalance, and generation of 4-HNE levels in the brain, could affect the behavior of the neural progenitor/stem cells.
Indeed, we demonstrated from independent experiments that DIO fish display a consistent and significant decrease in brain cell proliferation in most neurogenic niches including the ventral telencephalon, the preoptic area, the hypothalamus and the periventricular region of the pretectal nucleus (Fig. 5). It fits with another model of high caloric consumption in zebrafish resulting in decreased cerebellar proliferation 59 . In addition, in mammals, genetic and/or diet models of overweight/obesity were shown to result in decreased neural stem cell proliferation and subsequently to a lower number of newborn neurons generated 54,60-63 . Such a decreased neurogenesis was also observed in other metabolic disorders including hyperglycemia in both rodents and fish 18,64 . Further investigations would be required for determining the impact of overfeeding in newborn cell migration, differentiation and survival in both constitutive and regenerative conditions in zebrafish. As well, the identification of the misregulated extrinsic and intrinsic factors responsible for blunted neurogenesis should be further investigated, considering for example the Delta-Notch signaling pathway known for controlling NSC activity. In addition, in the offspring of mice that have followed a HFD, NSC upregulates Notch receptors and its downstream effector Hes, promoting NSC quiescence and limiting neurogenesis 65 . Also, given the links between neurogenesis and cognitive functions 50 , it is important to mention that a HFD zebrafish model was shown to display impaired cognitive functions as revealed by active avoidance test and by the disturbed expression of numerous genes involved in neuronal activity, anti-oxidative stress, and BBB functions 66 .
Consequently, we demonstrated for the first time in a same zebrafish model that overfeeding induces many detrimental effects on brain homeostasis as shown by BBB leakage, neuroinflammation, enhanced oxidative stress and impaired neurogenesis. These results also raised the question of the chicken or the egg between the first factors (inflammation? oxidative stress?) disrupting brain homeostasis and would allow to further investigate the mechanisms by which metabolic disorders disturb neurogenic activity and behavior.
A. borbonica herbal tea has been traditionally used as a natural remedy for its potential anti-inflammatory, anti-oxidant and anti-diabetic properties [29][30][31][32][33] . For this reason, we decided to work on aqueous extract. The subsequent HR-MS analysis allowed us to confirm previously published data demonstrating that A. Borbonica extracts contain high phenolic and flavonoid contents associated with antioxidant properties 29,33 . HR-MS and Folin-Ciocalteu assay analyses revealed that A. Borbonica infusion contains major phenolic acids including caffeic acids derivatives (dicaffeoylquinic acid and chlorogenic acid) and some glycosylated flavonoids (kaempferol hexosides and quercetin hexosides). In HFD mice, dicaffeoylquinic acids (150 mg/kg of body weight) improve metabolic parameters (i.e. liver and adipose tissue masses, decreased inflammatory factors, better hepatic lipid Scientific RepoRtS | (2020) 10:14496 | https://doi.org/10.1038/s41598-020-71402-2 www.nature.com/scientificreports/ synthesis and degradation) 67 . A study of Jung et al. 68 also showed the interesting effect of quercetin by regulating genes involved in lipid metabolism. As well, intraperitoneal injection of chlorogenic acid (100 mg/kg of body weight) was shown to ameliorate HFD-induced liver steatosis, insulin resistance and adipocyte hypertrophy in mice 69 . In addition, caffeic acids including dicaffeoylquinic acid and chlorogenic acid have been shown to exert antioxidant properties especially in the brain [70][71][72] . Although A. borbonica did not prevent from liver steatosis in our experimental conditions, and did not significantly decrease fasting blood glucose levels, a decreasing trend was observed (DIO vs CTRL: + 144.4% increase in fasting glycemia while DIO + A. borbonica vs CTRL: + 129% increase in fasting glycemia). Importantly, A. borbonica treatment significantly limited weight gain from week 3 to week 4 compared to DIO, without modulating feeding behavior (data not shown). This result is of peculiar interest given that the BMI of DIO + A. borbonica fish is higher than CTRL, but this increase is less significant than the one of DIO vs. CTRL (Fig. 7), while the size of DIO and DIO + A. borbonica remain unchanged (data not shown: 33.46 mm for DIO vs 33.04 mm for DIO + A. borbonica-). This result have to be paralleled with those in HFD mice for which treatment with chlorogenic acid and dicaffeoylquinic acid avoid or limited gain weight 67,69 .
The effects observed in our study are probably less spectacular given the low concentrations of these polyphenols. In addition, DIO + A. Borbonica fish are protected from BBB leakage and brain oxidative stress induced by DIO (Figs. 7, 8). These preventive effects could be attributed A. Borbonica anti-oxidant activity supported by its polyphenol content (phenolic acids, flavonoids and tannins) and are consistent with in vitro studies showing its anti-oxidant properties 29,33 . Interestingly, in almost all the neurogenic niches studied, DIO fish treated with A. borbonica still display blunted neurogenesis that nevertheless appeared less severe in the preoptic area and the hypothalamus. Consequently, A. borbonica did not fully prevent from neurogenic defects induced by DIO, showing that other mechanisms than BBB leakage and oxidative stress should be involved in such neurogenic impairments. It would be also interesting to determine the A. borbonica metabolites found within the brain fish in order to elucidate which compounds could be responsible for the central protective effects.

conclusion
To conclude, we have developed an overfeeding model in zebrafish that mimics the mammalian overweight state in the periphery and also in the central nervous system. To our knowledge, this is the first report showing the deleterious impact of DIO on brain homeostasis and especially considering the forebrain neurogenesis in zebrafish.
A. borbonica aqueous extract (mimicking herbal tea consumption) was shown to limit significantly weight gain in overfed zebrafish. The DIO protocol developed in this work could serve as a new physiological screening tool for identifying "anti-overweight" and "anti-obesity" drugs. It will allow the discovery of new preventive and therapeutic treatments against weight gain and associated central deleterious effects such as BBB leakage, neuroinflammation, oxidative stress and impaired neurogenesis. Although A. borbonica aqueous extract has only a limited effect on body weight and associated BMI, it prevents BBB leakage, cerebral oxidative stress and partly improve neurogenesis. Thus, natural compounds have a limited effect on the body weight but could prevent some central disorders induced by overfeeding.

Material and methods
Animals and ethics. Three to four month-old adult wildtype male and female zebrafish (Danio rerio) were obtained from our zebrafish facility and were maintained under standard conditions of temperature (28.5 °C), photoperiod (14 h dark/10 h light), pH (7.4) and conductivity (400 μS). The gender of the animal was performed visually according to sexual dimorphisms (males are thinner and females have a bigger belly). All experiments were conducted in accordance with the French and European Community Guidelines for the Use of Animals in Research (86/609/EEC and 2010/63/EU) and approved by the local Ethics Committee for animal experimentation of CYROI and the French Government (APAFIS_ 20191105105351_v10).

Diet-induced overweight/obesity (DIO) protocol. Adult zebrafish (3-4 months) were divided into 2
or 3 groups (control and DIO or CTRL, DIO, DIO + A. borbonica) at the same density between each group (10 to 20 fish maximum per 3.5 L tank, according to the experiment); males and females being mixed in each tank. The control group was fed once a day with dry food in the morning (15 mg/fish/day, GEMMA 300, Planktovie) and freshly hatched artemia (6 mg/fish/day, Artemia cysts; REF: B052-P) in the afternoon. The DIO and DIO + A. Borbonica group was fed six times a day with dry food (52.5 mg/fish/day) and three times with freshly hatched artemia (30 mg/fish/day) in the afternoon. These treatments were performed on a four-week period (see Suppl Fig. 1).

Body weight, body mass index (BMi) and fasting blood glucose measurements. Fish of each
group were weighted every week. They were captured using a net, quickly dried on a tissue paper, briefly weighed and immediately placed back into water. This procedure is usually done in less than 20 s. In addition, the body length of the fish was measured at the beginning (first day, during the weighing process) and at the end of the www.nature.com/scientificreports/ experiment (week 4, after euthanasia-see below-) from the tip of the mouth to the end of the tail (total length). The body mass index (BMI) was calculated by dividing the body weight (g) with the square of the length (cm 2 ). For blood glucose measurements, fish were fasted the day before and euthanized using ice water (2-4 °C) in order to avoid blood glucose fluctuations due tricaine use 73 . Fish were gently dried with a tissue and one eye was removed allowing the ocular cavity to fill with blood. The glycemia (mg/dl) was measured using a glucometer (One-Touch Ultra, LifeScan, France), as previously described 19 .
Blood collection for metabolic analyses. The fish were first euthanatized with tricaine, then one eye was removed and the blood was collected using a pipet that was previously equilibrated by 1X PBS and EDTA (1 mL of 1X PBS is added in EDTA blood tube and vortexed). The blood of 5 fish was pooled (around 10 µL) in one Eppendorf with 40 µL of 1X PBS 1 × EDTA to avoid its clotting. After that, the blood samples are centrifuged and the plasma was collected frozen at − 80 °C. For cryostat sections, the brain and the liver were dissected and cryopreserved by an overnight incubation in 1X PBS, containing 30% sucrose. Then, they were embedded in OCT matrix and cut using a cryostat at 12 µm thickness.
For qPCR or protein analyses, the tissues of interest were immediately dissected, snap-frozen and kept at − 80 °C.
immunostaining. For immunohistochemistry experiments, cryostat sections were rehydrated twice with 1X PBS containing 0.2% Triton (PBS-T). Antigen retrieval was performed using sodium citrate (pH 6) heated at 80 °C for 15 min. Sections were washed twice in PBS-T before being blocked in PBS-T containing 2% BSA. Next, sections were incubated with the following primary antibodies: rabbit anti zebrafish L-plastin (kindly provided by Dr Michael Redd 74

investigation of BBB permeability (evans Blue dye injection). For investigating BBB physiology,
Evans Blue dye was used as a tracer to monitor BBB permeability 77 . Briefly, fish were anesthetized with 0.02% tricaine and intraperitoneally injected with freshly prepared 1% Evans Blue dye diluted in 1X PBS (10 µL of 1% Evans Blue for 0.1 g of zebrafish body weight). Fish were allowed to recover for 10 min before being sacrificed. Then, fish heads were fixed with 4% PFA-PBS. After, the brains were dissected and imaged. protein extraction and dot blot. Zebrafish brain were lysed with Tris HCL buffer (50 mM pH7.4 EDTA 0.01 mM) and centrifuged (10,000 rpm, 4 °C for 5 min). Supernatants were kept at − 80 °C. Protein concentration was determined according to Bradford protein assay following manufacturer's protocol. For dot blot, 20 µg of protein were plotted on a nitrocellulose membrane. Following Ponceau S (Ponceau Red) staining, membranes were blocked in blocking buffer (5% milk in 1 × PBS containing 0. www.nature.com/scientificreports/ EDTA (1 mM), pH 7.4). After centrifugation (5,000 rpm, 4 °C for 10 min), the supernatant was used for protein quantification and enzymatic assays. Total protein concentration of lysates was quantified by the bicinchoninic acid assay (BCA). The catalase activity assay was estimated on 15-20 µg of protein lysates in 25 mM Tris-HCl (pH 7.5), using protocols previously described 78,79 . Blanks were measured at 240 nm just before adding 80 µL of H 2 O 2 (10 mM final) to start the reaction. Catalase activity was determined by measuring the absorbance at 240 nm and was calculated using a calibration standard curve of increasing amount of catalase between 12.5 and 125 units/mL. Catalase activity was expressed as international catalytic units per mg of proteins and then normalized in percentage versus the control condition.
Total SOD activity was determined using the cytochrome c reduction assay, as previously described 80 . In this method, superoxide radicals generated by the xanthine/xanthine oxidase system reduce cytochrome c, thereby leading to an increase in absorbance at 560 nm. 20 µL aliquot (about 10 µg of protein) of the lysates was combined with 170 µL reaction mixture (xanthine oxidase, xanthine (0.5 mM), cytochrome c (0.2 mM), KH 2 PO 4 (50 mM, pH 7.8), EDTA (2 mM) and NaCN (1 mM)). The reaction was monitored in a microplate reader (Fluostar OPTIMA, BMG Labtech France) at 560 nm for 1 min, at 25 °C. Total SOD activity was calculated using a calibration standard curve of SOD (up to 6 units/mg). Results were expressed as international catalytic units per milligram of cell proteins and then normalized in percentage versus the control condition.
Peroxidase activity of tissue lysates was assessed according to the protocol described by Everse and colleagues 81 . A reaction mixture was prepared with 200 µL of 50 mM citrate buffer/0.2% o-dianisidine and 20 µL of lysates (between 5 to 10 µg of protein). The reaction was initiated by adding 20 µL of 200 mM H 2 O 2 . Peroxidase activity was determined by measuring the absorbance at 450 nm at 25 °C for 3 min. Peroxidase activity was expressed as international catalytic units per mg of proteins and then normalized in percentage versus the control condition.
Chymotrypsin-like activity of the proteasome was assayed using fluorogenic peptide (Sigma-Aldrich, St Louis): Suc-Leu-Leu-Val-Tyr-7-amido-4-methylcoumarin (LLVYMCA at 25 mM), as described previously 78 . Analyses were carried out with 5-10 μg of protein in 25 mM potassium phosphate buffer (pH 7.5) containing LLVY-MCA, at 37 °C for a 0-30 min incubation period. The fluorescence of aminomethylcoumarin was determined at excitation/emission wavelengths of 350/440 nm using a microplate spectrofluorometer reader (Fluostar OPTIMA, BMG labtech France). Peptidase activity was measured in the absence or in the presence of 20 µM proteasome inhibitor MG132 (N-Cbz-Leu-Leu-leucinal) and the specific proteasome activity was obtained by subtracting the residual activity (not inhibited by MG132), i.e. total cellular peptidase activity and non-proteasomal peptidase activity.
RnA extraction and reverse transcription. Zebrafish brains were removed from the skull, pooled (n = 2), and stored at − 80 °C prior to RNA extraction. Three pools of 2 brains from control and DIO-treated fish were grinded with TissueLyser II (Qiagen, Chatsworth, CA) and RNA extraction was performed using RNA easy Mini Kit (Qiagen) according to manufacturer's protocol. Then, 2 µg of RNA were reverse transcribed into cDNA Gene expression analysis by qpcR. Semi-quantitative PCR experiments were performed using the Biorad CFX Connect Real-Time System (BR006305) using the SYBR green master-mix (Eurogentec) and specific zebrafish primers. Each PCR cycle was conducted for 15 s at 95 °C and 1 min at 60 °C. Melting curve analyses and PCR efficiency were performed to confirm correct amplification. Results were analyzed and the relative expressions of the pro-inflammatory cytokine genes (il1β, il6 and tnfα) and nfkb were normalized against the housekeeping ef1a gene. The sequences of the primers are provided in Table 1.

Microscopy.
Micrographs were obtained with an Eclipse 80i Nikon microscope equipped with a Hamamatsu digital camera (Life Sciences, Japan), and with a nanozoomer S60 (Hamamatsu). Pictures were adjusted for brightness and contrast in Adobe Photoshop. cell counting. For analyzing constitutive neurogenesis, proliferative activity was determined by quantification of PCNA-positive from 2 to 3 cryostat consecutive sections (12 µm thickness/section) by region of interest. Images were analyzed for detection of PCNA positive nuclei using ImageJ software (National Institutes of Health, Bethesda, MD; RRID: SCR_003070) by adjusting parameters (threshold, binary, and watershed).  AGC AGC TGA GGA GTG AT  CCG CAT TTG TAG ATC AGA TGG   Il1β  GCT GGA GAT CCA AAC GGA TA  ATA CGC GGT GCT GAT AAA CC   tnfα  GCG CTT TTC TGA ATC CTA CGT GCC CAG TCT GTC TCC TTC T   il6  TCA ACT TCT CCA GCG TGA TG  TCT TTC CCT CTT TTC CTC CTG   nfkb  CGG CCC ACT GTA GTT GTG  TGC GTT TCC GTT ATA AGT  www.nature.com/scientificreports/ Briefly, the parameters were set up as follows for each picture: threshold 65, 255, particle size 200-infinity. Minor modifications in these parameters could be slightly adjusted according to the experiments. In addition, ImageJ automated selection of PCNA-positive nuclei was manually double-checked and adjusted if necessary, for each picture. Neuroanatomical structures were identified with DAPI counterstaining. Cell counting was performed in blind conditions by two different people and the provided graphs correspond to the mean of proliferative cells per section. For determining the number of resting and amoeboid microglia, manual counting was performed on 2 consecutive brain cryosection (12 µm thickness/section) for each region of interest (ventral telencephalon and anterior hypothalamus). The counting was performed on a total of 5 to 6 fish per condition, and the provided graphs correspond to the mean of proliferative cells per section.
Behavioral analysis. The locomotor activity of zebrafish was monitored by the ZebraCube equipment (Viewpoint). Fish were placed in tanks within the ZebraCube equipment. Locomotor activity was recorded using the viewpoint software and inactivity, small and large activity counts, distance and duration were analyzed. A total number of 12 fish (from 3 independent experiments) were subjected to behavioral analysis. Individual fish were placed in separate tanks in an equal volume of water (750 mL in each tank, corresponding to a column height of 7 cm). The tanks were placed in the ZebraCube equipment with equivalent distance between them and a separation was placed between the tanks in order to avoid visual interaction between the fish from CTRL and DIO groups. The fish were allowed to discover freely this new environment for 10 min in order to adapt to the new space without excessive amount of stress. Then, the locomotor activity was recorded for 10 min: the movement of the fish was tracked as follows: the inactivity (< 4 mm/s), small activity (4-8 mm/s) or high activity (> 8 mm/s).

Instrumentation and LC-MS/MS Conditions. Polyphenols extracted from A. borbonica infusion were
identified by Ultra-high-performance liquid chromatography coupled with diode array detection and HESI-Orbitrap mass spectrometer (Q Exactive Plus, Thermo Fisher). Briefly, 10 µL of sample was injected using an UHPLC system equipped with a Thermo Fisher Ultimate 3,000 series WPS-3000 RS autosampler and then separated on a C18 column (5 µm, 4.6 mm × 100 mm, Thermo Fisher Scientific Inc.). The column was eluted with a gradient mixture of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at the flow rate of 0.450 mL/min, with 5% B at 0.00 to 0.1 min, 75% B at 0.1 to 7.1 min, 95% B at 7.2 to 7.9 min and 5% B at 8.0 to 10 min. The column temperature was held at 30 °C and the detection wavelengths were set to 280 nm and 310 nm.
For the mass spectrometer conditions, a Heated Electrospray Ionization source II (HESI II) was used. Nitrogen was used as drying gas. The mass spectrometric conditions were optimized as follows: spray voltage = 2.8 kV, capillary temperature = 350 °C, sheath gas flow rate = 60 units, aux gas flow rate = 20 units and S lens RF level = 50.
Mass spectra were registered in full scan mode from m/z 100 to 1,500 in negative ion mode at a resolving power of 70,000 FWHM at m/z 400. The automatic gain control (AGC) was set at 1e 6 . The Orbitrap performance in negative ionization mode was evaluated weekly and external calibration of the mass spectrometer was performed with a LTQ ESI negative ion calibration solution (PIERCE). Identification of the compounds of interest was based on their exact mass, retention time and MS/MS analysis. Data were acquired and processed by XCalibur 4.0 software (Thermo Fisher Scientific Inc.).

Identification and quantification of polyphenols in Antirhea borbonica infusion.
To determine total phenolic acids content in plant extract, Folin-Ciocalteu test was used 82 . Briefly, in a 96-well microplate, 25 μL of plant extract, 125 μL of Folin-Ciocalteu's phenol reagent (Sigma Aldrich) and 100 μL of sodium carbonate (Sigma Aldrich) were added and incubated at 50 °C for 5 min and then at 4 °C for 5 min. The absorbance was measured at 760 nm (FLUOstar Optima, BMG Labtech). A calibration curve between 12.5 − 300 µM was prepared using a standard solution of gallic acid (Sigma-Aldrich, Germany). Total phenol content was expressed as g gallic acid equivalent (GAE) per g plant powder.
Total flavonoids content was also measured using the aluminium chloride (AlCl 3 ) colorimetric assay and adapted from Zhishen et al. 83 . For this measurement, 100 μL of herbal tea extract were mixed in a 96-well microplate with 6 μL of 5% aqueous sodium nitrite (NaNO 2 ) solution. After 5 min, 6 μL of 10% aqueous AlCl 3 were added and the mixture was vortexed. Then, after 1 min incubation, 40 μL of 1 M NaOH were added. The absorbance was read at 510 nm (FLUOstar Optima, BMG Labtech). A calibration curve between 6.25-300 µM was prepared using a standard solution of epicatechin (Sigma-Aldrich). Total flavonoid content was expressed as g epicatechin acid equivalent (EE) per g plant powder.
The identification and the quantification analysis of caffeic acid, dicaffeoylquinic acid, caffeoylquinic acid, quercetin and kaempferol were achieved by an LC-MS/MS analysis as previously described by 29 , with minor modifications. For the mass spectrometry quantification, a mixed stock solution containing 10 µg/mL of each polyphenols was prepared in methanol.
Calibration curves were constructed by plotting the peak area of the analytes versus the concentration of the analytes with linear regression using standard samples at eleven concentrations. The calibration curves of each polyphenols had a correlation coefficient (R 2 ) of 0.99.

Statistical analysis.
Comparisons between two groups were performed using a statistical Student's t-test.