Ontogeny and nutritional control of adipogenesis in zebrafish (Danio rerio).

The global obesity epidemic demands an improved understanding of the developmental and environmental factors regulating fat storage. Adipocytes serve as major sites of fat storage and as regulators of energy balance and inflammation. The optical transparency of developing zebrafish provides new opportunities to investigate mechanisms governing adipocyte biology, however zebrafish adipocytes remain uncharacterized. We have developed methods for visualizing zebrafish adipocytes in vivo by labeling neutral lipid droplets with Nile Red. Our results establish that neutral lipid droplets first accumulate in visceral adipocytes during larval stages and increase in number and distribution as zebrafish grow. We show that the cellular anatomy of zebrafish adipocytes is similar to mammalian white adipocytes and identify peroxisome-proliferator activated receptor gamma and fatty acid binding protein 11a as markers of the zebrafish adipocyte lineage. By monitoring adipocyte development prior to neutral lipid deposition, we find that the first visceral preadipocytes appear in association with the pancreas shortly after initiation of exogenous nutrition. Zebrafish reared in the absence of food fail to form visceral preadipocytes, indicating that exogenous nutrition is required for adipocyte development. These results reveal homologies between zebrafish and mammalian adipocytes and establish the zebrafish as a new model for adipocyte research.

storage of energy as neutral lipid (i.e., triacylglycerol) in adipose tissues. Storing excess energy as neutral lipid is an evolutionarily conserved characteristic common to virtually all animals, providing a valuable energy source during periods of nutrient scarcity ( 2,3 ). Vertebrates are capable of storing neutral lipid in several tissues, with adipose tissue serving as the primary depot. The principal cellular component of adipose tissue is the adipocyte, a cell type specialized for storing fat in cytoplasmic neutral lipid droplets and endocrine control of energy balance ( 4,5 ).
Current knowledge of adipocyte development and physiology is largely derived from research using mammalian model systems. Mammals develop two general types of adipose tissues: white adipose tissue (WAT) and brown adipose tissue (BAT). WAT is more abundant and serves primarily as a site of energy storage and mobilization, while BAT primarily functions in energy expenditure in the form of thermogenesis. WAT and BAT form in distinct anatomic depots during mammalian development, with different depots displaying distinctive patterns of gene expression, endocrine sensitivity, and association with metabolic diseases ( 2 ). Adipocytes within WAT (white adipocytes) and BAT (brown adipocytes) both contain cytoplasmic neutral lipid droplets; however, mature white adipocytes typically contain a single large droplet (unilocular), while brown adipocytes contain multiple smaller lipid droplets. Previous studies have indicated that adipocytes develop from multipotent mesenchymal stem cells (MSCs). MSCs are classically considered to be derived from the mesoderm, although recent studies suggest that MSCs have additional developmental origins, such as the neural crest or neuroepithelium ( 6,7 ). In addition to producing adipocytes during development, MSCs are thought to populate distinct anatomical sites in adult animals, including bone Abstract The global obesity epidemic demands an improved understanding of the developmental and environmental factors regulating fat storage. Adipocytes serve as major sites of fat storage and as regulators of energy balance and infl ammation. The optical transparency of developing zebrafi sh provides new opportunities to investigate mechanisms governing adipocyte biology, however zebrafi sh adipocytes remain uncharacterized. We have developed methods for visualizing zebrafi sh adipocytes in vivo by labeling neutral lipid droplets with Nile Red. Our results establish that neutral lipid droplets fi rst accumulate in visceral adipocytes during larval stages and increase in number and distribution as zebrafi sh grow. We show that the cellular anatomy of zebrafi sh adipocytes is similar to mammalian white adipocytes and identify peroxisome-proliferator activated receptor ␥ and fatty acid binding protein 11a as markers of the zebrafi sh adipocyte lineage. By monitoring adipocyte development prior to neutral lipid deposition, we fi nd that the fi rst visceral preadipocytes appear in association with the pancreas shortly after initiation of exogenous nutrition. Zebrafi sh reared in the absence of food fail to form visceral preadipocytes, indicating that exogenous nutrition is required for adipocyte development. These results reveal homologies between zebrafi sh and mammalian adipocytes and establish the zebrafi sh as a new model for adipocyte research. The worldwide epidemic of obesity and associated complications, such as type II diabetes, hypertension, and cardiovascular disease, has resulted in the designation of obesity as a major public health challenge of our time ( 1 ). Obesity is a disorder of energy imbalance in which an excess of energy intake over expenditure leads to increased (dpf) and feeding commencing at approximately 5 dpf. Zebrafi sh larvae enter metamorphosis beginning approximately 14 dpf, resulting in an adult body plan by 28 dpf. Third, zebrafi sh are amenable to high-throughput forward genetic and chemical screens ( 18 ), facilitating the identification of new genes and molecules that regulate adipogenesis and energy balance. Fourth, previous studies by our group and others have indicated that lipid metabolism pathways are conserved between fi sh and mammals ( 3,(19)(20)(21)(22)(23). The existence of adipocytes in adult zebrafi sh has been documented ( 24,25 ), however zebrafi sh adipocyte development and physiology before adult stages have not been studied. As a consequence, the properties of zebrafi sh adipocytes have not been described, and tools for their in vivo observation and analysis have not been developed.
In contrast to zebrafi sh, the properties of adult adipose tissues in other aquacultured teleost fi shes have been investigated. Like their mammalian counterparts, adult fi sh adipocytes store neutral lipid in the form of triacylglycerol, and these stores are mobilized during periods of nutrient deprivation (26)(27)(28). Fat deposition in fi sh is also sensitive to nutrient availability ( 29,30 ), and the enzymatic machinery responsible for fat deposition and mobilization in fi sh appears to be similar to that found in mammals ( 27 ). Cultured fi sh adipocytes express genes that are homologous to mammalian adipocyte markers [e.g., Peroxisome proliferator-activated receptor ␥ ( Pparg ), CCAAT/enhancer binding protein ␣ , Lipoprotein lipase , and Leptin ] (31)(32)(33)(34) and are also responsive to many of the same endocrine signals as mammalian adipocytes (e.g., insulin and glucagon) ( 33,35 ). Recent studies have established methods for isolating stromovascular cells and preadipocytes from visceral adipose tissues in adult fi sh and have shown that these cells can differentiate into adipocytes under appropriate culture conditions ( 32,33 ). Taken together, these studies have provided important information about the biology of adult fi sh adipocytes and revealed extensive molecular and physiological homology between adult adipocytes in fi sh and mammals. However, previous analysis of fi sh adipocytes, like their mammalian counterparts, has focused on adult stages. Consequently, we understand very little of the developmental origins of fi sh adipocytes and the environmental factors that regulate their formation.
In this study, we defi ne methods to visualize zebrafi sh adipocytes in vivo, use these methods to describe the development of the zebrafi sh adipocyte lineage, and defi ne a novel role for exogenous nutrition in the early development of zebrafi sh adipocytes. Our results establish the zebrafi sh as a new model organism for studying adipocyte development and physiology.

Animals and experimental conditions
All zebrafi sh experiments were conducted in conformity with the Public Health Service Policy on Humane Care and Use of Laboratory Animals using protocols approved by the Institutional marrow ( 8 ) and adipose tissue stroma and vasculature (9)(10)(11), where they can produce new adipocytes and other cell types. Multipotent adipogenic precursor cells have also been found in circulating blood ( 12 ), suggesting that tissue-associated MSCs or their progeny might colonize and differentiate in distant anatomical niches. Mammalian MSCs can produce a bipotential early adipocyte precursor (adipoblast) that is capable of differentiating into committed white or brown preadipocytes. Upon appropriate stimulation, these preadipocytes can terminally differentiate into mature adipocytes, as indicated by accumulation of cytoplasmic neutral lipid droplets. However, the intermediary stages of in vivo differentiation between the MSC and the mature adipocyte are not fully resolved.
Mammalian adipocyte biology is strongly infl uenced by environmental factors, such as the nutrient supply. Mature adipocytes accumulate and mobilize neutral lipid as a function of nutrient availability, and nutrient availability during the early postnatal period can also exert a profound infl uence on the development of adipose tissues. For example, nutrient deprivation during early postnatal stages can lead to reduced fat mass and adipocyte cell number ( 13 ), while overnutrition during similar stages can result in increased number of both adipocytes and preadipocytes ( 14 ). Adipocyte number is specifi ed during childhood and adolescence and then remains constant during adulthood ( 15 ), indicating that early environmental infl uences on adipogenesis during development can have important lifelong impacts on adiposity. However, the molecular and cellular mechanisms underlying nutritional control of adipogenesis are not well understood.
Mammalian models have provided many important insights into adipocyte biology; however, analysis of mammalian adipogenesis has largely been performed using in vitro cell culture platforms that have the potential to deviate signifi cantly from in vivo adipogenesis programs ( 16 ). Furthermore, in vivo genetic analyses of adipogenesis in mammals have mostly been limited to reverse genetic approaches. As a result, our understanding of the mechanisms underlying adipocyte development and physiology remains incomplete, especially with respect to early developmental stages of this cell lineage. The gaps in our knowledge of adipogenesis are exemplifi ed by the fact that the developmental origins of MSCs and their adipocyte progeny have not been fully resolved. Similarly, the genetic and cellular mechanisms controlling development of MSCs and adipocytes in vivo remain largely unknown ( 2,17 ). These important gaps in our knowledge can be attributed in part to the inherent limitations and challenges of available mammalian model systems.
The zebrafi sh ( Danio rerio ) provides new opportunities to study adipogenesis through a combination of several unique attributes. First, zebrafi sh are optically transparent from the time of external fertilization through early adulthood, permitting in vivo observation of developmental and physiological processes. Second, zebrafi sh development occurs rapidly, with larvae hatching from their protective chorions at approximately 3 days postfertilization microscopy to locate representative adipocytes. Ultrathin sections were cut with a diamond knife (70-80 nm thickness), mounted on 200 mesh copper grids, and stained with 4% aqueous uranyl acetate for 15 min followed by Reynolds' lead citrate for 8 min ( 39 ). The sections were observed using a LEO EM-910 transmission electron microscope (LEO Electron Microscopy, Thornwood, NY) with accelerating voltage of 80 kV. Digital images were taken using a Gatan Orius SC 1000 CCD camera and DigitalMicrograph 3.11.0 software (Gatan, Pleasanton, CA).

Measurement of total lipid
Total lipid in whole zebrafi sh was measured using the Folch method ( 40 ). Briefl y, individual zebrafi sh were euthanized by overexposure to Tricaine, weighed to determine total mass as wet weight, snap frozen, and homogenized in 500 l 2:1 chloroform:methanol, and then washed with 100 l 0.9% NaCl. Homogenates were vortexed and centrifuged at 2,000 g for 30 min, and then the lower layer was extracted, dried to completion on preweighed aluminum micro-weighboats, and weighed to measure total lipid mass per fi sh. Mass measurements are reported as mean ± SD.

In vivo microscopy reveals spatial and temporal patterns of zebrafi sh adipogenesis
The optical transparency of the developing zebrafi sh provides new opportunities to monitor adipocyte ontogeny and physiology in a living vertebrate. However, meth-Animal Care and Use Committee of the University of North Carolina at Chapel Hill. Wild-type zebrafi sh (TL strain) were raised under a 14 h light/10 h dark cycle at a constant temperature of 28.0 ± 0.5°C using standard protocols ( 36 ). Zebrafi sh embryos were raised in 100 mm diameter Petri dishes through 5 dpf at a maximum density of 40 embryos per 20 ml egg water. At 5 dpf, larvae were transferred to static 2 liter tanks containing 1 liter of fresh egg water and fed to satiety twice per day with a 3:2 mixture of Active Spheres Golden Pearls (Brine Shrimp Direct, Ogden, UT) and spirulina powder (Aquatic Eco-Systems, Apopka, FL; cat. SP1). At 10 dpf, the diet was changed to a 3:2 mixture of Rotifer Size I Golden Pearls and spirulina powder, and tanks were placed on a slow drip of system water (pH 6.8, conductivity 850 S) on a recirculating zebrafi sh aquaculture system (Marine Biotech, Beverly, MA). At the onset of metamorphosis at 14 dpf, tanks were moved onto fl owing system water, and feedings were supplemented with a small amount of live artemia (Aquafauna Bio-Marine) and gradually weaned onto an exclusive live artemia diet over the course of several days. In addition to live artemia, adult zebrafi sh (>28 dpf) were fed a fl ake food mixture of fi ve parts TetraMin fl akes (Aquatic Eco-Systems; cat. 16623), two parts AquaTox fl akes (Aquatic Eco-Systems; cat. AX5), one part Spirulina fl akes (Aquatic Eco-Systems; cat. ZSF5), one part freezedried brine shrimp (Aquatic Eco-Systems; cat. SB113), and one part Cyclopeeze (Argent Laboratories, Redmond, WA; cat. F-CY-CL-FD30-CS). In some experiments, zebrafi sh were maintained in 100 M phenylthiourea (Lancaster Synthesis) to inhibit melanin synthesis.
For starvation and refeeding experiments, zebrafi sh were maintained in static egg water to avoid introduction of digestible material from the recirculating aquaculture system. Animals used in these experiments were exclusively fed either a 3:2 mixture of Active Spheres Golden Pearls and spirulina powder (for fi sh 5-10 dpf) or a 3:2 mixture of Rotifer Size I Golden Pearls and spirulina powder (for fi sh >10 dpf).

In vivo labeling and imaging of zebrafi sh adipocytes
Nile Red (Invitrogen; cat. N-1142) was dissolved in acetone at 1.25 mg/ml and stored in the dark at Ϫ 20°C. Vessels containing live unanesthetized zebrafi sh were supplemented with Nile Red to a fi nal working concentration of 0.5 g/ml egg water and then placed in the dark for 30 min. Zebrafi sh were then anesthetized in Tricaine (MS-222; Sigma-Aldrich), mounted in 3% methylcellulose, and imaged using a Leica MZ 16F fl uorescence stereomicroscope equipped with a green fl uorescent protein (GFP) long-pass emission fi lter set (HQ480/40× and HQ510LP), a RGB liquid crystal color fi lter module, and a Retiga 2000R Fast 1394 cooled digital monochrome camera (QImaging, Tucson, AZ). Oil Red O (ORO) staining ( 22 ) and staging of zebrafi sh embryos and larvae ( 37 ) were performed as previously described. Standard length (SL; the distance between the tip of the snout and the caudal peduncle) was measured using an eyepiece reticle and used to provide a robust common metric for larval and adult zebrafi sh staging (David Parichy, personal communication). SL is reported as mean ± 95% confi dence interval.

Toluidine blue staining and electron microscopy
Zebrafi sh were euthanized by prolonged exposure to Tricaine and then fi xed by immersion in 4% formaldehyde/2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer + 0.05% CaCl 2 (pH 7.4) overnight to several days at 4°C. The samples were postfi xed for 1 h in potassium ferrocyanide-reduced osmium ( 38 ), dehydrated through a graded series of ethanol washes, and embedded in Spurr's epoxy resin (Polysciences, Warrington, PA). Transverse 1 µm sections were cut in the region of the gut, stained with 1% toluidine blue/1% sodium borate, and examined by light cytes, we performed histological staining and transmission electron microscopy on visceral adipocytes in 28 dpf zebrafi sh. Toluidine blue-stained sections confi rmed that lipid droplets contained within zebrafi sh visceral adipocytes can achieve a wide range of sizes, some exceeding 100 m in diameter ( Fig. 2A ). Electron microscopy revealed that some adipocytes contained multiple small lipid droplets ( Fig. 2B ), a hallmark of early-stage terminal differentiation in mammalian white adipocytes ( 45 ). Congruently, mature adipocytes were found to contain a single large lipid droplet with an eccentric nucleus ( Fig. 2D ), similar to mature mammalian white adipocytes ( 45 ). Ultrastructural analysis revealed that zebrafi sh adipocytes, like their mammalian counterparts, are metabolically active as evidenced by enrichment in fi lamentous mitochondria and caveolae ( Fig. 2B-E ) ( 45 ). Adjacent adipocytes ods for visualizing zebrafi sh adipocytes in vivo have not been developed. The defi ning characteristic of the adipocyte is the ability to store energy as neutral lipid in large cytoplasmic droplets. The fl uorescent lipid probe Nile Red is quenched in an aqueous environment and emits fl uorescence upon excitation when incorporated into hydrophobic lipid environments. Importantly, the wavelengths of Nile Red fl uorescence excitation and emission maxima are reduced (blue shift) when the probe is incorporated into neutral lipid (e.g., triacylglycerol) compared with polar lipid (e.g., phospholipid bilayers) ( 42 ). This property of Nile Red fl uorescence permits the unambiguous detection of neutral lipid deposits within living cells and animals. We therefore used Nile Red to monitor adipocyte neutral lipid localization in live zebrafi sh as a function of developmental stage.
Zebrafi sh are born with a large yolk mass that serves as their principal nutrient supply through the onset of exogenous feeding that begins approximately 5 dpf (3.6 ± 0.1 mm SL). Consistent with previous reports, Nile Red staining of 5 dpf zebrafi sh revealed that the yolk mass is the major depot of neutral lipid in zebrafi sh embryos ( Fig. 1A , B ) ( 23 ). After the completion of yolk absorption at approximately 7 dpf, zebrafi sh were devoid of salient neutral lipid droplets until approximately 8 dpf (4.3 ± 0.1 mm SL) when we fi rst observed small neutral lipid droplets in the right visceral cavity ( Figs. 1C, D and 6K ). The right-sided visceral position of these initial adipocytes coincides with the stereotypic position of the pancreas ( 43 ), suggesting that the fi rst adipocytes form in close proximity to the pancreas. By 15 dpf (5.3 ± 0.1 mm SL), the number and size of neutral lipid droplets had increased and their anatomic distribution had expanded ( Fig. 1E, F ). Upon completion of larval-adult metamorphosis at 28 dpf (9.6 ± 0.4 mm SL), the visceral cavity continued to contain the largest neutral lipid depot, with smaller depots appearing in pectoral fi n plate, jaw, spinal column, subcutaneous, pericardial, and periorbital positions ( Fig. 3A ). In older adults, adipose depots were also observed in subcutaneous locations along the dorsal and ventral midline and near the base of other fi ns (data not shown). We occasionally observed smaller neutral lipid droplets in the liver (data not shown) and intestinal epithelium ( Fig. 3D ), suggesting that Nile Red could also be used to monitor neutral lipid droplets in these tissues.
To confi rm that the structures stained by Nile Red contain neutral lipid, we compared the in vivo Nile Red staining pattern in individual 15 dpf zebrafi sh to that of ORO, a terminal histochemical method for staining neutral lipid commonly used in zebrafi sh and other animals ( Fig. 1G-I ) ( 22,32,44 ). Zebrafi sh adipocyte lipid droplets that stained with Nile Red also stained with ORO, confi rming that zebrafi sh adipocyte lipid droplets stained by Nile Red contain neutral lipid ( Fig. 1G-I ).

Cellular anatomy of zebrafi sh adipocytes is conserved with mammalian white adipocytes
To defi ne the cellular anatomy of zebrafi sh adipocytes and disclose potential homologies to mammalian adipo-  35 dpf fi sh starved 7 days; P < 0.005). These reductions in body mass, length, and lipid content in 35 dpf zebrafi sh starved for 7 days were not accompanied by appreciable increases in mortality or lethargy. These large reductions in lipid content and body size in starved zebrafi sh are similar in magnitude to starvation-induced reductions previously observed in other teleost fi shes ( 26,35 ).
To test the ability of zebrafi sh adipocytes to deposit fat in response to refeeding, 35 dpf zebrafi sh that had been starved for 7 days were returned to a normal diet and monitored using Nile Red. After 1 day of feeding, Nile Red stained the intestinal epithelium, presumably labeling new lipid droplets derived from the absorbed dietary lipid ( Fig.  3D ). Within 4 days after feeding was restored, neutral lipid was apparent in the same anatomical depots that contained fat before nutrient deprivation ( Fig. 3E ). This result suggests that starvation does not signifi cantly alter the distribution of fat depots and that adipose depots can be reused upon refeeding. Fat deposition 4 days after refeeding was accompanied by a 172% increase in total lipid mass in individual zebrafi sh ( P < 0.001 vs. starved; Fig. 3F ), confi rming the Nile Red staining results. In addition to lipid accumulation, refeeding for 4 days resulted in a 198% increase in body mass (15.9 ± 2.3 mg; P < 0.001 vs. starved) and a 26% increase in SL (9.7 ± 0.7 mm SL; P < 0.001 vs. starved). These results demonstrate that refed fi sh can quickly recover lipid content, body mass, and SL equivalent to fi sh fed constantly. Taken together, these results show that zebrafi sh adipocytes are able to mobilize their fat stores during periods of nutrient scarcity and are able to rapidly deposit fat upon restoration of an exogenous nutrient supply.

Zebrafi sh fabp11a and pparg are expressed by zebrafi sh adipocytes
Nile Red is a useful vital marker of neutral lipid droplets in terminally differentiated zebrafi sh adipocytes; however, it does not permit visualization of the adipocyte lineage prior to neutral lipid deposition. We therefore sought to identify additional molecular markers of the zebrafi sh adipocyte lineage. Previous studies in mammalian models have identifi ed genes that are expressed by preadipocytes and differentiated adipocytes, including Pparg , a nuclear receptor that controls transcription of genes involved in adipocyte development and lipid metabolism ( 16,(47)(48)(49). The zebrafi sh genome encodes a single ortholog of mammalian Pparg ( pparg ) ( 50 ), and previous immunohistochemical analysis in adult zebrafi sh has suggested that Pparg protein is present in adipocytes, intestinal epithelium, and other tissues ( 24 ). To determine whether pparg is expressed by adipocytes during earlier stages of zebrafi sh development, we used Nile Red vital staining to document the location of adipocytes containing lipid droplets in live 15 dpf zebrafi sh and then processed the same animals for whole-mount in situ hybridization (WISH) using a riboprobe directed against zebrafi sh pparg mRNA. Zebrafi sh pparg mRNA was present in visceral cells that also stained with Nile Red, demonstrating that pparg is expressed in terminally differentiated adipocytes as early as 15 dpf ( Fig. often formed direct intercellular contacts, potentially permitting cell-cell communication via gap junctions ( Fig. 2E ) ( 46 ). Zebrafi sh adipocytes were often found in close proximity to blood capillaries, providing immediate access to lipoprotein-derived fatty acids and other factors available in the serum ( Fig. 2B ). Taken together, these data establish that the cellular anatomy of zebrafi sh visceral adipocytes is conserved with mammalian white adipocytes.

Zebrafi sh fat depots are mobilized in response to starvation and deposited in response to refeeding
The principal function of adipocytes is to store energy as fat that can be used during periods of nutrient scarcity. We therefore tested whether zebrafi sh adipocytes are capable of mobilizing their neutral lipid depots in response to starvation. Nile Red staining revealed that 28 dpf zebrafi sh raised on a normal diet developed adipose fat depots in multiple anatomical locations ( Fig. 3A ). To test the ability of zebrafi sh to mobilize its fat depots, we starved individual 28 dpf zebrafi sh and monitored fat depots longitudinally using Nile Red ( Fig. 3 ). After 4 days of starvation, all fat depots were reduced in size ( Fig. 3B ). Reduction of the major depots in the viscera and pectoral fi n plate appeared to be slower than other depots; however, we speculate that this is due to the larger initial size of those major depots. Nile Red imaging of 35 dpf zebrafi sh starved for 7 days revealed a complete depletion of all fat stores ( Fig. 3C ). The depletion of fat depots during 7 days of starvation was accompanied by a 67% reduction in the total lipid mass in individual zebrafi sh ( P < 0.0001; Fig. 3F ), indicating that the mobilized neutral lipid was rapidly metabolized. In addition to lipid depletion, starvation for 7 days caused a 58% reduction in body mass (12.6 ± 2.9 mg in fed 28 dpf fi sh vs. 5.3 ± 2.3 mg in 35 dpf fi sh starved 7 days; P < 0.001) and a 19% reduction in SL due to kyphosis (9.5 ± 0.4 mm SL in fed 28 dpf fi sh vs. 7.7 ± 1.2 mm SL in 4A , B ). Furthermore, transverse cryosections revealed that these adipocytes were located within the pancreas ( Fig.  4C ). In contrast, we did not detect pparg expression in any visceral cells lacking neutral lipid droplets ( Fig. 4A, B ). These results indicate that pparg is not robustly expressed in adipocyte precursors, however it remains possible that they express pparg at levels below the sensitivity of our assay. In accordance with previous reports, we also observed pparg expression in the intestinal epithelium ( Fig. 4B, C ) ( 24,51 ). Therefore, our results reveal that pparg is a marker of terminally differentiated zebrafi sh adipocytes during initial stages of zebrafi sh adipogenesis.
Fabp4 (also called aP2 and A-FABP ) is a lipid chaperone that serves multiple roles in lipid transport and metabolism, and is a marker of the mammalian adipocyte lineage ( 52 ). Fabp4 is expressed by mammalian preadipocytes and adipocytes both in vivo and in vitro ( 16,53 ) as well as in macrophages ( 54 ) and dendritic cells ( 55 ). The zebrafi sh genome encodes two homologs of mammalian Fabp4 , fabp11a and fabp11b . Recent phylogenetic analyses revealed that these two paralogous zebrafi sh genes are homologous to members of a subfamily of mammalian fatty acid binding proteins that includes Fabp4 , Fabp5 , Fabp8 , and Fabp9 ( 56,57 ). These data suggest that the last common ancestor of zebrafi sh and mammals possessed a single Fabp gene that was duplicated once in the zebrafi sh lineage, producing fabp11a and fabp11b , and duplicated multiple times in the mammalian lineage producing Fabp4 , Fabp5 , Fabp8 , and Fabp9 ( 57 ). Although mammalian Fabp4 and Fabp5 are expressed by adipocytes ( 16,53,58 ), these genes as well as Fabp8 and Fabp9 are also expressed in other cell lineages ( 52 ). Previous expression pattern analysis of zebrafi sh fabp11a and fabp11b have been limited to embryonic and early larval stages ( 56,57 ), and it remained unknown whether the zebrafi sh fabp11a and fabp11b genes are expressed within the adipocyte lineage. Zebrafi sh stained with Nile Red at 15 dpf were processed for WISH using a riboprobe directed against fabp11a or fabp11b mRNA. Colocalization of fabp11a transcript in differentiated adipocytes identifi ed by Nile Red staining established that fabp11a is expressed by zebrafi sh adipocytes ( Fig. 4D, E ). Transverse sections revealed that these fabp11a -expressing cells are located within the pancreas ( Fig. 4F ), similar to the location of pparg -expressing adipocytes in age-matched animals ( Fig. 4C ). We also detected abundant fabp11a transcript in distinct cells that lacked neutral lipid and were located near mature adipocytes ( Fig. 4D, E ). Based on their anatomic location within the pancreas ( Fig. 4F ) and lack of salient neutral lipid droplets, we speculate that these fabp11a -expressing cells in the pancreas are adipocyte precursors (hereafter referred to as preadipocytes). Taken together, our data identify fabp11a as a marker of juvenile zebrafi sh adipocytes and possibly also their preadipocyte precursors.
In order to determine if fabp11b is also expressed in the adipocyte lineage, we performed colocalization analysis between Nile Red staining and fabp11b expression. At 15 dpf, no colocalization between Nile Red and fabp11b was observed ( Fig. 4G, H ). Moreover, we did not detect fabp11b Fig. 3. Zebrafi sh fat depots are mobilized in response to starvation and deposited in response to refeeding. Zebrafi sh were starved for 7 days beginning at 28 dpf and then refed for 4 days. Individual zebrafi sh were stained with Nile Red and imaged daily to monitor neutral lipid deposits (arrowheads in A and E). A-E depict an individual representative animal. A: Zebrafi sh fed normally through 28 dpf develop salient neutral lipid depots in the viscera (v), pectoral fi n plate (pf), pericardial region (c), jaw (j), periorbital region (o), subcutaneous positions (s), and spinal column (sc). B: When starved for 4 days, neutral lipid depots were reduced in all locations, although the larger visceral and pectoral fi n plate depots were the last to be exhausted. C: After 7 days of starvation, all neutral lipid depots were depleted. D, E: Refeeding for 1 day was suffi cient to form transient neutral lipid deposits in the intestine (arrow in D), and refeeding for 4 days was suffi cient to reestablish neutral lipid depots in the same locations as before starvation (arrowheads in A and E). Anterior is to the left and dorsal at the top in all images. F: To confi rm these imaging results, we used the Folch method to extract and weigh total lipid from individual zebrafi sh at 28 dpf before starvation, at 35 dpf after 7 days of starvation, and at 39 dpf after 4 days of refeeding (6-9 individuals/ group). Results are shown as mean ± SD (***, P < 0.0001; **, P < 0.001). Bar = 1 mm. visceral adipose depot express fabp11a . We therefore assessed fabp11a mRNA expression by WISH in young adult zebrafi sh at 28 dpf when adipose depots have been established in the viscera, pectoral fi n plate, jaw, and spinal column ( Fig. 3A ). Our in situ hybridization analysis revealed fabp11a expression in distinct cells in all these anatomical locations ( Fig. 5G-J ), suggesting that fabp11a continues to label adipocytes in major adipose depots into adult stages.

Zebrafi sh adipogenesis requires exogenous nutrition
Nutrient availability during the early postnatal stages of mammalian ontogeny can exert a strong infl uence on adipocyte development ( 13,14 ). To determine if availability of exogenous nutrition infl uences zebrafi sh adipogenesis, we monitored adipogenesis in zebrafi sh fed normally beginning at 5 dpf (fed) and also zebrafi sh raised in the absence of an exogenous nutrient supply (starved). Zebrafi sh fed normally formed adipocyte neutral lipid droplets by 8 dpf as determined by Nile Red fl uorescence ( Fig. 1 ). In contrast, starved zebrafi sh did not form adipocyte neutral lipid droplets through 15 dpf (data not shown), indicating that an exogenous nutrient supply is required for the formation of adipocyte neutral lipid droplets.
To determine the stage of zebrafi sh adipocyte development at which the nutrient supply is required, we examined expression of the preadipocyte/adipocyte marker fabp11a in fed and starved zebrafi sh using WISH. By 8 dpf, fed zebrafi sh formed fabp11a -positive preadipocytes and adipocytes in association with the pancreas ( Fig. 6A , B ; 4.3 ± 0.1 mm SL). In contrast, zebrafi sh starved through 8 dpf ( Fig. 6D, E ; 3.7 ± 0.1 mm SL) or 10 dpf (data not shown; expression in any cells resembling adipocytes or preadipocytes from 1-16 dpf (data not shown). Consistent with previous reports, we detected fabp11b expression in retina during embryonic and early larval stages ( 57 ) and also observed a novel fabp11b expression domain in the gills beginning at 15 dpf (data not shown). These results indicate that zebrafi sh fabp11b is not expressed by the adipocyte lineage.
To identify potential developmental origins of the adipocyte lineage, we monitored fabp11a expression in zebrafi sh as a function of developmental stage using WISH. Consistent with previous reports, we observed fabp11a expression in the lens and diencephalon at 1 dpf ( Fig. 5A ) and in blood vessels by 2 dpf ( Fig. 5B ) ( 56 ). As early as 3 dpf, robust fabp11a expression was also detected in individual cells within the caudal hematopoietic tissue (CHT; Fig. 5C ), the site of hematopoiesis at this developmental stage ( 59 ). Individual cells expressing fabp11a were observed in association with the pancreas as early as 6 dpf ( Fig. 5D, E ). Based on the expression of fabp11a and position within the pancreas prior to the appearance of the fi rst terminally differentiated adipocytes in this location by 8 dpf, we infer that these cells are preadipocytes. Cells in this location continue to express fabp11a through 10 dpf (5.1 ± 0.2 mm SL; Fig. 5H ) and the onset of neutral lipid deposition ( Fig. 1D ). These results support the notion that fabp11a is expressed by preadipocytes as well as differentiated adipocytes in larval stages of zebrafi sh development ( Figs. 4, 5 ).
Different mammalian adipose depots can display distinct patterns of gene expression ( 2 ); therefore, we sought to determine if adipose depots in addition to the initial Fig. 4. Zebrafi sh adipocytes express pparg and fabp11a . Individual zebrafi sh were stained with Nile Red at 15 dpf, imaged, and then processed for WISH using riboprobe directed against zebrafi sh pparg , fabp11a , or fabp11b mRNA. Cells labeled by WISH stain purple, in contrast to the brown melanin pigment contained within melanophores. Nile Red and WISH staining patterns were compared with detect colocalization. Whole-mount (A, B, D, E, G, H) and transverse cryosections (C, F) from the trunk of the same individuals are shown. Both fabp11a and pparg mRNA colocalize with neutral lipid droplets within visceral adipocytes associated with the pancreas (black arrowheads). fabp11a mRNA is also observed in nearby cells lacking neutral lipid droplets (putative preadipocytes; white arrows in E) and blood vessels (black arrows in E), while pparg mRNA is also found in the intestinal epithelium (black arrow in C). G, H: In contrast, visceral adipocytes containing neutral lipid droplets (black arrowheads in G) did not express fabp11b mRNA. Swim bladder (sb), gall bladder (gb), intestine (in), and pancreas (pan) are indicated. Dorsal is to the top in all panels and anterior to the right in A, B, D, and E. Bars = 100 m in A, B, D, E, G, and H and 50 m in C and F.

A zebrafi sh model for adipocyte research
Our current understanding of adipocyte biology is largely derived from rodent models and cell culture platforms. Although these model systems have been instrumental in revealing molecular mechanisms governing adipocyte differentiation and physiology, many important aspects of adipocyte development and adipose depot formation remain poorly understood. In this study, we provide a foundation for using the zebrafi sh model to address these gaps in our knowledge of adipogenesis. In vivo imaging and in situ hybridization analyses establish the temporal and spatial pattern of preadipocyte and adipocyte development in the zebrafi sh. Furthermore, we demonstrate that an exogenous nutrient supply is required for adipocyte development and that mature adipocytes are able to mobilize and deposit lipid as a function of nutrient availability. The anatomical and molecular features of zebrafi sh adipocytes are homologous to those of mammalian white adipocytes, suggesting that results obtained in the zebrafi sh model can be translated to mammals. This is consistent with the notion that teleost fi shes lack brown adipocytes, which evolved with homeothermy and nonshivering thermoregulation in mammals ( 2 ). We anticipate that the optical transparency of the zebrafi sh allowing 3.7 ± 0.1 mm SL) did not develop any fabp11a -expressing cells in the pancreas, suggesting that exogenous nutrition is required for the normal development of visceral preadipocytes. Interestingly, zebrafi sh starved through 8 dpf did possess supernumerary fabp11a -expressing cells in the CHT and the heart ( Fig. 6F, I, J ), regions typically devoid of fabp11a -expressing cells in age-matched fed animals ( Fig. 6C, G, H ).
To determine if the effect of starvation on adipocyte development is reversible, we began feeding starved zebrafi sh at 8 or 10 dpf and monitored formation of adipocyte lipid droplets using Nile Red ( Fig. 6K; data not shown). Animals starved through 8 or 10 dpf lacked fabp11a -expressing visceral preadipocytes at the initiation of feeding ( Fig. 6A,  D; data not shown). However, 2 or 3 days of feeding was suffi cient to robustly establish neutral lipid droplets in visceral adipocytes in zebrafi sh starved through 8 dpf ( Fig.  6K ; 3.8 ± 0.1 mm SL) or 10 dpf (data not shown; 3.9 ± 0.1 mm SL), respectively. Visceral adipocytes that formed in starved animals after feeding exhibit characteristics of normal adipocytes, including expression of fabp11a ( Fig. 6L, M ) but not fabp11b (data not shown). In conclusion, we demonstrate that the absence of exogenous nutrition results in reversible abrogation of adipocyte lineage development resulting in a reduction of preadipocytes and adipocytes associated with the pancreas. high-resolution in vivo imaging, combined with its amenability to genetic and chemical screens, will lead to the identifi cation of novel factors and mechanisms governing adipocyte development and physiology.

Markers of the zebrafi sh adipocyte lineage
The utility of the zebrafi sh to the analysis of adipocyte biology will depend heavily on our ability to visualize adipocytes in situ. This study uses the vital fl uorescent lipid probe Nile Red to visualize neutral lipid depots in living zebrafi sh. Although Nile Red is fl uorescent when incorporated into any hydrophobic lipid environment, its excitation and emission wavelength maxima are reduced when incorporated into neutral lipid ( 42 ). Visualization of Nile Red staining using a long-pass GFP fi lter permits the visualization of individual neutral lipid droplets by their yellow fl uorescence emission, while the polar lipid throughout the rest of the animal emits red fl uorescence ( Figs. 1, 3, 4, 6 ). Nile Red and other fl uorescent lipid probes should continue to be useful tools for in vivo monitoring of lipid dynamics in adipocytes and other cell types in the zebrafi sh.
Nile Red allows visualization of adipocytes containing neutral lipid droplets, however visualization of the adipocyte lineage prior to the onset of neutral lipid accumulation requires additional markers. This study is the fi rst to identify pparg as a marker of adipocytes during initial stages of zebrafi sh adipogenesis ( Fig. 4A-C ). Previous in vitro analysis in other teleost species (i.e., Pagrus major ) revealed that pparg is expressed in cultured preadipocytes as well as adipocytes ( 34 ). Although we are unable to exclude the possibility that zebrafi sh preadipocytes express pparg below our applied level of detection, zebrafi sh pparg appears to be robustly expressed only in mature adipocytes after the onset of neutral lipid deposition. This study is Fig. 6. Zebrafi sh adipogenesis is regulated by exogenous nutrition. Whole-mount preparations (A, D, G, I) and transverse cryosections through the pancreas (B, E), heart (C, F), or CHT (H, J) of 8 dpf zebrafi sh that were either fed since 5 dpf (fed; A-C, G-H) or never fed (starved; D-F, I-J) and then processed for WISH using fabp11a riboprobe to reveal the location of preadipocytes (black arrowheads). Cells expressing fabp11a mRNA stain purple, in contrast to the brown melanin pigment contained within melano-Although the development of adipose depots occurred in an ordered manner, the mobilization of adipose depots in response to starvation was qualitatively uniform across depots ( Fig. 3 ). This indicates that the mechanisms underlying the temporal order of adipose depot formation do not strongly infl uence the order in which adipose depots are depleted during periods of nutrient deprivation. Similarly, fat deposition in response to refeeding occurred in a qualitatively uniform manner across depots ( Fig. 3 ). This suggests that starvation for 7 days does not signifi cantly affect the distribution of adipose depots in adult zebrafi sh. However, future analyses will be required to defi ne the molecular and cellular consequences of starvation on adipose depots.

Nutritional regulation of zebrafi sh adipocyte development
Early postnatal dietary manipulations can contribute signifi cantly to adipocyte development and predisposition to metabolic disorders in mammals ( 13,14 ); however, the mechanisms underlying these events remain unclear. When zebrafi sh are fed beginning at 5 dpf, fabp11a -expressing preadipocytes appear in the pancreas as early as 6 dpf ( Fig. 5 ) and begin to accumulate neutral lipid as early as 8 dpf ( Fig. 1 ). We observed that fabp11aexpressing pancreatic preadipocytes fail to form in zebrafi sh larvae starved through 8 or 10 dpf, suggesting that the exogenous nutrient supply is an important regulator of adipogenesis in zebrafi sh as well as mammals ( Fig. 6 ). Interestingly, the lack of fabp11a -expressing preadipocytes in the pancreas of fi sh starved through 8 dpf was accompanied by an elevated number of cells expressing fabp11a in the CHT and heart. It remains unclear whether the fabp11a -expressing cells in the CHT and heart represent preadipocytes or other cell types. The mammalian homologs of zebrafi sh fabp11a (i.e., Fabp4 , Fabp5 , Fabp8 , and Fabp9 ) are expressed in several different cell types, including macrophages and dendritic cells ( 52 ), and it is possible that zebrafi sh fabp11a is also expressed in these lineages.
Based on our data, we propose two models for the developmental origins and nutritional regulation of visceral adipocytes. In the fi rst model, adipocytes that initially appear in the pancreas are derived from fabp11a -negative adipoblast precursors that are also located in the pancreas or in the adjacent stroma or vasculature. In response to exogenous feeding, these local adipoblasts could differentiate into fabp11a -expressing visceral preadipocytes and adipocytes. Consistent with this notion, previous studies in mammals have identifi ed adipogenic MSCs residing in adipose tissue stroma and vasculature (9)(10)(11). In the second model, adipocytes that initially appear in the pancreas are derived from fabp11a -positive preadipocytes or adipoblasts that originate in the CHT. These CHT-associated precursors might be mobilized into circulation in response to exogenous feeding and then colonize and complete differentiation within the pancreas. This model is consistent with previous studies establishing that the mammalian bone marrow is a rich source of MSCs ( 8 ). The CHT is the also the fi rst to analyze zebrafi sh fabp11a expression during adipogenesis, and we fi nd that fabp11a is expressed by mature adipocytes in multiple anatomic depots ( Figs. 4,5 ). We also fi nd that fabp11a is expressed by cells in the pancreas that lack neutral lipid droplets, and we speculate that these fabp11a -positive cells may be preadipocytes. In contrast, we did not detect expression of the paralogous gene fabp11b in zebrafi sh adipocytes, however we cannot exclude the possibility that fabp11b is expressed by adipocytes at later developmental stages or at levels below the sensitivity of our WISH assays. Lack of expression of the paralogous fabp11b gene by zebrafi sh adipocytes, combined with the fact that mammalian homologs Fabp4 and Fabp5 are expressed by adipocytes ( 53,58 ), suggests that the Fabp homolog in the last common ancestor of fi sh and mammals was also expressed in adipocytes. Subsequent gene duplication and subfunctionalization in derivative vertebrate lineages would have allowed for adipocyte expression to be lost in a subset of duplicated genes (i.e., zebrafi sh fabp11b and mammalian Fabp8 and Fabp9 ).

Formation of zebrafi sh adipose depots
Our results establish that zebrafi sh adipocytes develop in several anatomic depots. Mammals also form distinct adipose depots with distinct molecular features and relationships to metabolic disease susceptibility ( 2 ). The mechanisms underlying this developmental and physiological heterogeneity remain largely unknown and might represent useful targets for depot-specifi c therapeutic interventions. The appearance of adipose depots during zebrafi sh development occurs in an ordered manner, with the fi rst and largest depot forming in the viscera beginning at approximately 8 dpf ( Figs. 1, 3 ). Although fabp11a and pparg can serve as a marker of zebrafi sh adipocytes, their expression in other cell types (e.g., blood vessels, heart, and intestine; Figs. 4, 5 ) demands the identifi cation of additional specifi c molecular markers of zebrafi sh adipocytes. These tools will be essential for defi ning the heterogeneity between and within zebrafi sh adipose depots and interpreting results from genetic and chemical tests aimed at manipulating zebrafi sh adipogenesis.
A striking feature of zebrafi sh adipogenesis is the initial formation of adipocytes within the pancreas. Similar pancreatic localization of initial adipose depots has been observed during ontogeny of other teleost species ( 60 ), however the mechanisms underlying this developmental pattern remain unknown. One possibility is that a local signal produced in the pancreas is able to promote adipocyte differentiation. For example, insulin is known to be a potent pro-adipogenic factor in fi sh as well as mammals ( 33,61 ), and high local concentrations of insulin around the pancreas during early larval stages may potentially stimulate adipogenesis in nearby precursors. Another nonexclusive possibility is that the adipocyte precursors located in or around the pancreas might be especially sensitive to a systemic proadipogenic signal. As discussed below, additional knowledge about the developmental origins and nutritional regulation of these early visceral adipocytes will help distinguish between these possibilities. site of hematopoiesis in the zebrafi sh larvae and contains hematopoietic cells as well as undefi ned cells with fi broblast morphology ( 59 ). It is therefore possible that zebrafi sh MSCs reside within the CHT where they produce adipocyte precursors that can be mobilized and recruited to peripheral locations upon appropriate stimulation. This is congruous with previous reports identifying multipotent adipogenic precursors in circulating blood ( 12 ). Importantly, the effects of starvation on adipogenesis were reversible: feeding starved animals for 2 days was suffi cient to form pancreatic adipocytes ( Fig. 6K ). This suggests that adipocyte precursors, whether they reside within the CHT or within the viscera, remain competent to respond to nutritional cues by differentiating within the pancreas. We anticipate that lineage tracing analysis of zebrafi sh adipocyte precursors will help distinguish between these models.
Zebrafi sh research has traditionally focused on embryonic and early larval stages of development prior to the onset of exogenous feeding; consequently, very little is known about the roles of exogenous nutrition on zebrafi sh larval development and physiology. Our results establish that adipocyte development depends on the availability of an exogenous nutrient supply as early as 8 dpf, only 3 days after fi rst feeding. It is likely that other developmental and physiological processes are also sensitive to the nutrient supply during early feeding stages. For example, we have previously shown that an exogenous nutrient supply is required as early as 6 dpf to facilitate robust metabolic and innate immune responses to the community of commensal microorganisms residing in the intestine (microbiota) ( 21 ). Interestingly, the intestinal microbiota has been identifi ed as an important environmental factor that regulates fat storage in adult mice. The presence of a microbiota stimulates fat deposition via microbial processing of dietary nutrients and regulation of host metabolism ( 62,63 ). We anticipate that the zebrafi sh model will be a useful tool to defi ne the respective roles of the diet, the microbiota, and other environmental factors on early stages of adipocyte development.