Transcriptomic and metabolomic profiling of chicken adipose tissue in response to insulin neutralization and fasting
- First Online:
- 6.2k Downloads
Domestic broiler chickens rapidly accumulate adipose tissue due to intensive genetic selection for rapid growth and are naturally hyperglycemic and insulin resistant, making them an attractive addition to the suite of rodent models used for studies of obesity and type 2 diabetes in humans. Furthermore, chicken adipose tissue is considered as poorly sensitive to insulin and lipolysis is under glucagon control. Excessive fat accumulation is also an economic and environmental concern for the broiler industry due to the loss of feed efficiency and excessive nitrogen wasting, as well as a negative trait for consumers who are increasingly conscious of dietary fat intake. Understanding the control of avian adipose tissue metabolism would both enhance the utility of chicken as a model organism for human obesity and insulin resistance and highlight new approaches to reduce fat deposition in commercial chickens.
We combined transcriptomics and metabolomics to characterize the response of chicken adipose tissue to two energy manipulations, fasting and insulin deprivation in the fed state. Sixteen to 17 day-old commercial broiler chickens (ISA915) were fed ad libitum, fasted for five hours, or fed but deprived of insulin by injections of anti-insulin serum. Pair-wise contrasts of expression data identified a total of 2016 genes that were differentially expressed after correction for multiple testing, with the vast majority of differences due to fasting (1780 genes). Gene Ontology and KEGG pathway analyses indicated that a short term fast impacted expression of genes in a broad selection of pathways related to metabolism, signaling and adipogenesis. The effects of insulin neutralization largely overlapped with the response to fasting, but with more modest effects on adipose tissue metabolism. Tissue metabolomics indicated unique effects of insulin on amino acid metabolism.
Collectively, these data provide a foundation for further study into the molecular basis for adipose expansion in commercial poultry and identify potential pathways through which fat accretion may be attenuated in the future through genetic selection or management practices. They also highlight chicken as a useful model organism in which to study the dynamic relationship between food intake, metabolism, and adipose tissue biology.
KeywordsMicroarray Chicken adipose tissue Fasting Insulin neutralization Fatty acid metabolism Glucose metabolism Adipogenesis
Very low density lipoprotein
CCAAT: Enhancer-binding protein alpha
Peroxisome proliferator-activated receptor
Insulin receptor β-subunit
Insulin receptor substrate-1
Src homology 2 domain-containing substrate
False discovery rate
Platelet derived growth factor
Mesenchymal stem cell
Cyclic adenosine monophosphate
Diacylglycerol O-acyltransferase homolog 2
Dehydrogenase kinase 4
Acyl-Coenzyme A binding domain containing 5
Type I angiotensin II receptor
Polyunsaturated fatty acids
Non-esterified fatty acid
Adipose triglyceride lipase
Liver X receptor
Preproglucagon class B transcript
Branched chain amino acids.
The domestic chicken provides a widespread and relatively inexpensive source of dietary protein for humans. In addition to its role as a food animal, the chicken has a long history as a valuable model research organism. These dual considerations led to the selection of chicken as the first agricultural animal model to be sequenced at the genome level. While chickens have been used heavily for studies of developmental biology and immunology, a number of traits make them a viable model for studies of adipose biology, obesity and insulin resistance. Commercial broiler chickens, in particular, rapidly accumulate excess adipose tissue as a result of genetic selection for growth and are considered “obese” relative to leaner egg-laying or wild strains of chickens(rev. in). Chickens mimic the early stage of type 2 diabetes in humans, exhibiting both hyperglycemia (up to 200 mg/dL in the fasting state) and resistance to exogenous insulin[4, 5]. Like humans, but unlike rodents or pigs, chickens rely on liver rather than adipose tissue for the majority of de novo lipid synthesis[6, 7, 8]. Most metabolic genes are conserved with humans, and a number of the quantitative trait loci (QTLs) that have been linked to fatness in chickens contain genes implicated in human susceptibility to obesity or diabetes. Chickens also represent a model for studying mechanisms of adipocyte hyperplasia during development, a process that may exacerbate adult obesity. During at least the first several weeks after hatch, chicken adipose tissue expands more through adipocyte hyperplasia than hypertrophy, and an early increase in adipocyte number is a common feature of some lines genetically selected for excess adiposity[10, 11]. Finally, the egg presents opportunities to directly manipulate the developmental milieu and study the consequences on adipose metabolism via in ovo injection.
Relatively little is known about regulation of adipose tissue deposition and metabolism in chicken. Because of its relative importance in lipogenesis, most studies have focused on the role of liver in adipose expansion. Several genetic lines of fat and lean chickens have been developed through phenotypic selection, most of which have both elevated plasma levels of very low density lipoprotein (VLDL) and lower levels of plasma glucose, reflecting the importance of hepatic lipogenesis and glucose consumption in fat accretion. Reciprocally, phenotypic selection for low plasma glucose simultaneously selects for fatness. Both chicken and mammalian adipocytes develop through a sequence of molecular triggers including activation of CCAAT-enhancer-binding protein alpha (CEBPα) and peroxisome proliferator-activated receptor gamma (PPARγ). A clear point of divergence, however, is their responsiveness to insulin. Unlike in mammals, insulin has minimal effect on glucose uptake in chicken adipose tissue. In fact, an avian homolog of the insulin-sensitive glucose transporter GLUT4 has not been identified in the current chicken genome database. Insulin does, however, stimulate uptake of acetate, which is the preferred substrate for de novo lipogenesis in chicken adipocytes, although the magnitude of the effect is relatively modest. Insulin signaling appears to proceed through tissue specific cascades in chicken metabolic tissues. In liver, insulin elicits a signaling cascade that parallels the response in mammals, including tyrosine phosphorylation of insulin receptor β-subunit (IRβ), insulin receptor substrate-1 (IRS-1) and Src homology 2 domain-containing substrate (Shc) and activation of phosphatidylinositol 3-kinase (PI3K)[16, 17]. The situation in skeletal muscle is more complex. Tyrosine phosphorylation of IRβ and IRS-1 and PI3K activity are not regulated by insulin, whereas events downstream of PI3K (e.g. Akt and P70S6K activation) are accordingly sensitive. We recently reported that insulin also does not elicit a classical IRβ initiated cascade in chicken adipose tissue, including the downstream steps of Akt and P70S6K activation. Insulin also does not inhibit lipolysis in chicken adipose tissue; glucagon, is the primary lipolytic hormone (rev. in).
In the present study we simultaneously characterized the effects of a short term (5 hours) fast or neutralization of insulin action (5 hours) on adipose tissue of young (16–17 day-old), fed commercial broiler chickens. The goals of this study were two-fold. First, we sought to identify pathways activated by feed restriction, reasoning that they may highlight potential strategies for control of fatness through either genetic selection or improved management practices. Simultaneously, we sought to understand the contribution of insulin, if any, into chicken adipose physiology. No experimental model of diabetes exist in chicken: total pancreatectomies are not achievable, and alloxan and streptozotocin are inefficient at destroying pancreatic chicken beta-cells (rev. in). The two treatments were compared to distinguish potential insulin-specific changes from those that could be mimicked by fasting through changes in nutrient availability. Both treatments were shown previously to elicit significant alterations in several plasma metabolic and endocrine parameters; in the studies reported herein, samples of abdominal adipose tissue were issued from the same experiment. Tissue metabolomics was combined with microarrays to bridge the gap between gene expression, metabolic and physiological responses, and to identify the composite effects of both fasting and insulin deprivation on chicken adipose tissue.
Fold change verification of gene expression by RT-PCR
Microarray fold change
Quantitative PCR fold change
Fasting vs. Fed
Ins-Neu vs. Fed
Ins-Neu vs. Fasting
Fasting vs. Fed
Ins-Neu vs. Fed
Ins-Neu vs. Fasting
F-box protein 8
Dual specificity phosphatase 5
BCL2/adenovirus E1B 19 kDa interacting protein 3
Pre-B-cell leukemia homeobox 3
Interleukin 10 receptor, beta
Early growth response 1
NGFI-A binding protein 1 (EGR1 binding protein 1)
Pyruvate dehydrogenase kinase, isozyme 4
Angiotensin II receptor, type 1
Gene ontology (GO) and KEGG annotation for representative clusters of differentially expressed genes
GO term (Biological Process, level 6 or 7) or KEGG pathway name
1, 3 and 4 (731 genes)
Positive regulation of protein metabolic process
Negative regulation of cellular macromolecule biosynthetic process
Triglyceride metabolic process
Negative regulation of gene expression
Proteolysis involved in cellular protein catabolic process
Negative regulation of kinase activity
Regulation of transcription, DNA-dependent
Antigen receptor-mediated signaling pathway
Regulation of phosphate metabolic process
Regulation of kinase activity
Regulation of I-kappaB kinase/NF-kappaB cascade
Ubiquitin mediated proteolysis
PPAR signaling pathway
Fatty acid metabolism
2 (557 genes)
Monosaccharide catabolic process
DNA dependent DNA replication
Hexose metabolic process
Glucose metabolic process
Regulation of cell shape
Nucleoside triphosphate metabolic
Glycolysis / Gluconeogenesis
Citrate cycle (TCA cycle)
6 (402 genes)
Glyoxylate and dicarboxylate metabolism
5 and 7 (250 genes)
Enzyme linked receptor protein signaling pathway
Regulation of cellular protein metabolic process
Negative regulation of macromolecule metabolic process
Transmembrane receptor protein serine/threonine kinase signaling pathway
Negative regulation of protein metabolic process
Positive regulation of cellular metabolic process
PPAR signaling pathway
MAPK signaling pathway
Global biological responses to fasting and to insulin neutralization were further characterized using KEGG pathway matching, based on genes with statistically significant differential expression (FDR < 0.05) and absolute fold-change ≥1.5. Genes altered exclusively by fasting represented a wide range of cellular pathways, indicating significant effects of even a five hour fast on adipose function and metabolism in chicken. Fasting exerted significant effects on pathways related to carbohydrate, amino acid and lipid metabolism and synthesis. Within the categories related to lipid metabolism, fasting up-regulated expression of genes involved in fatty acid oxidation (e.g., acyl-CoA oxidase 1 (ACOX1), acetyl-CoA carboxylase beta (ACACB), carnitine palmitoyltransferase 1A (CPT1A)) and down-regulated expression of genes that control fatty acid, cholesterol and triacylglycerol synthesis (e.g., 1-acylglycerol-3-phosphate O-acyltransferase 9 (AGPAT9), ATP citrate lyase (ACLY), farnesyl diphosphate synthase (FDPS), acetyl-Coenzyme A carboxylase alpha (ACACA) and acetoacetyl-CoA synthetase (AACS)). Fasting also up-regulated expression of many genes involved in proteolysis and amino acid degradation. In addition to pathways highlighted by KEGG analysis, fasting down-regulated a number of genes (e.g., TGFβ, BMP) that mediate mesenchymal stem cell (MSC) commitment, an early step in the formation of new adipocytes ( Additional file1). Finally, a number of phosphodiesterases were up-regulated with fasting, presumably in response to the increased plasma glucagon and subsequent elevations in cyclic adenosine monophosphate (cAMP; Additional file1). Collectively, these categories indicate that chicken adipose tissue responds to a relatively short duration (five hour) fast with sweeping changes in gene expression that suppress synthesis and storage of lipids and other macromolecules and up-regulate mobilization and metabolism of fatty acids and proteins.
Shared effects of fasting and insulin-neutralization on differential gene expression
Fold change Insu-Neu vs. Fed
Fold change Fasting vs. Fed
pyruvate dehydrogenase kinase, isozyme 4
aldehyde oxidase 1
pleckstrin homology domain containing, family H (with MyTH4 domain) member 2
F-box protein 8
angiotensin II receptor, type 1
protein-L-isoaspartate (D-aspartate) O-methyltransferase domain containing 1
proteasome (prosome, macropain) activator subunit 4
islet cell autoantigen 1, 69 kDa
inositol hexakisphosphate kinase 2
ubiquitin-like, containing PHD and RING finger domains, 2
acyl-Coenzyme A binding domain containing 5
similar to hypothetical protein
zinc finger protein 217
Interferon (alpha, beta and omega) receptor 1
diacylglycerol O-acyltransferase homolog 2 (mouse)
endonuclease/exonuclease/phosphatase family domain containing 1
ankyrin repeat domain 9
dihydrolipoamide S-succinyltransferase (E2 component of 2-oxo-glutarate complex
protein tyrosine phosphatase type IVA, member 3
heat shock 70 kDa protein 5 (glucose-regulated protein, 78 kDa)
nucleolar protein family A, member 2 (H/ACA small nucleolar RNPs)
macrophage stimulating 1 receptor (c-met-related tyrosine kinase)
GRAM domain containing 2
dehydrodolichyl diphosphate synthase
dynein, light chain, LC8-type 2
chromatin licensing and DNA replication factor 1
fumarylacetoacetate hydrolase domain containing 1
DOT1-like, histone H3 methyltransferase (S. cerevisiae)
BTB (POZ) domain containing 11
Unique effects of insulin neutralization on differential gene expression
glucagon receptor precursor
t-complex 11 (mouse)-like 2
chaperonin containing TCP1, subunit 3 (gamma)
membrane associated guanylate kinase, WW and PDZ domain containing 1
SET domain containing (lysine methyltransferase) 7
suppression of tumorigenicity 13 (colon carcinoma) (Hsp70 interacting protein)
LSM14A, SCD6 homolog A (S. cerevisiae); similar to LSM14 homolog A (SCD6, S. cerevisiae)
AHA1, activator of heat shock 90 kDa protein ATPase homolog 1 (yeast)
target of EGR1, member 1 (nuclear)
antizyme inhibitor 1
Fold change of fasting and insulin neutralization on adipose tissue metabolites
Correlated patterns of gene expression and metabolite abundance were extracted using hierarchical clustering to interconnect treatment effects on transcripts and metabolites. Clusters 2 and 3 contained genes and metabolites with lower abundance in fasted vs. fed or insulin neutralized tissue. The two clusters differed with respect to the insulin neutralized group: cluster 3 contained analytes at intermediate levels between fasted and fed, while cluster 2 contained those at levels comparable to or greater than fed. Twelve of the 17 metabolites with statistically suggestive or significant effects of treatment, including all of the amino acids and amino acid derivatives ( Additional file4), were present in cluster 2 along with a set of genes that included the p85α regulatory subunit of PI3 kinase (PIK3R1), as well as ME, malonyl CoA decarboxylase and ELOVL6. Cluster 3 contained several metabolites including both NAD + and NADPH and was significantly enriched in GO annotations related to carbohydrate metabolism and in the KEGG pathways TCA cycle, glycolysis/gluconeogenesis, pyruvate metabolism and steroid biosynthesis (p < 0.05; Additional file4). Clusters 7 and 8 consisted of genes and metabolites with higher levels in fasted than in the other two treatment groups. These clusters were significantly enriched in GO categories PPAR signaling and negative regulation of cellular biosynthesis and also contained citrate and pyruvate (p < 0.05; Additional file4).
Despite roles as both a domestic food animal of worldwide economic importance and a widely used model organism with relevance for human obesity and insulin resistance, few studies have examined regulation of gene expression in chicken adipose tissue. To our knowledge, no studies of nutritional regulation of chicken adipose tissue at the genomic level have been reported in the published literature. Likewise, although insulin is the most well-defined hormonal mediator of metabolism in mammalian adipose tissue, its role in chicken remains to be clarified. Therefore the current study addressed two objectives: 1) characterize the transcriptomic and metabolomic response to energy manipulation as a step toward enhanced understanding of adipose biology in chicken; and 2) identify the effects of insulin on chicken adipose tissue by including a group of birds in which insulin action was blocked by immunoneutralization with an anti-insulin antibody. We sought to both identify potential new targets for genetic selection or management strategies to reduce fat accumulation in commercial broilers and to further develop chicken as a model organism for studies of human obesity.
Although intrinsic lipogenic activity is low in chicken adipose tissue, genes involved in fatty acid synthesis and storage were suppressed and those in fatty acid mobilization and oxidation were up-regulated by fasting. The 40 down-regulated genes with fold changes greater than three were significantly enriched for the GO annotation lipid biosynthetic process (FDR <0.05), including genes that control triglyceride synthesis (DGAT2 and AGPAT9) and fatty acid synthesis (ACACA, ACLY and ME), elongation (ELOVL6), and desaturation (FADS1). AGPAT9 and DGAT2 catalyze the initial and final steps, respectively, of de novo triglyceride synthesis. ACLY is the main enzyme for synthesis of cytosolic acetyl-CoA, which is carboxylated to malonyl-CoA by ACACA, the rate-limiting step in fatty acid synthesis. Reducing equivalents for the conversion of malonyl-CoA to palmitate are supplied by malic enzyme (ME). ELOVL6 catalyzes elongation of palmitate to stearate and appears to play a key role in insulin sensitivity[30, 31]. Finally, FADS1 is rate-limiting for polyunsaturated fatty acids (PUFA) biosynthesis and was recently implicated in control of fasting glucose homeostasis in humans. Genes altered by fasting in adipose tissue in this study overlapped with those shown to be differentially expressed in chicken liver after 16 or 48 hours of fasting, including ACLY, ACOX1, BCAT1 and PDK4. These authors used a different array platform than ours, which precludes precise quantitative comparisons. However, among the genes changed in both studies, the fold changes observed in adipose tissue were consistently greater than those in liver, despite the longer duration of fasting in that study. For example, PDK4 expression was up-regulated ~ 18-fold by a five hour fast in adipose tissue, but only ~ 1.5-fold after a 16 hour fast in liver. While differences in sensitivity between the two array platforms must be kept in mind, these data suggest that adipose tissue metabolism in chicken is at least as sensitive to energy status as hepatic metabolism. Our results indicate that both fatty acid synthesis and storage are dynamically regulated by energy status in chicken adipose tissue, despite its modest (~ 15%) contribution to the amount of stored fatty acids.
Both fasted and insulin-neutralized birds exhibited significant increases in plasma glucagon. Parallel elevations in plasma NEFA suggested that this resulted in significant lipolysis of stored triacylglycerol in both treatment groups. During fasting, a considerable percentage of the liberated fatty acids are re-esterified in adipocytes, and only a small fraction traditionally have been thought to be oxidized in the mitochondria of adipocytes through beta oxidation. However, recent studies in mice and in human adipose tissue demonstrate that in some conditions fatty acid oxidation in white adipose tissue is considerable and may be an important determinant of obesity[35, 36, 37]. Consistent with this concept, we found significant increases in a number of key enzymes that mediate mobilization of fatty acids and their oxidation, including the rate-limiting enzymes in both mitochondrial and peroxisomal fatty acid oxidation (CPT1A and ACOX1, respectively). We measured tissue levels of beta-hydroxybutyrate, a ketone product of beta oxidation, to confirm that changes in gene expression had functional consequences and found them to be significantly elevated in adipose tissue of fasted vs. fed chickens. Levels were numerically but not statistically higher in insulin-neutralized adipose tissue (data not shown). Qualitatively, fasting-induced changes in gene expression resemble those induced by the fibrate class of drugs, which activate PPARα and promote fatty acid oxidation in white adipose tissue and are used clinically to treat hyperlipidemia[35, 37, 38, 39, 40]. These data suggest that dietary activation of PPARα, for example through supplementation with fatty acids that preferentially bind and activate this member of the PPAR family, may be a means to attenuate fat deposition in commercial broilers. Such action may underlie the reduced abdominal fat mass reported in broilers that were fed diets rich in n-3 PUFA.
Both fasting and insulin neutralization elicited marked upregulation of PDK4. PDK4 is a nutrient sensing fuel switch that phosphorylates and inactivates pyruvate dehydrogenase, which shifts fuel use from glucose to fatty acids and spares glucose for the brain during periods of fasting[23, 43, 44, 45]. PDK4 also enhances glycerol synthesis in white adipose tissue by shunting pyruvate into glyceroneogenesis, at least in the fed state. Hepatic and skeletal muscle expression of PDK4 is increased by fatty acids, acetyl CoA, NADH and the diabetic state and decreased by insulin and pyruvate (rev. in[23, 47]). Little is known about PDK4 in chicken, but a recent study suggests it acts as a glycogen sensor in muscle and thus plays comparable roles to those in mammals. In mouse white adipose tissue, PDK4 expression was shown to be induced by activation of p38MAPK, which we found to be significantly up-regulated with fasting and, to a lesser extent, with insulin neutralization. Although PDK4 was up-regulated in both treatment groups, and both groups showed evidence of increased lipolysis, only fasted chickens presented a gene expression signature and tissue beta-hydroxybutyrate levels that were clearly indicative of fatty acid oxidation. Although we did not measure malonyl-CoA levels, we predict that they were reduced with fasting, but not insulin neutralization, based on reduced expression of ACACA. Malonyl-CoA allosterically binds and inhibits CPT1A, minimizing fatty acid transport and subsequent oxidation in mitochondria. With insulin neutralization, increased PDK4 may thus be more aligned with the demand for glycerol needed to re-esterify fatty acids liberated by lipolysis. Additional experiments are needed to confirm that manipulation of PDK4 alters fatty acid oxidation in chicken adipose tissue and to delineate its relative contributions to fatty acid oxidation and glyceroneogenesis under varying metabolic states. If manipulation of PDK4 does alter fatty acid oxidation, our results highlight this pathway as a potential target for reducing fatness, which has relevance for both poultry and humans.
Microarray data indicate that the effects of fasting in chicken adipose tissue extend beyond metabolism. GO analysis highlighted pathways such as cell cycle and cytokine-cytokine receptor interaction that are most likely related to changes in the stromal vascular fraction, which contains proliferating preadipocytes and cells of the immune system. In particular, a number of genes that regulate multiple steps in adipogenesis were significantly altered by fasting. Chickens rapidly accumulate abdominal fat after hatch, and until approximately 7 weeks of age this is due more to formation of new adipocytes than to adipocyte hypertrophy. Adipocytes arise from mesenchymal stem cells in a two stage process of lineage commitment to an adipocyte fate, followed by differentiation of fibroblast-like preadipocytes into mature fat-storing cells. Members of both the Wnt (MSC lineage commitment) and TGFβ/BMP (MSC lineage commitment and preadipocyte competence) signaling pathways were significantly regulated by fasting. Fasting down-regulated expression of CEBPα and PPARγ, two transcription factors that orchestrate the cascade of gene expression changes that lead to terminal adipocyte differentiation. Expression of other adipogenic mediators including fibroblast growth factor 2 (FGF2), fibroblast growth factor receptor 1 (FGFR1), and nuclear receptor corepressor 1 (NCOR1) were also significantly regulated by fasting. Collectively, these changes suggest that adipocyte number in chickens is dynamically tied to energy status, at least in young chicks (such as those used herein) that are rapidly forming new adipocytes. An elegant study by Arner et al. concluded that adipocyte number in humans is a major determinant of adult fat mass and is determined during early childhood. Less is known about this process in humans due to the limitations of sampling adipose tissue, particularly during development and from different abdominal depots. In light of what appears to be sensitive regulation of adipogenesis by nutritional state, chickens may thus be particularly valuable models in which to elucidate mechanisms of adipocyte hyperplasia during development that would inform the study of human obesity.
It is worth noting that, despite the uncertainty about insulin signaling in chicken adipose tissue, fasting altered the expression of several messengers encoding elements of the insulin signaling cascade. Expression of PIK3CB, which encodes the catalytic p110 subunit of PI3K, was up-regulated with fasting, while PIK3R1, which encodes the regulatory p85 subunit, was down-regulated. Such regulation could maintain some insulin signals despite a fall in plasma insulin level. CBLB and CRK, which mediate insulin signals that are associated with lipid rafts, were also up-regulated with fasting. In mammals, this pathway stimulates glucose uptake independently of PI3K activation, which may shed light on the apparent refractoriness of PI3K activity to insulin that was described in chicken skeletal muscle. Therefore, the potential effects of insulin on lipid storage and energy utilization appear to be defended in the fasting state, when insulin levels fall, by enhanced insulin sensitivity at the post-receptor level. Additional studies are needed to confirm this effect and to further explore the potential of PI3K-independent effects of insulin on glucose utilization in chicken adipose tissue.
Insulin is not considered to be a key regulator of glucose metabolism in chicken adipose tissue, although it does induce glucose disposal in chicken liver and muscle. It is therefore not surprising that the majority of genes significantly altered by both insulin neutralization and fasting are not related to glucose metabolism and lipid synthesis. The main exception is DGAT2, which catalyzes the final step in esterification of fatty acids into triglycerides. In fact, DGAT2 showed the most extreme down-regulation (6.1- and 10.5-fold, insulin-neutralized and fasted, respectively) in each treatment group, which is surprising considering that other genes related to lipogenesis were only regulated by fasting. Suppression of DGAT2 expression may be due to feedback by lipolysis, which appeared to be increased in both treatment groups based on plasma NEFA levels. In general, our data indicate that insulin deprivation altered fatty acid and glucose metabolism in a manner comparable to fasting but to a lesser extent, such that most genes involved in these pathways did not exhibit statistically significant changes in expression. For example, cluster analysis (Figure2) revealed that some genes upregulated by fasting were also increased by insulin neutralization (clusters 1, 3 and 4); these three clusters were enriched with genes in the KEGG pathways for fatty acid metabolism and PPAR signaling, including both ACOX1 and CPT1A, among others. Similarly, among genes that were downregulated by fasting, clustering discriminated a set of genes (Figure2, cluster 2) with a trend to also be decreased (albeit to a lesser extent than in fasting) by insulin deprivation. Interestingly, this cluster was significantly enriched in functions related to carbohydrate metabolism, suggesting that insulin does play some role in chicken adipose glucose metabolism. Comparable trends appeared in the metabolomic data. For example, stearate and palmitate (the only fatty acids measured by our MS platform) were lower (although not significantly) in both fasted and insulin neutralized compared to fed birds ( Additional file3). While the purpose of our study design was to determine the specific effects of insulin on chicken adipose tissue, we cannot exclude the possibility that some of the overlapping changes in gene expression were secondary to systemic factors, such as hyperglucagonemia present in both treatment groups. In vitro experiments using primary adipocytes or adipose explants will be useful to confirm specific effects of insulin on genes identified herein.
Of the 13 changes in expression that were unique to insulin neutralization, the most interesting responses were up-regulation of GCG, which encodes preproglucagon (fold change = 2.91), and down-regulation of the glucagon receptor (LOC425670, fold change = −2.77). The proglucagon system in avians is more complex than in mammals. The avian preproglucagon locus encodes two distinct precursor proteins that yield different peptides through alternative posttranslational processing: the class A transcript (PGA) yields glucagon and glucagon-like peptide-1 (GLP-1), while the class B transcript (PGB) additionally produces glucagon-like peptide-2 (GLP-2) and is more like the mammalian transcript (rev. in). Adipose tissue expresses both transcripts, with PGA being slightly more abundant, and is the third highest preproglucagon expressing tissue in chicken, behind pancreas and the proventriculus. We used transcript-specific QPCR to determine that only the PGB transcript was up-regulated by insulin neutralization (data not shown). Additional experiments are necessary to delineate which of the encoded peptides are up-regulated in parallel, but the coincident down-regulation of the glucagon receptor suggests a paracrine glucagon axis in chicken adipose tissue, and one that is regulated by insulin. In support of this concept, plasma glucagon (presumably derived largely from pancreas) was elevated comparably in both treatment groups, while GCG expression in adipose tissue was only up-regulated by insulin neutralization.
Tissue metabolomic analysis highlighted effects of insulin neutralization that were divergent from fasting and not readily apparent from microarray data. Most of the tissue amino acids that were measured were higher with insulin-neutralization but lower with fasting when each group was compared to ad libitum fed controls. This pattern parallels the levels of αNH2NPN levels in blood. Low levels in fasted adipose tissue were most likely due to oxidation of the carbon skeletons for cellular energy through the tricarboxylic acid cycle (TCA) cycle and/or for glyceroneogenesis, in the absence of dietary glucose. Increased amino acid catabolism was reflected in the differential expression profiles of the fasted vs. fed comparison (Figure3; Additional file1). In the insulin neutralized group, however, glucose supply from food was maintained and preferentially oxidized for energy. Elevated amino acids in the insulin neutralized group may also reflect reduced utilization due to the lack of insulin’s anabolic effects, particularly on the proliferating cell population within adipose tissue. The metabolomics approach used here measured only metabolite pool sizes at the time that tissues were harvested, rather than the effect of fasting or insulin neutralization on the rates of metabolism through glycolysis and the TCA cycle. The latter would be much more informative with respect to the dynamic impact of treatment, but requires the use of isotopic labeling (e.g., by feeding 13C-labelled glucose) which was not performed in this study. Nonetheless, we were able to demonstrate significant effects on some metabolites that inform the parallel changes in gene expression, particularly in relation to amino acid metabolism. Combined clustering of metabolomic and gene expression together identified a set of genes correlated with many amino acid levels, including PIK3R1, ME and MCD.
In summary, we determined that adipose tissue metabolism in the chicken is regulated by energy status and, to a lesser extent by insulin. Although adipose tissue is not a primary site of lipogenesis in chicken, the rate-limiting genes for fatty acid synthesis were suppressed by fasting. Likewise, fasting appeared to increase aspects of insulin sensitivity based on expression profiles, despite the view that chicken adipose tissue is relatively insensitive to insulin. Consistent with this paradigm, insulin neutralization significantly altered the expression of only a few genes related to glucose and lipid metabolism. Nonetheless, a considerable number of genes were altered by insulin neutralization, many of which thus far have unclear roles in adipose biology. Expression profiles suggest that even short term fasting alters fat storage in broilers by enhancing the oxidation of fatty acids. The initiating events that trigger upregulation of the corresponding genes are unclear, but there is considerable evidence for activation of PPARa, LXRa, and potentially other transcription factors that are activated by fatty acid ligands. Further studies are warranted to identify these triggers because of their potential impact on fat storage. Our data also suggest that broiler chicks may be an informative model organism in which to investigate dietary effects on adipose development in light of what appears to be a relationship between energy intake and adipogenesis. The results of this study thus have dual benefit for both the poultry industry and for studies of obesity in humans.
Male broiler chicks (ISA 915, Institut de Sélection Animale, Saint Brieuc, France) from which samples were collected for this study were hatched and raised under standard conditions, as originally described by Dupont and in accordance with the guidelines for Care and Use of Agricultural Animals in Agricultural Research and Teaching. Briefly, at 16–17 days of age, chicks of similar body weights were either allowed to continue feeding (ad libitum fed controls), fasted for five hours, or fed but injected at 0, 2 and 4 hours with porcine anti-insulin serum (insulin neutralized). Both the fed and fasted groups received injections of normal porcine serum as a vehicle control. Abdominal adipose tissue samples were harvested and rapidly snap-frozen in liquid nitrogen, pulverized, then stored at -80°C until analysis. Adipose samples from five birds in each group were used for both microarray and metabolomic analyses.
Total RNA was isolated from chicken adipose samples using the RNeasy Lipid kit and incorporating an on-column DNase treated with the RNase-free DNase Set according to the manufacturer's protocol (Qiagen.com). RNA quality and concentration were measured using the Experion System (Bio-Rad.com); only RNA passing recommended standards of quality was used for further studies. Transcriptome profiling was performed by Genome Quebec (Montreal, Canada) using the Affymetrix GeneChip Chicken Genome Array (San Diego, CA). Microarray data from this study are deposited in the Gene Expression Omnibus (GEO) under the accession number GSE35581. For real time RT-PCR validation, cDNA was synthesized using the iScript cDNA Synthesis kit (Bio-Rad.com). Commercially designed and validated primer sets (QuantiTect) and the associated SYBR Green master mix (Qiagen.com) were used to assay gene expression on a CFX96 real-time PCR detection system (Bio-Rad.com). All samples were analyzed in triplicate and normalized to ß-tubulin. Relative differences in gene expression were determined using the 2-ΔΔCT method and statistical differences were tested by analysis of variance (ANOVA).
Liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS)
Abdominal adipose tissue samples from five birds in each treatment group (the same five birds used for expression profiling) were extracted by placing tissue in a mortar containing liquid nitrogen and then powdering with a pestle. Portions (8–40 mg) of the powered tissue were weighed into 1.5 mL centrifuge tubes. Chilled methanol (0.5 mL at −80°C) and internal standard (5 μL of 1.7 mM benzoic acid in negative mode or 4.25 mM tris(hydroxymethyl)aminomethane in positive mode) were added to each tube. Each tube was mixed thoroughly by vortexing for two minutes, and the metabolites were extracted from the tissue for 30 min at 4°C. The tubes were then centrifuged (5 min, 4°C, 16.1 rcf) and supernatant (210 μL) was split into two autosampler vials. One of these samples was immediately placed on the LC-MS/MS for analysis, while the other was stored at −80°C for analysis in the opposite polarity ion mode on the following day.
Samples were placed in an autosampler tray chilled to 4°C, and 10 μL of each was injected onto an LC column for analysis. The chromatography method for positive ion mode was reported previously by Bajad and coworkers, with one exception that the column was cooled to 10°C. The chromatography method for negative ion mode was performed as reported by Waters and coworkers, except the gradient was allowed to run 50 min instead of 45 min to allow more thorough equilibration of the column. The eluent was introduced directly into the MS via an electrospray ionization (ESI) source fitted to a Finnigan TSQ Quantum Discovery Max triple quadrupole MS (Thermo Electron, Waltham, MA) through a 0.1 mm internal diameter fused silica capillary. The spray voltage was 4500 V in positive mode or 3000 V in negative mode. The sheath gas (nitrogen) was set to 40 psi, and the capillary temperature was set to 290°C. The collision cell gas (argon) was set to a pressure of 1.5 mTorr. Samples were analyzed using selected reaction monitoring (SRM) mode with a scan width of 1 m/z and a scan time of 0.05 s. The SRM parameters for most metabolites have been published previously. This method was used to scan for almost 300 metabolites. Xcalibur software (Thermo Scientific, Waltham, MA) was used to manually assess the elution time of the correct LC spectral peak for each metabolite-specific SRM. The Quan Browser utility in Xcalibur was then used to integrate the LC spectral peak area (in ion counts) for each detected compound, and these data were exported to a Microsoft Excel spreadsheet for further processing.
Statistical analysis of the microarray data was performed using R 2.9.0 and routines contained in Bioconductor (bioconductor.org). GC robust multi-array average (GCRMA) was used to normalize and scale the raw data from CEL files. The normalized data were filtered for low expression by removing any probes with normalized expression less than 3 in at least 5 arrays. Statistical significance of gene expression differences were analyzed by one-way ANOVA and empirical bayes using the limma package. Differential expression was defined based on false discovery rate (FDR) adjusted p-value <0.05. False discovery rate for differential expression and for GO and KEGG enrichment testing was controlled using the Benjamini-Hochberg method. Venn diagrams of differentially expressed genes were plotted to visualize the number of differentially expressed genes for each treatment comparison and their intersections. Hierarchical clustering of significant genes was performed using the hclust function and a hierarchical clustering heatmap was created using heatmap.2 in the gplots package. Hierarchical clustering also was used to identify correlated patterns of gene expression and metabolites. The Database for Annotation, Visualization and Integrated Discovery (DAVID, version 6.7) and ClueGO, a Cytoscape plug-in, were used for Gene Ontology (GO) at level 6 and 7 and KEGG analysis of differentially expressed genes[62, 63]. Statistical analysis of metabolomic data was performed using an analysis tool that we developed specifically for metabolomic data analyses (Ernest et al., manuscript in review). The script (metabR), written in the language R, uses linear mixed-effect modeling to normalize metabolomics data containing both fixed- and random-effect confounding variables. The script averages any replicate measurements (statistical sampling) made on experimental units and performs ANOVA to test for statistical differences between experimental groups.
- 2.Hillier LW, Miller W, Birney E, Warren W, Hardison RC, Ponting CP, Bork P, Burt DW, Groenen MAM, Delany ME, Dodgson JB, Chinwalla AT, Cliften PF, Clifton SW, Delehaunty KD, Fronick C, Fulton RS, Graves TA, et al: Sequence and comparative analysis of the chicken genome provide unique perspectives on vertebrate evolution. Nature. 2004, 432 (7018): 695-716. 10.1038/nature03154.CrossRefGoogle Scholar
- 5.Simon J: Chicken as a useful species for the comprehension of insulin action. Crit Rev Poultry Biol. 1989, 2: 121-148.Google Scholar
- 9.Nadaf J, Pitel F, Gilbert H, Duclos MJ, Vignoles F, Beaumont C, Vignal A, Porter TE, Cogburn LA, Aggrey SE, et al: QTL for several metabolic traits map to loci controlling growth and body composition in an F2 intercross between high- and low-growth chicken lines. Physiol Genomics. 2009, 38 (3): 241-249. 10.1152/physiolgenomics.90384.2008.CrossRefPubMedGoogle Scholar
- 16.Dupont J, Tesseraud S, Derouet M, Collin A, Rideau N, Crochet S, Godet E, Cailleau-Audouin E, Metayer-Coustard S, Duclos MJ, et al: Insulin immuno-neutralization in chicken: effects on insulin signaling and gene expression in liver and muscle. J Endocrinol. 2008, 197 (3): 531-542. 10.1677/JOE-08-0055.CrossRefPubMedGoogle Scholar
- 17.Dupont J, Derouet M, Simon J, Taouis M: Nutritional state regulates insulin receptor and IRS-1 phosphorylation and expression in chicken. Am J Physiol - Endocrinol Metab. 1998, 274 (2): E309-E316.Google Scholar
- 18.Dupont J, Tesseraud S, Derouet M, Collin A, Rideau N, Crochet S, Godet E, Cailleau-Audouin E, Métayer-Coustard S, Duclos MJ, et al: Insulin immuno-neutralization in chicken: effects on insulin signaling and gene expression in liver and muscle. J Endocrinol. 2008, 197 (3): 531-542. 10.1677/JOE-08-0055.CrossRefPubMedGoogle Scholar
- 21.Roberts LD, Murray AJ, Menassa D, Ashmore T, Nicholls AW, Griffin JL: The contrasting roles of PPARdelta and PPARgamma in regulating the metabolic switch between oxidation and storage of fats in white adipose tissue. Genome Biol. 2011, 12 (8): R75-10.1186/gb-2011-12-8-r75.PubMedCentralCrossRefPubMedGoogle Scholar
- 22.Mandrup S, Færgeman NJ, Knudsen J: Structure, function and phylogeny of acyl-CoA binding protein. Cellular proteins and their fatty acids in health and disease. Edited by: Duttaroy AKaS F. 2004, Wiley-VCH gmbH & Co, Kaga, Weinheim, 151-172.Google Scholar
- 24.Kalupahana NS, Massiera F, Quignard-Boulange A, Ailhaud G, Voy BH, Wasserman DH, Moustaid-Moussa N: Overproduction of angiotensinogen from adipose tissue induces adipose inflammation, glucose intolerance, and insulin resistance. Obesity. 2012, 20 (1): 48-56. 10.1038/oby.2011.299.PubMedCentralCrossRefPubMedGoogle Scholar
- 25.Suyeon Kim MS-B, Annie Q-B, Florence M, Michele T, Gerard A, Jung Han K, Naima M-M, Voy Brynn H: The adipose renin-angiotensin system modulates systemic markers of insulin sensitivity and activates the intrarenal renin-angiotensin system. J Biomed Biotechnol. 2006, 2006 (27012): 2006-Google Scholar
- 26.Kim S, Urs S, Massiera F, Wortmann P, Joshi R, Heo YR, Andersen B, Kobayashi H, Teboul M, Ailhaud G, et al: Effects of high-Fat diet, angiotensinogen (agt) gene inactivation, and targeted expression to adipose tissue on lipid metabolism and renal gene expression. Horm Metab Res. 2002, 34 (11/12): 721,725-CrossRefGoogle Scholar
- 27.Elbein SC, Kern PA, Rasouli N, Yao-Borengasser A, Sharma NK, Das SK: Global gene expression profiles of subcutaneous adipose and muscle from glucose-tolerant, insulin-sensitive, and insulin-resistant individuals matched for BMI. Diabetes. 2011, 60 (3): 1019-1029. 10.2337/db10-1270.PubMedCentralCrossRefPubMedGoogle Scholar
- 32.Dupuis J, Langenberg C, Prokopenko I, Saxena R, Soranzo N, Jackson AU, Wheeler E, Glazer NL, Bouatia-Naji N, Gloyn AL, et al: New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet. 2010, 42 (2): 105-116. 10.1038/ng.520.PubMedCentralCrossRefPubMedGoogle Scholar
- 35.Goto T, Lee J-Y, Teraminami A, Kim Y-I, Hirai S, Uemura T, Inoue H, Takahashi N, Kawada T: Activation of peroxisome proliferator-activated receptor-alpha stimulates both differentiation and fatty acid oxidation in adipocytes. J Lipid Res. 2011, 52 (5): 873-884. 10.1194/jlr.M011320.PubMedCentralCrossRefPubMedGoogle Scholar
- 39.Tsuchida A, Yamauchi T, Takekawa S, Hada Y, Ito Y, Maki T, Kadowaki T: Peroxisome proliferator–activated receptor (PPAR)α activation increases adiponectin receptors and reduces obesity-related inflammation in adipose tissue. Diabetes. 2005, 54 (12): 3358-3370. 10.2337/diabetes.54.12.3358.CrossRefPubMedGoogle Scholar
- 41.Krey G, Braissant O, L’Horset F, Kalkhoven E, Perroud M, Parker MG, Wahli W: Fatty acids, eicosanoids, and hypolipidemic agents identified as ligands of peroxisome proliferator-activated receptors by coactivator-dependent receptor ligand assay. Mol Endocrinol. 1997, 11 (6): 779-791. 10.1210/me.11.6.779.CrossRefPubMedGoogle Scholar
- 50.Ruderman NB, Saha AK, Vavvas D, Witters LA: Malonyl-CoA, fuel sensing, and insulin resistance. Am J Physiol - Endocrinol Metab. 1999, 276 (1): E1-E18.Google Scholar
- 61.Benjamini Y, Hochberg Y: Controlling the false discovery rate: a practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B. 1995, 57 (1): 289-300.Google Scholar
- 63.Bindea G, Mlecnik B, Hackl H, Charoentong P, Tosolini M, Kirilovsky A, Fridman WH, Pages F, Trajanoski Z, Galon J: ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics. 2009, 25 (8): 1091-1093. 10.1093/bioinformatics/btp101.PubMedCentralCrossRefPubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.