Biochemical Journal

Review article

Molecular mechanisms of metabolic regulation by insulin in Drosophila

Aurelio A. Teleman

Abstract

The insulin signalling pathway is highly conserved from mammals to Drosophila. Insulin signalling in the fly, as in mammals, regulates a number of physiological functions, including carbohydrate and lipid metabolism, tissue growth and longevity. In the present review, I discuss the molecular mechanisms by which insulin signalling regulates metabolism in Drosophila, comparing and contrasting with the mammalian system. I discuss both the intracellular signalling network, as well as the communication between organs in the fly.

  • Akt
  • Drosophila metabolism
  • insulin signalling
  • target of rapamycin (TOR)
  • total body glucose
  • total body lipid

INTRODUCTION

In the past decade, Drosophila has been one of the important model systems for studying the insulin and IGF (insulin-like growth factor)/TOR (target of rapamycin) signalling pathway (Figure 1). For instance, some components of the pathway such as Rheb, Tsc1 (tuberous sclerosis complex 1) and Tsc2 (tuberous sclerosis complex 2) were first placed into the TOR pathway in the fly [15]. Such work has been supported by the powerful genetic tools available in Drosophila, allowing researchers to quickly generate loss-of-function animals to query gene function, and epistasis experiments to query gene relationships, both of which are fundamental for elucidating signalling networks.

Figure 1 The ‘core’ intracellular insulin signalling pathway in Drosophila

Functional relationships between components are indicated. Arrows indicate activation, but not necessarily indicate direct physical interactions, whereas bar-ended lines indicate inhibitory interactions. Broken lines indicate indirect interactions or interactions requiring further study. Red arrows indicate transcriptional regulation. Details of each interaction are described in the main text, and outlined in Table 1.

More recently, Drosophila has also become a model system for studying how insulin signalling regulates organismal and cellular metabolism (Table 1). Although the metabolic regulation of Drosophila is interesting in itself, work in this field has also been sparked by an increased interest in organismal metabolism due to the emerging world-wide epidemic of Type 2 diabetes and obesity. At first, one might be sceptical that the metabolic regulation of flies and humans have anything in common, yet a surprising amount of conservation and parallelism between the two systems has emerged. As discussed below, the conservation occurs not only at the level of the molecular components of the insulin signalling pathway, but also at the level of the physiological outputs of the pathway. This is probably because multicellular organisms needed to regulate their response to varying nutrient conditions early in evolutionary history. Indeed Drosophila, like mammals, regulate their circulating sugar levels and store excess energy in the form of glycogen and lipids [6]. These energy stores are then mobilized when energy is needed [7]. In the light of this, in this review I will highlight both the similarities and the differences between the fly and mammalian systems.

View this table:
Table 1 Summary of insulin genes with demonstrated metabolic phenotypes in Drosophila

The metabolic phenotypes are given for the loss-of-funtion condition unless otherwise stated.

INSULIN MEDIATES MUCH OF THE NUTRIENT-DEPENDENT SIGNALLING IN DROSOPHILA

Two sets of observations support the conclusion that most of the nutrient sensing in Drosophila takes place via the insulin pathway. First, mutations in components of the insulin pathway phenocopy the effects of nutrient deprivation. For instance, mutation of the kinase TOR phenocopies amino acid deprivation, leading to a block in tissue growth, reduced nucleolar size and cell cycle arrest despite the presence of nutrients [8]. As another example, reduction in S6K [RPS6 (ribosomal protein S6)-p70-protein kinase] activity in the Drosophila brain leads to similar changes in feeding behaviour to those observed upon fasting, including an increased rate of food intake and reduced aversion to unappealing food [9]. Secondly, artificially elevating signalling through the insulin pathway is sufficient to bypass the need for nutrients for many cell-biological processes. This means that signalling through the pathway is ‘epistatic’ to nutrient availability. For instance, activation of PI3K (phosphoinositide 3-kinase) signalling bypasses the usual nutrient requirement of many larval cell types for growth and DNA replication [10]. Thus PI3K activation can lead to cell growth and proliferation despite a lack of nutrient availability, causing death of the animal [10].

It is worth mentioning that in Drosophila, organismal growth is restricted to the larval stages of development, which can therefore be considered similar to the childhood and teenage years in humans. In contrast, after metamorphosis, the adult fly no longer increases in size. This means that genetic manipulations that alter insulin signalling in the fly during larval stages result in both growth and metabolic phenotypes, whereas genetic manipulations that alter insulin signalling in the adult only result in metabolic consequences.

Metabolic regulation is inherently a complex system, designed to maintain homoeostasis in a robust way, yet to be responsive to varying inputs. As a consequence, it is rich in feedback loops and interconnections, both within a single cell and between organs. In the following sections I will first describe the intracellular insulin/IGF signalling network in flies, together with the metabolic phenotypes that have been reported for components of the pathway. I will then address the roles and interconnections of various organs involved in insulin signalling in the fly. Finally, I will point out some differences between the Drosophila and mammalian insulin systems.

COMPONENTS OF THE INSULIN PATHWAY AND METABOLIC PHENOTYPES

ILPs (Insulin-like peptides)

Signalling through the insulin/IGF pathway commences upon binding of ligand to the insulin receptor, in Drosophila InR (insulin-like receptor) (Figure 1). Drosophila has seven ILPs, termed ILP1–7, which are homologues of human insulin and IGFs. Mammalian insulin and Drosophila ILPs are homologues at the protein level [11], as well as being functionally equivalent; mammalian insulin can activate the Drosophila InR [12,13] and Drosophila extract has insulin bioactivity in mice [14]. Although it was initially surprising that Drosophila has seven ILPs, as the complexity of Drosophila insulin signalling tends to be lower than that of its mammalian counterpart, it is now apparent that mammals also have a large number of ILPs as at least nine have been identified in rodents {insulin, Igf1, Igf2, relaxin, Insl3, Ins4, Ins5 and Ins6 (for a review see [15])}. The seven Drosophila ILPs are similar at the amino acid level with the exception of ILP6, which has some structural differences that might distinguish it functionally from the others [16]. However, each ILP has a different expression pattern and unique regulation, suggesting that the functions of the various ILPs are not overlapping. ILPs 1, 2, 3 and 5 are expressed in seven median NSCs (neurosecretory cells) of the Drosophila brain [11,17,18], ILPs 4, 5 and 6 are expressed in the midgut, ILP2 is expressed in the imaginal discs and the salivary gland [11,17] and ILP7 is expressed in the ventral nerve cord of the brain [11]. In the larva, expression of ILPs 2, 3 and 5 reduce upon fasting [17,19] in a manner analogous to that of human insulin, whereas, surprisingly, expression of ILPs 6 and 7 increase [19].

Overexpression of any of the seven ILPs during larval development leads to increased body size [17] indicating that all seven ILPs can activate the insulin receptor. Larval ablation of the NSCs that produce ILPs 2, 3 and 5 yields adults that are developmentally delayed, reduced in body size and have elevated circulating sugars, total body glycogen and lipid levels [18,20,21]. Ablation of the same cells in adults leads to increased trehalosaemia without the growth phenotypes [21]. Selective knockdown of ILP2 expression in the NSCs by RNAi (RNA interference) yields adults with elevated total body trehalose levels, but none of the other defects, possibly due to compensatory changes in ILP 3 and 5 expression [22].

These physiological effects of ablating the NSCs cells, reducing insulin levels systemically and hence causing elevated circulating sugars and lipid accumulation, are analogous to the effects seen in diabetic patients or in mice when there is generalized insulin resistance, such as in insulin receptor knockouts (for a review see [15]). Furthermore, the attenuated growth observed in animals with reduced insulin signalling (either via ablation of the NSCs or other manipulations as described below) parallel the reduction in body size seen in mice mutant for insulin and IGF receptors [23].

Although it is probable that the various ILPs have distinct functions, this has not yet been carefully studied and the role of each individual ILP in fly physiology is only starting to be elucidated. For example, overexpression of ILPs 2 and 4 in adult NSCs induce behavioural changes that are not induced by ILP3. When larvae are starved of nutrients, they undertake a set of hunger-induced behaviours, including an enhanced feeding rate. This is caused by reduced insulin production from the NSCs. Expression of ILPs 2 and 4 are able to suppress this hunger-driven feeding activity, whereas expression of ILP3 does not [9]. The above example also illustrates that in Drosophila, ILPs form part of a homoeostatic feedback loop. Nutrient conditions regulate expression of several ILPs, and expression of ILPs in turn regulate feeding behaviour. The generation of knockout flies for each of the individual ILPs should shed light on the function of each ILP.

Three secreted inhibitors of ILP function have recently been described, Imp-L2 (imaginal morphogenesis protein-late 2, also known as Ecdysone-inducible gene L2) [24], dALS (Drosophila acid-labile subunit, also known as convoluted) [25] and NLaz (neural lazarillo) [26]. These will be described in detail in the section on insulin signalling in tissues.

Insulin receptor and IRS (insulin receptor substrate)

Drosophila has one insulin receptor, InR, which is similar in sequence to the mammalian insulin receptor, except that it contains 400 additional amino acids at the C-terminus. This extension contains three YXXM motifs similar to those found in mammalian IRS1, allowing Drosophila InR to bind PI3K in the absence of an IRS [27,28]. Whereas complete removal of InR in Drosophila throughout development causes early larval lethality [12,29], a more mild reduction in InR levels, specifically in the adult, leads to live adults with increased total body free sugar and lipid levels [30].

Upon ligand binding, InR autophosphorylates [28], recruiting the IRSs Chico and Lnk [31,32]. Although a large number of adapter proteins have been described as binding to the mammalian insulin/IGF receptors {IRS1, IRS2, IRS3, IRS4, SHC, CBL, APS, SH2B, GAB1, GAB2, DOCK1 (dedicator of cytokinesis or downstream of Crk-180 homologue 1), DOCK2 and CEACAM1 (carcinoembryonic antigen-related cell-adhesion molecule 1); for a review see [33]}, only Chico (the fly homologue of IRS1), dreadlocks (which is involved in axon guidance [34]) and Lnk [the fly homologue of SH2B (SH2B adaptor protein 1)] have been described as binding to the Drosophila InR. As Drosophila has homologues for some of these other adapter proteins, future work might shed light on whether they are also involved in transducing signals downstream of the fly insulin receptor. Both Chico and Lnk mutant adults are strongly reduced in size and viability, and have elevated total body lipid levels [31,32], whereas simultaneously mutation of Chico and Lnk is lethal, suggesting these two proteins transduce the majority of the signal downstream of InR. The metabolic phenotypes observed in InR, Chico and Lnk mutants, which are essentially insulin-resistant in the whole body, are consistent with the metabolic phenotypes observed upon ablation of the NSCs as described above, which also causes a reduction in insulin signalling.

In mammals, a number of connections have been described between insulin and Ras signalling. In mammals, activation of the insulin receptor induces interaction of the GRB2 (growth-factor-receptor-bound protein 2)–SOS (Son of sevenless) complex with IRS1 thus activating Ras and subsequently ERK (extracellular-signal-regulated kinase) [35,36]. Furthermore, Ras-GTP binds to and activates the catalytic subunit of PI3K [37]. These interconnections are starting to be studied in the fly, yielding insights into the physiological relevance of these interactions. Although insulin treatment of Drosophila cells leads to activation of ERK [38], mutating the consensus binding site for the Ras pathway adaptor Drk (downstream of receptor kinase)/GRB2 in the Chico protein does not interfere with growth [39]. As Ras is required for viability, this suggests that either Ras is activated in a Chico-independent manner in Drosophila, perhaps via the C-terminal extension of the Drosophila insulin receptor, or that the effect of insulin signalling on Ras is modulatory and not required for insulin-mediated growth. A study has shown that mutating the Ras-binding domain in the PI3K Drosophila p110 protein (also known as PI3K92E), the catalytic subunit of PI3K, in vivo is dispensable for viability but is required for maximal PI3K signalling, leading to a phenotype of small flies with dramatically lowered egg production [40]. This lowered egg production might reflect a defect in lipid metabolism given lipids are required in large amounts in eggs. As PI3K is required for viability, these results suggest the impact of Ras on insulin signalling is modulatory in function.

PI3K, PTEN (phosphatase and tensin homologue deleted on chromosome 10), PDK1 (phosphoinositide-dependent kinase 1) and Akt/PKB (protein kinase B)

Upon phosphorylation of InR and Chico, PI3K (p110) [41,42], together with its adapter subunit p60 (also known as PI3K21B) [43], is recruited to the cell membrane and activated. This leads to the generation and accumulation of PtdIns(3,4,5)P3 at the cell membrane. The kinase activity of PI3K is opposed by the phosphatase activity of the tumour suppressor PTEN. Step (Steppke), a member of the cytohesin GEF (guanine-nucleotide-exchange factor) family, has now been found to act upstream of PI3K to regulate growth and metabolism [44,45]. Step mutant flies are small, and blocking the function of the Step homologues in mice causes elevated gene transcription of insulin-repressed gluconeogenic genes, as well as inhibition of glycogen and fatty acid synthesis [44,45]. The exact mechanism by which Step affects PI3K function remains to be elucidated.

Accumulation of PtdIns(3,4,5)P3 recruits the two kinases PDK1 [also known as PK61C (protein kinase 61C)] and Akt to the plasma membrane, via their lipid-binding PH (pleckstrin homology) domains, and leads to their phosphorylation and activation (for a review see [46]). These proteins and interactions are highly conserved between flies and mammals. The most dramatic phenotypes observed in mutants for these components are growth-related. Overexpression of PI3K, Akt or PDK1 leads to tissue overgrowth, whereas loss-of-function leads to tissue undergrowth and lethality [41,42,47,48]. Because of this, it is difficult to reveal any ‘metabolic’ phenotypes. In contrast, mutation of the counteracting phosphatase PTEN has the opposite size effects and inactivating mutations in PTEN are not lethal [4952]. In PTEN mutants, which can be considered insulin ‘hyperactive’ in the entire body, it would be expected that we would see the opposite phenotype from that in diabetic patients. Indeed both total body glycogen and lipid levels are reduced in PTEN mutant adult flies [39] (although, as will be discussed below, in one specific cell type in Drosophila, the nurse cells, PTEN mutation leads to the opposite effect, namely formation of enlarged lipid droplets [53]).

Akt activity is regulated by two additional inputs. First, recent work has shown that Akt is dephosphorylated and inactivated by the phosphatase PP2A (protein phosphatase 2A) via its PP2A-B′ subunit, Wdb (widerborst) [54]. Mutation of Wdb leads to activation of Akt and to the same lipid-droplet phenotype in nurse cells as seen in PTEN mutants [53,54]. It would be interesting to know whether, at the whole organism level, Wdb mutants have the same phenotypes of reduced body glycogen and lipid levels as are observed in PTEN mutants. Secondly, Akt is phosphorylated and activated by TOR-C2 (TOR complex 2) both in Drosophila and in mammals [55]. This regulation appears to be modulatory, as removal of TOR-C2 activity by mutation of the essential component Rictor (rapamycin-insensitive companion of Tor) only resulted in mild growth impairment and no observable metabolic effects ([55]; V. Hietakangas, personal communication).

Akt targets and FOXO (forkhead box O) proteins

In mammals, Akt phosphorylates a large number of proteins involved in metabolic control, including GSK-3β [glycogen synthase kinase 3β; the Drosophila orthologue is known as Sgg (shaggy)], TBC1D4 (TBC1 domain family, member 4), the FOXO transcription factors, Tsc2, PRAS40 (proline-rich Akt substrate of 40 kDa; the Drosophila orthologue is known as Lobe), 6-phosphofructo-2-kinase and ATP-citrate lyase [46,56]. A number of these interactions have also been studied in Drosophila.

One important target of Akt in Drosophila is the single homologue of the FOXO transcription factors, dFOXO [57,58]. In Drosophila, as in mammals, phosphorylation of FOXO by Akt leads to its retention in the cytoplasm, inhibiting its nuclear transcriptional activity [57,59]. As one of the principle transcription factors in the insulin signalling pathway, FOXO has a profound impact on animal metabolism and a large number of functional studies have been performed in the fly [5766]. These studies all converge at the concept that part of the anabolic effect of insulin results from blocking FOXO activity, which otherwise promotes the conservation of energy or, in extreme cases, even catabolism. Overexpression of dFOXO in larvae phenocopies starvation, leading to growth arrest and causing larvae to wander away from food [58]. In contrast targeted overexpression of dFOXO in fly tissues reduces their size by reducing cell number [57]. Analogously, growth suppression can be seen when endogenous FOXO is activated by reducing Akt activation [60,61]. Along these lines, one study showed that flies infected with Mycobacterium marinum undergo a process similar to wasting; they progressively lose metabolic stores, in the form of fat and glycogen, and become hyperglycaemic. This process is mediated in part via dFOXO, as dFOXO mutants were found to exhibit less wasting [62]. Finally, a recent report showed that dFOXO is activated upon amino acid starvation of the animal and that this activation is required for the animals to survive the adverse conditions [63]. Surprisingly, dFOXO functions both autonomously and non-autonomously in cells; expression of dFOXO in adult head fat body causes reduces ILP2 secretion from the NSCs via an unknown mechanism [64] and causes the same metabolic consequences as in flies with reduced insulin production, including increased longevity [65]. This will be discussed in more detail in the section on tissue specificity below. A further Drosophila scaffolding protein, called Melt (melted), has been found to modulate the ability of Akt to inhibit FOXO in vivo, possibly by encouraging the physical interaction between Akt and FOXO [66]. Melt mutants display elevated FOXO activity and reduced total body lipid levels, a phenotype which could be rescued by removal of FOXO [66].

FOXO exerts its effects by modulating transcription of a very large number of target genes. One canonical target of dFOXO is the translational repressor 4E-BP (4E-binding protein, in Drosophila also known as Thor). Upon activation, 4E-BP binds eIF4E (eukaryotic initiation factor 4E) and blocks recruitment of the ribosome to the 5′ end of mRNAs. As a result, cellular translation rates are dampened [67]. Although this has been suggested to regulate tissue growth, no growth defects are observed in 4E-BP-null flies [68], nor in 4E-BP1 or 4E-BP2-null mice [69,70]. Instead, both 4E-BP-null flies and 4E-BP1-null mice have elevated metabolic rates [68,69], which is consistent with the idea that protein synthesis is an energetically expensive process and that its regulation impacts overall organismal energy balance. Studies have shown that FOXO regulates close to 2000 genes, half of which are regulated in a tissue-specific manner [19,71]. Among these are many components of the translational apparatus, as well as mitochondrial components, leading to an overall reduction in translation and in mitochondrial biogenesis upon dFOXO activation. Furthermore, dFOXO was found to regulate expression of Myc (the Drosophila orthologue is also known as diminutive), which in turn will also affect mitochondrial biogenesis [19,72]. Finally, dFOXO also regulates expression of an acid lipase (dLip4; Drosophila lipase 4) [73], and so is involved in lipid homoeostasis. In summary, FOXO regulates expression of a large number of genes that have an impact upon cellular metabolism. Furthermore, a large number of metabolism-related targets for FOXO1 have been described in mammals (for a review see [74]), which have not yet been studied in depth in the fly.

A second important phosphorylation target of Akt is the tumour suppressor Tsc2. This phosphorylation presents a conundrum. Tsc2 inhibits TOR-C1 (TOR complex 1) activity both in flies and in mammals via the small GTPase Rheb. Tsc2 acts as a GAP (GTPase-activating protein) for Rheb, thereby inhibiting it and, in turn, inhibiting TOR (Figure 1). In both systems, inhibition of Tsc2 therefore leads to TOR hyperactivation and consequently strong effects on tissue growth and metabolism [5,7577]. Phosphorylation of Tsc2 by Akt can inhibit Tsc2 function, at least when Tsc2 is overexpressed in mammalian cell culture, in Drosophila cell culture or in vivo in the fly [7881]. However, the in vivo functional significance of this phosphorylation event is not clear, as flies lacking Akt phosphorylation sites on Tsc2, or flies simultaneously lacking Akt phosphorylation sites on Tsc1 and Tsc2, are viable and normal in size, with only very mild metabolic defects [82,83]. Therefore, although this phosphorylation does occur in vivo, the functional consequences of this regulation are not yet clear. Further work, perhaps with ‘knockin’ mice, should shed light on whether this reflects a difference between mammals and flies, or whether phenotypes are being masked by redundant regulatory mechanisms such as the possible mechanism mediated via PRAS40/Lobe. In an analogous manner to Tsc2, mammalian PRAS40 was identified as a TOR-C1 inhibitor that is phosphorylated and inhibited by Akt [84,85]. However, the functional role of PRAS40 is not completely clear, as it has also been reported that PRAS40 is actually a TOR-C1 target, required for signalling downstream of the complex [8688]. PRAS40 and Tsc2 therefore could represent two redundant pathways by which Akt activates TOR. Knockdown of the Drosophila orthologue Lobe in tissue culture does activate phosphorylation of TOR-C1 targets, suggesting that Lobe/PRAS40 function might be conserved in flies [85]. However, Lobe mutants display phenotypes suggestive of Notch pathways function, rather than TOR-C1 function [89], so further work will be required to shed light on the possible role of Lobe in TOR-C1 signalling.

In mammals, Akt phosphorylates and inactivates GSK-3β, leading to the dephosphorylation and activation of glycogen synthase and hence to an acceleration of glycogen synthesis [90]. The Drosophila orthologue of GSK-3β, Sgg, was also shown to be phosphorylated downstream of PI3K signalling in vivo [91]. Future work will shed light on whether this regulation has metabolic consequences.

Another phosphorylation target of mammalian Akt is the kinase SIK2 (salt-inducible kinase 2; the Drosophila orthologue is CG4290) [92]. Upon feeding, Akt phosphorylates SIK2, leading to phosphorylation and degradation of the transcriptional co-activator TORC (in mammals also known as CRTC2 [CREB (cAMP-response-element-binding protein)-regulated transcription co-activator 2]. As a consequence, TORC no longer functions as a CREB co-activator and inducer of genes involved in catabolism and gluconeogenesis. Recently, this signalling cascade was found to be conserved in Drosophila. The Drosophila TORC homologue is phosphorylated by Drosophila SIK2 in an Akt-dependent manner [93]. TORC mutant adult flies are sensitive to starvation and oxidative stress [93]. Surprisingly, TORC mutant flies have reduced glycogen and lipid stores [93], rather than the elevated levels of nutrient stores that might be expected from mutation of a catabolic gene.

The Tsc1/2, Rheb and TOR signalling cassette

A central regulator of cellular metabolism is the anabolic kinase TOR. TOR exists in one of two complexes, termed TOR-C1 and TOR-C2, with some components that are shared and some that are distinct between the two complexes (for a review see [94]). TOR-C1 is traditionally thought of as a regulator of cell size and cell growth, as manipulation of TOR-C1 activity affects cell size in eukaryotes from yeast to humans. However, TOR also regulates processes that at first seem quite disparate, including carbohydrate metabolism, lipid metabolism and autophagy. One way to unify these functions is to consider TOR-C1 principally as a regulator of cellular metabolism, controlling a cell's decision on whether to use energy and nutrients or whether to conserve them. Growth can then be considered a metabolic process. To grow, a cell needs biochemical building blocks, including amino acids, proteins, lipids and carbohydrates, which in turn need to by synthesized via metabolic pathways. Furthermore, protein synthesis is a highly energetically expensive process, using 35–80% of a cell's energy [95,96]. Therefore cell size and growth are readouts or consequences of a cell-metabolic decision. When TOR-C1 activity is high, this leads to accumulation of glycogen, lipids and cell mass, whereas low TOR-C1 activity in a cell leads to the catabolism of carbohydrate and lipid stores, turnover of excess mass and, in extreme cases, autophagy. For this reason, I also mention growth phenotypes in the present review, as a readout of cellular metabolism.

In Drosophila, TOR-C1 function is conserved relative to mammals. Genetic manipulation of TOR activity results in significant changes in tissue and animal size. Loss-of-function reduces tissue size, by reducing cell size and cell number, whereas gain-of-function leads to the opposite effects [8,97]. Loss of TOR also leads to lipid vesicle aggregation in the larval fat body [8] and mild impairment of TOR-C1 activity in the whole larva also results in decreased lipid stores in adipose tissue and reduced circulating sugar levels [98]. This is consistent with the phenotype of mice lacking one of the TOR-C1 targets, S6K1, which are hypersensitive to insulin and are protected against obesity [99]. Strong reduction in TOR-C1 activity in the Drosophila adipose tissue induces autophagy [100] (for a review see [101]) and TOR-C1 activation is sufficient to suppress starvation-induced autophagy [102].

Mechanistically, one of the main cellular processes that TOR-C1 regulates is translation. This occurs at a number of levels. TOR-C1 regulates ribosome biogenesis by controlling the production of both components of the ribosome, ribosomal proteins and rRNA. In mammals and in the fly, TOR-C1 activity promotes the function of the transcription factor TIF-IA (transcriptional intermediary factor-IA), which is required for RNA polymerase I-mediated expression of rRNA [103,104]. Although the regulation is conserved, the mechanistic details might be different in the two systems, as phosphorylation of TIF-IA by mTOR influences TIF-IA subcellular localization in mammalian cells [104], whereas this does not seem to be the case in Drosophila (A. A. Teleman, unpublished work). Furthermore, in mammals, TOR-C1 also regulates activity of another rRNA transcription factor, UBF (upstream binding transcription factor) [105]; however, Drosophila contains no obvious homologue. In Drosophila, TOR-C1 also regulates transcription of genes involved in ribosome synthesis and assembly [106]. This was recently shown to occur via regulation of the transcription factor Myc [19]. Inhibition of TOR-C1 leads to a rapid reduction in Myc protein levels, and consequently in expression of genes involved in ribosome biogenesis. Consistent with this result, Myc activity is also required for TOR-C1 to drive tissue growth [19,72]. Close connections between TOR-C1 and Myc appear to also exist in mammals [107]. In addition to regulating ribosome biogenesis, in mammals TOR-C1 also regulates expression of tRNAs by RNA polymerase III. Mammalian TOR-C1 phosphorylates the transcription factor Maf1, causing it to relocalize to the nucleolus where it activates expression of tRNA genes [108]. Whether this is also the case in Drosophila has not yet been investigated. Finally, TOR-C1 regulates translation initiation and elongation. Both in mammals and in flies TOR-C1 phosphorylates the inhibitor of translation initiation 4E-BP [109] (for reviews see [67,94,110]). This leads to the dissociation of 4E-BP from the initiation factor eIF4E, allowing recruitment of the ribosome to the cap complex at the 5′ end of cellular mRNAs, and an increase in overall cellular translation rates. TOR-C1 also phosphorylates and activates S6K, which phosphorylates the 40S ribosomal protein S6 [109] (for reviews see [67,94,110]). S6K requires phosphorylation at two sites for full activation and the second site is phosphorylated by PDK1. Activation of S6K was initially thought to regulate translation of mRNAs containing 5′ TOPs (terminal oligopyrimidine tracts), which include many mRNAs encoding components of the translation apparatus, although this has now been shown not to be the case, as mice lacking S6K have normal translation of 5′ TOP mRNAs [111]. Therefore it is likely that TOR-C1 regulates translation of 5′ TOP mRNAs in an S6K-independent manner, which remains to be explored. Nonetheless, mammalian S6K phosphorylates two other regulators of translation initiation and elongation, eIF4B (eukaryotic initiation factor 4B) and eEF2K (eukaryotic elongation factor-2 kinase) [112,113].

Regulation of translation by these TOR-C1 effectors has a strong impact on organismal size and metabolism. TIF-IA mutants arrest as first instar larvae and fail to grow [103]. Drosophila deficient in S6K exhibit a strong delay in development and a severe reduction in body size [114]. Although the S6K mutant phenotype is less severe than that of dTOR mutants, S6K appears to be a key mediator of TOR-C1 activity, as overexpression of S6K in vivo can rescue dTOR the viability of mutant animals [8] and removal of one copy of S6K can rescue the lethality of Tsc1/2 loss-of-function [115]. As mentioned above, mutants for the other TOR target, 4E-BP, do not display growth abnormalities, but rather have elevated metabolic rates, leading to reduced organismal lipid levels, consistent with translation being an energetically expensive process [68].

In addition to translation, TOR-C1 regulates a number of other cellular processes. TOR-C1 inhibits autophagy in the Drosophila fat body, and perhaps in other tissues as well [100,102]. Although the molecular mechanism by which this occurs is not completely clear, it is known to be S6K-independent [102] and that Atg1 (autophagy-specific gene 1), an important regulator of autophagy, is involved. TOR-C1 physically interacts with Atg1 and promotes its phosphorylation, although more work will be required to understand the implications of this regulation [116]. There are probably other molecular mechanisms by which TOR-C1 regulates autophagy and these remain to be explored. In addition, Atg1 was shown to inhibit S6K phosphorylation [117] suggesting a regulatory feedback loop between TOR activity and autophagy. Autophagy is important for lipid mobilization as inhibition of autophagy leads to increased lipid storage, indicating this is probably one means by which TOR-C1 regulates metabolism [118].

TOR signalling also stimulates bulk endocytic uptake and targeted uptake of nutrients, although again the mechanism is not yet completely elucidated [119]. This may occur in part via inhibition of the endocytic degradation of nutrient transporters such as Slif (slimfast) [119]. Additionally, calderon [also known as Orct2 (organic cation transporter 2)], an organic cation transporter similar to transporters believed to function in the uptake of sugars and other organic metabolites, was shown to be transcriptionally induced by S6K activity [120]. Calderon is required cell autonomously and downstream of S6K for normal growth and proliferation of larval tissues [120].

Finally, a recent study in mammalian cells showed that TOR-C1 activity is required for Akt-induced activation of de novo lipid synthesis [121]. Inhibition of mTOR-C1 by rapamycin prevented Akt-dependent accumulation of unsaturated and saturated fatty acids as well as phosphatidylcholine and phosphatidylglycerol [121]. This probably occurs via regulation of SREBP [sterol-regulatory-element-binding protein, also known in Drosophlia as HLH106 (helix-loop-helix 106)] by TOR-C1. SREBP is a transcription factor that regulates expression of genes involved in sterol biosynthesis. The nuclear accumulation of mouse SREBP1 in response to Akt activation was prevented by rapamycin [121]. This interaction is probably conserved in flies as the study showed that silencing of Drosophila SREBP blocked the induction of cell growth caused by PI3K [121].

It is clear that the regulation of TOR-C1 is complex and not completely understood (for reviews see [67,110,122,123]). Briefly, TOR-C1 is activated in response to growth factors, especially insulin, and nutrient availability. Activation of TOR-C1 in response to insulin signalling occurs via Akt, but the exact mechanism is still unclear and, as described above, this activation might occur via two redundant pathways: by phosphorylation and inhibition of the Tsc1/2 complex or by phosphorylation and inhibition of PRAS40/Lobe. The Tsc1/2 complex acts as a GAP, and inhibitor, for the small GTPase Rheb, which in turn activates TOR. Rheb was reported to be activated by the GEF TCTP (translationally controlled tumour protein) [124]; however, a later study called this into question [125]. In mammalian cells TOR-C1 is also regulated by cellular energy status via AMPK (AMP-activated protein kinase). High intracellular AMP levels activate AMPK, which directly phosphorylates Tsc2, leading to Tsc2 activation and TOR-C1 inhibition. Whether activation of AMPK also leads to TOR-C1 inhibition in the fly remains to be directly tested. TOR-C1 is also regulated by the availability of amino acids in mammals and in Drosophila. Three factors have been reported to mediate the amino acid availability signal to TOR-C1, Vps34 (also known in Drosophila as PI3K59F), the Rag GTPases and MAP4K3 [MAPK (mitogen-activated protein kinase) kinase kinase kinase-3; the Drosophila orthologue is CG7097]. In mammals, activity of the class III PI3K Vps34 is regulated by amino acid availability via a calcium-dependent mechanism [126]. This leads to accumulation of PtdIns3P in cells, which is thought to cause the recruitment of proteins recognizing PtdIns3P to early endosomes, forming an intracellular signalling platform that leads to TOR-C1 activation [127,128]. This feature of the pathway may be specific for vertebrates, as flies mutant for Vps34 have been reported not to show TOR-C1 signalling defects [129]. Recently, two groups discovered that Rag GTPases mediate amino acid signalling to TOR-C1, both in mammals and in flies [130,131]. The emerging picture is that amino acids change the GDP/GTP loading of the Rag GTPases, thereby stimulating the binding of Rag heterodimeric complexes to TOR-C1. This in turn causes TOR-C1 to change its intracellular localization, perhaps relocalizing it to vesicles containing the activator Rheb. Finally, Lamb and co-workers identified human MAP4K3 in cell culture as a kinase that is activated by amino acid sufficiency, and in turn it is necessary for S6K activation in response to amino acids [132]. This mechanism also appears to be conserved from flies to humans; flies mutant for MAP4K3 have reduced TOR-C1 activity levels and display TOR-C1 hypomophorphic phenotypes (B. Bryk, K. Hahn, S. M. Cohen and A. A. Teleman, unpublished work). The relationships between Vps34, the Rag GTPases and MAP4K3 remain to be investigated. In addition, novel regulators of TOR-C1 have been identified, whose mechanisms of action are not yet clear. TOR-C1 activity was found to be down-regulated upon exposure of cells to hypoxic conditions [133]. This down-regulation was mediated by Ptp61F (protein tyrosine phosphatase 61F), although the mechanism remains to be elucidated [133]. TOR-C1 activity is also modulated by the amino acid transporter Path (pathetic). Path mutants are reduced in size and yield genetic interactions consistent with reduced TOR-C1 activity [134]. When expressed in Xenopus oocytes, Path allows extracellular amino acids to activate TOR-C1. Surprisingly, however, Path has very low transport capacity for amino acids, suggesting Path may be acting as an amino acid sensor which regulates TOR-C1 via an unknown intracellular mechanism [134].

Outputs of the insulin signalling pathway

As described above, the ‘upstream’ signalling events in the insulin signalling pathway are now fairly well understood. In brief, they lead to activation of a number of kinases such as Akt and TOR via a relay of phosphorylation events. In comparison, however, at least in Drosophila, the outputs of the pathway are not as clearly defined. Activation of the kinases in the insulin cascade, in one way or another, has an impact on cellular metabolism by regulating the enzymes that act as effectors, controlling lipid and carbohydrate biosynthesis and breakdown. One well-understood example is the regulation of GSK-3β by Akt. Activation of Akt leads to phosphorylation and inactivation of GSK-3β, leading to the dephosphorylation and activation of glycogen synthase [90]. However, there are probably a large number of links between the insulin pathway and metabolic enzymes that are not yet known. Work in the future on this ‘downstream’ part of the insulin signalling pathway will be fundamental for us to understand how metabolic pathways are regulated by insulin. This will require the coming together of multiple fields of biological research, including cell signalling, biochemistry and metabolomics. Furthermore, such research will probably require the study of phenotypes that are more subtle and quantitative in nature, compared with the phenotypes the field has studied so far. This is because the few ‘master regulator’ kinases of the insulin pathway, such as Akt and TOR, control both directly and indirectly a large number of effector enzymes. Therefore mutation of these master regulatory kinases leads to obvious and easily studied phenotypes. In contrast, single mutations of each effector enzymes will only lead to a subset of effects. Nonetheless, an understanding of the molecular connections between insulin signalling components and cellular metabolic enzymes will be necessary for us to have a complete picture of how insulin regulates metabolism.

Intracellular feedback loops

The molecular mechanisms regulating animal metabolism need to simultaneously meet two constraints. First, the system needs to be exquisitely sensitive and reactive to nutrient conditions of the animal, in order to respond quickly to environmental changes. This is best illustrated by insulin levels and insulin intracellular signalling in humans; the reaction to changing blood glucose levels occurs within minutes to maintain the glucose concentration in human blood between 80 and 110 mg/dl [135]. Secondly, the system needs to be robust and to function properly in a wide range of individuals with differing genetic makeups. This ‘engineering’ problem of making a machine that is both flexible and inflexible is probably partly solved by utilizing the large number of feedback loops present in the system. In the case of insulin signalling, there are at least three feedback loops worth mentioning.

First, activation of the insulin signalling pathway leads to activation of PDK1 and TOR-C1, both of which phosphorylate and activate S6K. S6K, in mammalian cells, then phosphorylates IRS1, inhibiting its function and down-regulating signalling through the pathway [110]. This mechanism is probably conserved in Drosophila, as removal of S6K in Drosophila cell culture leads to increased Akt phosphorylation [136]. Thus this is one negative feedback loop which attenuates signalling through the pathway. Furthermore, reducing S6K phosphorylation specifically should lead to increased signalling through Akt and TOR, and therefore increased phosphorylation of the remaining TOR targets.

A second feedback loop occurs when activation of the insulin signalling pathway leads to activation of Akt and inhibition of the FOXO transcription factor. As the insulin receptor is a transcriptional target of FOXO, signalling through the insulin receptor leads to its own transcriptional down-regulation, attenuating signalling. Conversely, low insulin signalling leads to increased transcription of insulin receptor levels, sensitizing the system to renewed insulin signalling. This occurs both in mammalian cells and in Drosophila [137].

One further mechanism is the antagonistic relationship between activation of TOR-C1 and activation of TOR-C2. This can be seen in Drosophila cells by monitoring phosphorylation of S6K at Thr398 (a readout for TOR-C1 activity) and phosphorylation of Akt at Ser505 (a readout for TOR-C2 activity) when Rheb, Tsc1 or Tsc2 are knocked-down [136]. For instance, knockdown of Rheb in Drosophila S2 cells causes decreased phosphorylation of S6K and increased phosphorylation of Akt, indicating that TOR-C1 activity is reduced, whereas TOR-C2 activity is increased [136]. In mammalian cells, this also appears to be the case, as mouse embryonic fibroblasts lacking Tsc1 or Tsc2 have elevated TOR-C1 activity and severely blunted TOR-C2 activity [138]. As a consequence, a negative feedback loop is created between Akt and TOR (Figure 1). Activation of Akt leads to activation of TOR-C1 and inactivation of TOR-C2. This causes a reduction in the ability of TOR-C2 to enhance Akt activation via phosphorylation of its hydrophobic motif (Ser505 in Drosophila) [55].

Although the metabolic and growth consequences of these three feedback loops have not been carefully explored, they constitute an integral part of the insulin signalling network and therefore warrant attention.

INSULIN SIGNALLING IN TISSUE CONTEXT

The insulin signalling pathway described above can be considered the core pathway, which functions in almost all cells and tissues of the animal. In addition to this core pathway, there are tissue-specific upstream regulators and downstream effectors of the pathway. Furthermore, the metabolic consequences of manipulating insulin signalling depend strongly on the identity of the tissue in which the manipulation takes place. Therefore it is essential to consider insulin signalling in a tissue-specific context. Work in the mouse has been at the forefront of our understanding of how various tissues contribute to metabolic regulation. Nonetheless, recent studies in the fly have also taken tissue specificity into account. Future work will probably take tissue specificity increasingly into account, as tissue-specific manipulations are relatively easy in Drosophila.

Tissue-specific effects of insulin signalling

A wealth of mouse knockout studies have shown that the physiological effects of reduced insulin signalling differ enormously depending on the tissue in which the manipulation takes place (for a review see [139]). For instance, mice completely lacking the insulin receptor in the whole body, or lacking one of the two insulin genes, die soon after birth due to highly elevated circulating glucose levels. Along the same lines, mice lacking insulin receptor specifically in muscle develop a metabolic syndrome with increased fat stores and hypertriglyceridemia, confirming the central role of this tissue in glucose disposal. In stark contrast, however, mice lacking insulin receptor in white and brown adipose tissue have a 30% decrease in whole body triacylglycerol (triglyceride) content and have normal levels of circulating lipids, non-esterified fatty acids (free fatty acids) and glycerol. This is similar to leanness observed [upon adipose-specific removal of TOR-C1 function [via knockout of the Raptor (regulatory associated protein of mTOR) protein] in mice [140], indicating that insulin and TOR-C1 signalling help promote fat accumulation in this tissue. These differing effects reflect the function of each individual organ in organismal metabolic control. Furthermore, they highlight an antagonistic relationship, in which insulin signalling in muscle promotes the reduction of total body nutrient stores, whereas insulin signalling in adipose tissue promotes accumulation of nutrient stores.

Although it has not been tested directly, various studies in Drosophila suggest that the antagonistic relationship described above also holds true for the fly. Reduction of insulin signalling in the entire animal, for instance via ablation of the NSCs that produce ILPs 2, 3 and 5, yields animals that have elevated levels of circulating sugars and total body glycogen and lipids [18,20,21]. As another example, flies lacking the IRS Chico similarly have elevated whole body lipid levels [31]. In contrast, reducing insulin signalling specifically in adipose tissue, as observed both in Melt and miR-278 (microRNA-278) mutant animals, has the opposite effect, leading to reduced adiposity [66,141]. The same is observed with conditions that up-regulate insulin signalling. Flies mutant for PTEN, an inhibitor of the pathway, have reduced total body glycogen and lipid levels [39]. In contrast, adipose-specific activation of TOR activity leads to increased total body lipid levels [66].

In addition to these results, which parallel those in the mouse, other studies have highlighted that other tissues respond in a specific manner to insulin signalling in the fly. For example, although PTEN mutants have reduced total body glycogen and lipid levels, the nurse cells in the developing oocytes of PTEN mutants actually have enlarged lipid droplets [53], suggesting that nurse cells respond differently to the PTEN mutation compared with other tissues of the animal. Another recent study looked at the transcriptional response of adipose tissue and muscle to altered insulin signalling, and found that insulin signalling modulates the expression of approx. 2000 genes in each tissue [19]. However, only half of these genes were affected in both tissues, and the other half were specific for each tissue, reflecting the different role of each tissue in organismal metabolism.

Communication between tissues

Metabolic regulation by insulin takes place within the context of an intricate network of communication between various tissues (Figure 2). For instance, inhibition of insulin signalling specifically in larval muscles autonomously reduces muscle size and non-autonomously reduces the size of the entire body, most likely by regulating feeding behaviour [72]. Although this is not seen in mice [142], this highlights the importance of communication between tissues, which will be discussed below.

Figure 2 Communication between Drosophila tissues in the insulin signalling pathway

NSCs are one tissue that produces ILPs, which are secreted to influence insulin signalling in all other tissues of the animal. ILP transcription is a central node of regulation, modulated by a number of inputs including FOXO and S6K activity, as well as sNPF signalling from nearby neurons. Furthermore, FOXO activation in the fat body of the head regulates ILP expression in NSCs by an unknown mechanism and serotonin (5-hydroxytryptamine) signalling from nearby neurons regulates secretion of ILP2. ILP function is inhibited by a number of secreted factors including dALS, Imp-L2 and NLaz. Several feedback loops exist in the animal. For instance, insulin signalling in the prothoracic gland promotes ecdysone production, which in turn inhibits insulin signalling in other tissues such as the peripheral fat body. Insulin signalling in gut promotes expression of the α-1,4-glucosidase Tobi. Further details are described in the main text. a.a., amino acids; EcR, ecdysone receptor.

Brain

One important organ for insulin signalling in Drosophila is the brain. The brain contains seven NSCs in each hemisphere, which express ILPs 1, 2, 3 and 5 [11,17,18]. These NSCs have axon terminals in the larval ring gland and on the aorta, where the ILPs are secreted into the haemolymph. Reduced ILP2 expression in NSCs correlates with reduced insulin signalling in peripheral tissues, observed by reduced phosphorylation of Akt or increased nuclear localization of FOXO [64,143146], which is consistent with a model in which ILPs secreted from NSCs activate insulin signalling in most tissues of the animal. Secretion of ILPs from these cells has profound effects on other tissues of the animal, affecting tissue growth, lipid metabolism, carbohydrate metabolism and animal longevity. Knockdown of ILP2 expression in the NSCs by RNAi yields animals with elevated total body trehalose levels [22]. Ablation of the NSCs causes developmental delay, growth retardation, elevated carbohydrate levels in larval haemolymph and an extension of lifespan [18,20]. Recently, an α-1,4-glucosidase named Tobi (target of brain insulin) was identified as a peripheral transcriptional target of ILP signalling emanating from the brain [147]. Ablation of NSCs leads to a 17-fold down-regulation of Tobi expression in the gut. Consistent with its glucosidase function, Tobi overexpression causes a reduction in total body glycogen levels in adult flies [147].

A large number of factors have been reported to regulate ILP expression by NSCs, suggesting this is a central node of metabolic regulation in Drosophila. Neurons adjacent to the NSCs express sNPF (short neuropeptide F), the fly orthologue of mammalian neuropeptide Y [143]. In an elegant study, Lee et al. [143] showed that activation of sNPF signalling, via NPFR1 (sNPF receptor), in NSCs leads to activation of ERK signalling, inducing expression of ILPs 1, 2 and 3. This connection was confirmed in cell culture using CNS (central nervous system)-derived neural BG2-c6 cells, and is therefore direct. Thus increased sNPF production by sNPFnergic neurons resulted in increased insulin signalling in peripheral tissues like fat body, leading to a reduction in larval haemolymph carbohydrates. In contrast, reduced sNPF signalling caused an elevation in circulating sugars and increased median lifespan by 16–21%. Another set of neurons adjacent to the NSCs, serotonergic neurons, were also shown to regulate ILP activity [145]. This study focused on the function of NS3 (nucleostemin 3), a nucleostemin family GTPase. NS3 mutants have elevated serotonin signalling in serotonergic neurons, leading to inhibition of ILP2 secretion, but not expression, from NSCs by an unknown mechanism. This causes NS3 mutant animals to have low insulin signalling in peripheral tissues, to grow slowly and to reach an adult weight of only ~60% that of control animals [145,148]. Adjacent to the adult brain in the Drosophila head is fat tissue. Activity of the transcription factor FOXO in this adult head fat body somehow regulates ILP2 expression in NSCs [64]. Increased activity of FOXO in head fat leads to reduced ILP2 expression in NSCs, leading to reduced insulin signalling in peripheral tissues, such as peripheral fat, leading to increased animal lifespan. The mechanism by which FOXO activity in head fat causes reduced expression of ILP2 in NSCs remains to be investigated.

In addition to these inputs from nearby cells, several factors within the NSCs themselves have been found to regulate ILP expression. NSCs have high JNK (c-Jun N-terminal kinase) activity levels [146]. As in other tissues, this JNK activity was shown to activate FOXO in a cell autonomous manner in NSCs, causing a reduction in ILP2 expression and consequently a reduction in growth and an extension of lifespan [146]. Likewise, increased S6K activity in NSCs has been proposed as a mechanism to increase ILP expression [9], although further work will be necessary to test this directly. In both cases, the mechanism by which S6K and FOXO activity in NSCs regulates ILP expression remains unclear. A recent study showed that PKA (protein kinase A) activity in NSCs also affects insulin signalling in peripheral tissues, and regulates production of ecdysone in the prothoracic gland [149]. This presumably occurs via modified ILP expression in the NSCs caused by the altered PKA activity, although this was not tested directly. Finally, overexpression of Drosophila p53 in NSCs causes nuclear accumulation of FOXO in NSCs. Consistent with the studies mentioned above, this leads to reduced ILP2 expression, reduced insulin signalling in peripheral tissues and increased lifespan [144]. Future work will shed light on whether p53 loss-of-function in NSCs has the opposite effect to that seen with overexpression.

Ring gland

The ring gland in Drosophila is a central endocrine organ which produces ecdysteroids and juvenile hormones. It is composed of three parts, the prothoracic gland, the corpus allatum and the CC (corpus cardiacum). A number of connections involving insulin signalling have been described between the ring gland and other organs of the fly.

The CC secretes several factors that counteract insulin signalling. A recent study showed that CC cells produce Imp-L2, which binds ILP2 inhibiting its function and reducing growth non-autonomously [24]. This function is particularly necessary when animals are in adverse nutritional conditions. Under these conditions, Imp-L2 mutant animals are not able to reduce their insulin signalling in peripheral tissues such as fat body, leading to increased mortality [24]. In addition, CC cells, together with fat body cells, produce the Drosophila homologue of the vertebrate IGF-binding protein acid-labile subunit, dALS. dALS binds dILPs in a tertiary complex with IMP-L2, antagonizing ILP function, hence controlling animal growth as well as carbohydrate and fat metabolism [25]. Finally, CC cells produce AKH (adipokinetic hormone), a functional homologue of glucagon [150]. CC cells express ATP-sensitive potassium channels, the targets of sulfonylureas, such as glyburide and tolbutamide, used to stimulate glucagon secretion in diabetic patients. Exposure of Drosophila to tolbutamide led to a 40% increase in haemolymph glucose in a manner dependent on CC cells. This indicates that CC cells regulate glucose homoeostasis using glucose-sensing and response mechanisms similar to islet cells [150].

The second component of the ring gland, the prothoracic gland, is the site of synthesis of ecdysone. Ecdysone counteracts insulin signalling in a number of tissues including the fat body. Increased ecdysone levels lead to reduced PI3K activity in the fat body, assayed via the PtdIns(3,4,5)P3 level, as well as by increased nuclear FOXO [151,152]. The mechanistic details of how ecdysone signalling counteracts insulin signalling are not known. Nonetheless, the physiological results of this inhibition are clear. During metamorphosis, increased ecdysone titres lead to the induction of autophagy in the fat body, which is normally suppressed by TOR signalling [100,102]. Experimental manipulations that lead to increased ecdysone during larval life inhibit the growth-promoting function of insulin, leading to reduced growth of the entire animal [151,152]. Surprisingly, ecdysone is part of a feedback loop involving insulin signalling. Synthesis of ecdysone is positively regulated by PI3K and TOR activity in the prothoracic gland [151153]. This has several physiological consequences. First, as TOR can sense nutrients directly, independently of insulin signalling, this couples the availability of nutrients to developmental timing, which is regulated by ecdysone. When TOR activity is reduced in the prothoracic gland, the ecdysone peak that marks the end of larval development and the beginning of metamorphosis is abrogated, extending the duration of growth. Thus the developmental delay induced by food deprivation works in part via TOR signalling in the prothoracic gland [153]. A second consequence is that ILP2 produced by the NSCs is less ‘powerful’ than ILP2 produced in other tissues, due to two counteracting functions; ILP2 produced by NSCs acts directly on peripheral tissues to regulate their growth and metabolism but also increases insulin signalling in the prothoracic gland, which the NSCs innervate and this leads to increased ecdysone signalling which counteracts ILP function [149].

Fat body

In mammals, adipose tissue not only acts as a fat reservoir but also functions as an endocrine organ. The same holds true in Drosophila. The fat body appears to act as a sensor of nutrients, in particular amino acids, and secretes factors that non-autonomously regulate insulin signalling in other tissues. This was discovered by studying mutants for the cationic amino acid transporter called Slif [154]. When Slif is knocked-down in fat body, this causes a non-autonomous reduction in insulin signalling in other tissues, such as salivary gland cells, and a reduction in the total body size of the animal. These effects are partially rescued by co-expressing S6K in adipose tissue, implicating the TOR pathway in this process. Indeed, down-regulation of TOR function in adipose tissue also non-autonomously reduces insulin signalling and growth in other tissues [154]. This suggests that removal of Slif in fat body causes a reduction in intracellular amino acids, which is sensed by TOR. The authors of that study go on to show that knockdown of Slif in adipose tissue strongly suppresses expression of dALS in adipose tissue, suggesting that dALS might function as the secreted second messenger [154]. However, subsequent work has shown that dALS functions as an ILP antagonist, and not as an agonist, as would be required by this model [25]. Therefore further work will be required to identify the secreted messenger by which fat body signals to other tissues.

A third secreted factor, in addition to dALS and IMP-L2, was recently identified as an inhibitor of insulin signalling in Drosophila. Stress-responsive JNK signalling, as well as starvation were found to induce expression of the secreted lipocalin, NLaz, in fat body [26]. NLaz mutant larvae have reduced insulin signalling levels, as measured via localization of a reporter for PI3K activity, GFP–PH [GFP (green florescent protein) fused to the PH domain of GRP1 (general receptor of phosphoinositides-1)]. Conversely, overexpression of NLaz in fat body caused a decrease in the GFP–PH signal. Overexpression of NLaz in fat body also led to a decrease in GFP–PH signal in nurse cells of the Drosophila oocyte, demonstrating non-autonomy of NLaz action [26]. NLaz mutant animals display all the canonical phenotypes of animals with increased systemic insulin signalling, such as larger size and reduced stores of glucose, trehalose, glycogen and triacylglycerols [26]. This is a second mechanism, in addition to its effects on ILP2 secretion from NSCs, by which JNK signalling regulates insulin signalling and hence metabolism in Drosophila.

SOME DIFFERENCES BETWEEN MAMMALS AND FLIES

Although the insulin signalling pathway is strikingly conserved between flies and mammals, there are also differences between the two systems, some of which are mentioned above. Some further differences are also described in this section.

Insulin and Ras

In mammalian cells, Ras activation has been implicated as an important mediator of the mitogenic effects of insulin stimulation [36,155]. Results so far in Drosophila have seen only a modulatory effect of Ras on the physiological effects of insulin (discussed above) and this might reflect a difference between Drosophila and mammals. Alternatively, the Drosophila results agree with those presented in a study performed in CHO (Chinese-hamster ovary) cells, where inhibition of Ras and MAPK activation downstream of insulin signalling does not have an impact on proliferation (assayed via DNA synthesis), activation of S6K or glycogen synthase [156], suggesting that in this system the effect of Ras is also modulatory.

GLUT4 (glucose transport protein 4) and glucose

In mammals, one of the principal effects of insulin signalling is to promote cellular glucose update via the GLUT4 transporter [157]. In response to insulin signalling, GLUT4 translocates to the cell surface resulting in an increased glucose uptake. In Drosophila, this appears not to be the case. Insulin stimulation does not alter [3H]-labelled 2-deoxyglucose uptake in Drosophila Kc cells [158], nor does manipulation of PI3K or Akt activity in S2 cells [159]. However, one difference between Drosophila and mammals is that the major circulating sugar in Drosophila is trehalose, not glucose. Knocking-down ILP2 expression in the NSCs by RNAi yields animals with elevated total body trehalose levels [22], indicating that insulin signalling is indeed involved in homoeostasis of circulating trehalose. Therefore it would be interesting to test whether cellular uptake of trehalose is induced upon insulin stimulation.

Glucocorticoids

In mammals, signalling through the glucocorticoid receptor is known to have strong effects on the insulin signalling pathway [160]. Drosophila has no obvious orthologue of the glucocorticoid receptor [161], suggesting that one major layer of regulation might be missing in the fly.

Reduced complexity of the pathway

Although Drosophila contains a homologue for virtually all components of the insulin signalling pathway, the overall complexity of the pathway is somewhat reduced in flies compared with mammals. This is because flies often have a single homologue that corresponds to multiple mammalian ones. For instance, humans have two S6K proteins, three 4E-BP proteins and multiple FOXOs, whereas flies only have a single S6K, 4E-BP and FOXO protein. On the one hand, this means that removal of gene function in the fly is more likely to give strong and clear effects, as there is less redundancy compared with the mammalian system. On the other hand, direct comparison between the fly protein functions and the mammalian ones may not be easy, as the proteins in the mammalian system may have specialized functions.

CONCLUSIONS

Some aspects of mammalian metabolic control, such as the leptin signalling pathway, do not exist in the fly. These mechanisms can be considered additional layers of metabolic regulation that were added during mammalian evolutionary on top of the more basic ones found in all animals. Clearly these mechanisms cannot be studied in the fly. Insulin signalling, however, appears to be both present and highly conserved, both at the molecular level and the physiological level, probably due to the basic cellular functions it regulates. Therefore it is fruitful to exploit the powerful genetic tools available in the Drosophila model organism in order to effectively study the molecular mechanisms by which insulin regulates metabolism.

FUNDING

The work in my laboratory was supported by a Helmholtz Young Investigator grant and an European Union FP7 grant under the MITIN (integration of the system models of mitochondrial function and insulin signalling and its application in the study of complex diseases) project.

Abbreviations: 4E-BP, 4E-binding protein; AKH, adipokinetic hormone; AMPK, AMP-activated protein kinase; Atg1, autophagy-specific gene 1; CC, corpus cardiacum; CREB, cAMP-response-element-binding protein; dALS, Drosophila acid-labile subunit; dLip4, Drosophila lipase 4; eIF4B, eukaryotic initiation factor 4B; eIF4E, eukaryotic initiation factor 4E; ERK, extracellular-signal-regulated kinase; FOXO, forkhead box O; GAP, GTPase-activating protein; GEF, guanine-nucleotide-exchange factor; GFP, green florescent protein; GLUT4, glucose transport protein 4; GRB2, growth-factor-receptor-bound protein 2; GSK-3β, glycogen synthase kinase 3β; IGF, insulin-like growth factor; ILP, insulin-like peptide; Imp-L2, imaginal morphogenesis protein-late 2; InR, insulin-like receptor; IRS, insulin receptor substrate; JNK, c-Jun N-terminal kinase; MAP4K3, MAPK kinase kinase kinase-3; MAPK, mitogen-activated protein kinase; Melt, Melted; NLaz, neural lazarillo; NS3, nucleostemin 3; NSC, neurosecretory cell; Path, pathetic; PDK1, phosphoinositide-dependent kinase 1; PH, pleckstrin homology; PI3K, phosphoinositide 3-kinase; PKA, protein kinase A; PP2A, protein phosphatase 2A; PRAS40, proline-rich Akt substrate of 40 kDa; PTEN, phosphatase and tensin homologue deleted on chromosome 10; RNAi, RNA interference; S6K, RPS6 (ribosomal protein S6)-p70-protein kinase; Sgg, shaggy; SIK2, salt-inducible kinase 2; Slif, slimfast; sNPF, short neuropeptide F; NPFR1, sNPF receptor; SREBP, sterol-regulatory-element-binding protein; Step, steppke; TIF-IA, transcriptional intermediary factor-IA; Tobi, target of brain insulin; TOP, terminal oligopyrimidine tract TOR, target of rapamycin; TOR-C1, TOR complex 1; TOR-C2, TOR complex 2; Tsc1, tuberous sclerosis complex 1; Tsc2, tuberous sclerosis complex 2; Wdb, widerborst

References

View Abstract