Increased extracellular adenosine in Drosophila that are deficient in adenosine deaminase activates a release of energy stores leading to wasting and death

A genetic deficiency of adenosine deaminase (ADA) is accompanied by greatly elevated levels of its substrates adenosine and deoxyadenosine. This deficiency causes severe combined immunodeficiency (SCID) in both humans and mice, as well as pathological changes in other tissues (Blackburn et al., 1998; Hershfield and Mitchell, 2001). Elevated levels of extracellular adenosine are also observed during multiple organ failure (Hasko et al., 2002), the cause of 50–80% of all deaths in surgical intensive care units.

Adenosine is an endogenous purine nucleoside that has multiple functions. It is a structural component of nucleic acids, but it also plays an important role in physiological regulation both inside and outside the cell. Inside the cell, it acts as an intermediate in the biosynthetic pathway leading to ATP. However, it can be released from cells or formed extracellularly from ATP/ADP/AMP. The extracellular form plays an important role in controlling various processes, including neuromodulation (Dunwiddie and Masino, 2001), vasoconstriction and hormone release (Nyce, 1999). The release of adenosine in some tissues has been shown to be stimulated by hypoxia (Buck, 2004). Specific inflammatory stimuli, such as bacterial products, are also capable of triggering adenosine release from immune cells (Bodin and Burnstock, 1998). Adenosine regulates cellular functions by binding to G-protein-coupled adenosine receptors (A1, A2a, A2b and A3 in mammals) that can regulate intracellular cAMP (Latini and Pedata, 2001).

The role of adenosine as a signaling molecule is in many ways indistinguishable from that of a hormone and it is sometimes called a ‘stress hormone’ (Hasko et al., 2002). Compared with typical hormones, which are usually secreted from a storage pool or synthesized from a precursor in response to a specific external stimulus, adenosine formation is typically a consequence of the metabolic status within the secreting cell, as is the case for regulatory metabolites. Therefore, adenosine can be seen as a regulatory metabolite that has extended its sphere of action beyond the cell that generates it (Newby, 1985).

In animals, there are two groups of enzymes with adenosine deaminase activity, referred to as ADA1 and ADA2. Although the protein responsible for ADA1 activity, which when defective is associated with human SCID, has been known for long time (Valerio et al., 1984), the protein with ADA2 activity has only recently been isolated (Zavialov and Engstrom, 2005). ADA2 is encoded by human CECR1 gene, which is a member of a novel family of adenosine deaminase-related growth factors (ADGFs) (Maier et al., 2005).

We have established a Drosophila model to study the effects of increased levels of adenosine by targeting Drosphila ADGFs using homologous recombination mutagenesis (Dolezal et al., 2003). Six Drosophila ADGF genes are homologs of human CECR1 and thus belong to the family of ADA2 genes (Dolezelova et al., 2005). We showed that a product of the ADGF-A gene is a secreted, active adenosine deaminase (Zurovec et al., 2002) and that it is the major source of the ADA activity during Drosophila larval development (Dolezal et al., 2005).

The adgf-a mutation leads to dramatically increased levels of adenosine and deoxyadenosine in the larval hemolymph (insect blood) and to larval or pupal death, associated with fat body disintegration and formation of melanotic capsules in the late third-instar larvae (Dolezal et al., 2005). Melanotic capsule formation is accompanied by a massive release of hemocytes from the lymph gland. Both intracellular (by cell uptake) and extracellular (i.e. signaling) effects of the increased adenosine concentration might play roles in the adgf-a phenotype.

Drosophila possesses one homolog of the adenosine receptor (AdoR) (Dolezelova et al., 2007), which activates cAMP and calcium signaling similarly to mammalian A2B receptor by coupling to the G-protein subunits Gα_s and Gα_q. A rescue of the adgf-a mutant phenotype by the adoR mutation showed that a significant part of the mutant phenotype is caused by the signaling effect of the extracellular adenosine (Dolezal et al., 2005).

The increased level of extracellular adenosine in the adgf-a mutant could have various consequences, and the actual reason for the larval lethality of the adgf-a mutant had not been determined in our previous studies. Therefore, we decided to use a large genetic screen for genes suppressing the adgf-a mutant phenotype. The results of this screen and the work presented here demonstrate that the increased concentration of extracellular adenosine triggers an energy mobilization by increasing hemolymph glucose via adenosine receptor signaling, and that a deregulation of this process in the adgf-a mutant leads to its death.

To identify genes associated with the effects of elevated adenosine in vivo, we performed a genetic screen for dominant suppressors of the adgf-a mutant phenotype in flies (supplementary material Fig. S1). We used a Bloomington deficiency kit for chromosome X and II to screen for deletions that, when crossed to the adgf-a mutation, would show a significant effect on the viability of the adgf-a mutants. First, we simply screened vials for an increased presence of pupae homozygous for the adgf-a mutation. Adult flies homozygous for adgf-a almost never occur and there was a very low number of pupae, which usually appear a couple days after their heterozygous siblings. Thus, the presence of a higher number of homozygous pupae at the same time or shortly after the heterozygous siblings indicated suppression of the adgf-a mutant phenotype. This was often accompanied by further development through metamorphosis to adult flies.

We screened a total of 49 deletions for chromosome X and 73 deletions for chromosome II, and found 11 regions that suppress the adgf-a phenotype (supplementary material Table S1). We used microdeletions and transposable element insertions to further analyze the identified regions to find the particular genes with the suppressing effect in each region. The originally identified deficiency Df(1)Exel9050 in the cytological localization 10D5–10D6 was molecularly defined by the transposon insertions P{XP}d00034 and PBac{RB}FucT6^e02394, which were used to produce this deficiency (Parks et al., 2004) (Fig. 1A). The deficiency spanned over three genes, including the Phosphorylase Kinase γ gene (PhK-γ; Fig. 1A). We tried an available P element insertion P{EP}PhKgamma[EP779] (designated EP779 hereafter), which was inserted in the second exon of an alternatively spliced variant designated the PhK-γ-RC transcript (Fig. 1A). The results showed that the heterozygous EP779 insertion suppresses the adgf-a mutant phenotype and that this suppression effect is stronger when the EP779 insertion is homozygous (Fig. 1B). The effect of this insertion is more apparent at the morphological level in the rescued animals. The adgf-a mutant larvae show extensive fat body disintegration and, eventually, a complete degeneration associated with melanotic capsule formation (Dolezal et al., 2005) (Fig. 1C). Homozygous EP779 insertion largely suppresses this phenotype when the double mutants EP779; adgf-a are fatter than the single adgf-a mutants, and the extensive disintegration of fat body typical for the adgf-a mutants is not observed in the double mutants (Fig. 1C).

Phosphorylase kinase (PhK) is an enzyme that has a key function in glycogenolysis, where it catalyzes the phosphorylation of the glycogen phosphorylase, which in turn catalyzes the conversion of glycogen to glucose-1-phospahate. It has four subunits, and the γ-subunit is responsible for the catalytic activity. A loss of the zygotic expression of Drosophila PhK-γ does not affect normal embryonic and larval development; some pupal lethality is observed but the majority of mutant animals eclose to adults (Bahri and Chia, 1994).

On the basis of the suppression effect of the EP779 insertion on the adgf-a mutant phenotype, we concluded that carbohydrate metabolism might be affected in the adgf-a mutant larvae and pursued further investigation in this direction. Buch and colleagues (Buch et al., 2008) showed that the glycogen stores were significantly reduced when adult flies were kept on a pure yeast diet rather than on a cornmeal-based fly food. Therefore, we tried to rear the adgf-a mutant larvae on yeast (a protein-rich diet) supplemented with different amounts of carbohydrates. We found a striking sensitivity of the adgf-a mutants to the sugar content in the diet. When larvae were grown on a yeast diet supplemented with 5% sucrose they expressed the typical adgf-a mutant phenotype (Fig. 2A). However, when the mutant larvae were grown on a pure yeast diet without any additional sugar supplement, they all died during the larval stages whereas the heterozygous or wild-type larvae grew normally on this diet (Fig. 2A and data not shown).

By contrast, an increase of sucrose content to 10% led to a significant increase in the mutant pupation and especially in the eclosion of adult flies compared with those on a diet containing 5% sucrose (Fig. 2A). The late third-instar larvae were indistinguishable from the wild type on the 10%-sucrose diet; similarly to larvae rescued by the mutation in PhK-γ (Fig. 1C), they did not show any signs of fat body disintegration (image not shown).

In agreement with the suppression screen, the EP779 insertion in PhK-γ, both heterozygous and homozygous, also suppressed the adgf-a mutant phenotype on the 5%-sucrose diet (Fig. 2D). The suppression effect was very weak on a pure yeast diet and only when homozygous insertion was used (Fig. 2D).

The number of circulating hemocytes is greatly increased in the adgf-a mutant and this is associated with a melanotic capsule formation (Dolezal et al., 2005). The sugar-enriched diet decreased the number of circulating hemocytes to a normal level (Fig. 2B) and the mutants very rarely formed melanotic capsules (data not shown). It seems then that the fat body of the adgf-a mutant larvae reared in more restrictive conditions degenerates due to the excessive release of stores, and that the described behavior of hemocytes is a secondary reaction to this degeneration. When the fat body does not degenerate, i.e. on a diet containing 10% sucrose, the hemocytes do not express such behavior.

There are five additional ADGF genes with proven or putative adenosine deaminase activity in Drosophila (Zurovec et al., 2002). Therefore, we decided to test whether the other members of the ADGF family also play a role similar to adgf-a in larval sensitivity to dietary restriction. We used the available combinations of mutants produced by homologous recombination: adgf-c+adgf-d and adgf-a2+adgf-b (Dolezal at al., 2003). Neither of the two combinations showed sensitivity to the dietary restriction (Fig. 2C), demonstrating that none of the four genes (ADGF-A2, ADGF-B, ADGF-C and ADGF-D) played an important role in the observed effects, at least not during larval stages. It is important to stress here that besides ADGF-A, only the ADGF-D gene is expressed during larval development, particularly in the late third-instar larvae, under normal conditions (Zurovec et al., 2002).

We used duplications in the ADGF-A2, ADGF-A, ADGF-B cluster as the adgf-a mutant, particularly lines adgf-a^Karel and adgf-a^Gerda (Dolezal et al., 2003) (see Methods) in which only the ADGF-A gene is mutated on both sides of the duplication. Therefore, we wanted to verify that the observed sensitivity to dietary restriction was solely due to the adgf-a mutation. So, we tested a combination of adgf-a^Am2 over adgf-a^Fr3, which were the true single mutants for the ADGF-A gene recovered by a reduction of the duplications (Dolezal et al., 2003) (depicted as ‘single adgf-a’ in Fig. 2C). This combination fully phenocopied the adgf-a^Karel/adgf-a^Gerda mutant (described as adgf-a in the rest of the text; compare Fig. 2A and 2C), demonstrating that the observed effects were solely due to the mutation in the ADGF-A gene.

We took advantage of the lethal effects of dietary restriction on the adgf-a larvae to test the role of the extracellular adenosine signaling in these effects. We had previously shown that a mutation in the single Drosophila adenosine receptor (AdoR) suppressed the deleterious effect of elevated levels of adenosine in the adgf-a mutant (Dolezal et al., 2005). As expected, the adoR mutation suppressed the lethality of the adgf-a mutant on the yeast diet supplemented with 5% sucrose, which was similar to cornmeal fly food (Fig. 2D). The same mutation also greatly suppressed the enhanced lethality of adgf-a on the pure yeast diet. Fig. 2D shows that almost all the adoR adgf-a double mutants pupate compared with no pupation of either single adgf-a mutant or mutants homozygous for adgf-a and heterozygous for adoR. A significant number of the double mutants also emerged as adults.

The rescue by the mutation in PhK-γ and the sensitivity to dietary restriction indicated that energetic metabolism was affected in the adgf-a mutant. Therefore, we measured the levels of circulating carbohydrates in the hemolymph of the control and mutant larvae as one of the markers of energetic metabolism. We collected hemolymph of the pre-wandering third-instar larvae of selected genotypes reared on different diets, eliminated the cells (mainly circulating hemocytes) by centrifugation, and measured the levels of hemolymph glucose and trehalose in parallel with adenosine levels in the same sample using mass spectroscopy analysis.

Similar to our previous observation (Dolezal et al., 2005), although the extracellular adenosine concentration never exceeded 0.2 μM in the hemolymph of the wild-type or heterozygous adgf-a larvae, it increased to 2.5 μM in the adgf-a mutants (Fig. 3A, top). The glucose level in the hemolymph was doubled in the adgf-a mutants compared with wild-type larvae (Fig. 3A, middle). The increased survival of the adgf-a mutant reared on 10% sucrose was not associated with a decrease of extracellular adenosine because its level increased similarly with both the 5%- and 10%-sucrose diets (Fig. 3A, top). The level of glucose was also similarly increased in the adgf-a mutant reared on diets containing either 5% or 10% sucrose (Fig. 3A, middle).

The significant increase of circulating glucose in the adgf-a mutants is in agreement with the results above, suggesting an affected energetic metabolism. The blocking of adenosine signaling by the adoR mutation decreases the level of circulating glucose in the adgf-a mutant to a normal level (Fig. 3B, middle), which is in agreement with the suppression of the sensitivity to the dietary restriction. Although the level of extracellular adenosine decreases in the adoR adgf-a double mutant compared with adgf-a, its level is still a magnitude higher than that of the wild type or heterozygous adgf-a control (Fig. 3B, top) showing that it is the blockage of AdoR signaling that is responsible for the observed rescue effects. Interestingly, the extracellular adenosine is similarly increased in the EP779; adgf-a double mutant and in adoR adgf-a (Fig. 3B, top) but, compared with adoR, the mutation in PhK-γ does not lower the glucose level in the hemolymph of the adgf-a mutant (Fig. 3B, middle).

We also measured trehalose, a non-reducing sugar formed from two glucose units, which is present both in circulation and tissues and serves as energy storage (glucose can be rapidly produced by trehalase) and protectant molecule in various stresses in prokaryotes, yeasts, plants and invertebrates (Elbein et al., 2003). The trehalose level in the hemolymph did not change in the adgf-a mutant compared with heterozygous siblings both on 5%- and 10%-sucrose diets (Fig. 3A, bottom) nor in the adoR adgf-a double mutant (Fig. 3B, bottom); there was a small increase in trehalose concentration in the EP779; adgf-a double mutant. Our measurement using mass spectroscopy showed levels of trehalose that were a magnitude higher than those of glucose, which is typical for most insects and other invertebrates and is associated with the role of trehalose (Wyatt and Kalf, 1957; Elbein et al., 2003; Chung, 2008). Data on Drosophila differ in different reports, sometimes showing same levels of glucose and trehalose (e.g. Lee and Park, 2004; Buch et al., 2008).

We collected the adgf-a mutant larvae and their heterozygous siblings reared on cornmeal diet at different times to produce a developmental profile of carbohydrate stores, i.e. glycogen and trehalose. Fig. 4A shows that normal larvae progressively accumulated glycogen stores while feeding, reaching a peak before they start wandering. Then after the larvae stop feeding, which was associated with a decrease in hemolymph glucose (Fig. 3C, middle), the glycogen stores started a rapid decline. At least part of the glycogen utilization probably goes into the production of trehalose, the concentration of which increased in the hemolymph during wandering (Fig. 3C, bottom).

Fig. 4A demonstrates that the adgf-a mutant larvae accumulated glycogen stores more slowly than heterozygous siblings and that they were not able to reach the pre-wandering peak present in the heterozygous siblings. There was a similar amount of glycogen stores in the prepupal stage of those adgf-a mutants reaching this stage as in heterozygous siblings. Similarly to glycogen, the total trehalose also accumulated slower in the adgf-a mutant than in normal larvae but eventually similar levels were reached.

We also tested an effect of fasting on the carbohydrate stores. Fig. 4B shows that the speed of the glycogen breakdown was similar in the adgf-a and normal larvae, i.e. after 4 hours half of the stores were utilized. Interestingly, the EP779 mutation in PhK-γ slowed down the glycogen breakdown in the adgf-a mutant (Fig. 4B). Thus, the slower glycogen breakdown achieved by the EP779 insertion might help the adgf-a larvae to accumulate more glycogen stores, which is in agreement with the rescue effect found in the genetic screen. Although there were not such dramatic changes in trehalose levels during starvation (larvae seem to preferentially use glycogen stores), EP779 insertion also resulted in a different speed from normal and adgf-a larvae.

Because an adipokinetic hormone (AKH) stimulates the release of energy stores in Drosophila (Lee and Park, 2004), we decided to test whether the glucose release observed in the adgf-a mutant was due to the action of AKH. Using the UAS/Gal4 system (Brand and Perrimon, 1993), the AKH producing cells in corpora cardiaca were ablated by expressing the pro-apoptotic gene Grim (UAS-Grim) (Zhou et al., 1997) driven by the AKH-Gal4 driver (Isabel et al., 2005) which is expressed exclusively in the AKH-producing cells in the corpora cardiaca (Fig. 5C). The complete ablation of these cells in otherwise wild-type larvae was vital (Isabel et al., 2005) (Fig. 5A). However, it did not lead to the increased survival of adgf-a mutant larvae (there was a small increase in number of adult flies on cornmeal and 5%-sucrose diet). The ablation of AKH-producing cells in the adgf-a mutants (Fig. 4C) resembled the adgf-a phenotype for all types of diets (Fig. 4A,B), demonstrating that AKH was not significantly involved in the observed effects of adenosine on the energy metabolism of larvae.

AdoR, similarly to AKH receptors, is coupled to G proteins. Although AdoR expression was previously observed using in situ hybridization in the ring gland where AKH-producing cells are also located, it was not detected in the fat body (Dolezelova et al., 2007). However, the AdoR expression was very weak, producing faint signals. Therefore, we used the more sensitive RT-PCR method to check the expression in fat body tissue isolated from the third-instar larvae. Fig. 5D demonstrates that the AdoR was indeed expressed in this tissue. This suggests that it is direct AdoR signaling to the fat body, working parallel to AKH signaling and using the same downstream machinery of G proteins, that is responsible for the observed effects of adenosine.

We previously showed that a deregulation of the adenosine levels due to the deficiency of Drosophila ADGF-A leads to larval or pupal death associated with a degeneration of larval fat body (Dolezal et al., 2005). We used a genetic screen, which led us to the identification of a mutant in the γ-subunit of the phosphorylase kinase (PhK-γ) suppressing the adgf-a mutant phenotype. PhK activates through the phosphorylation of glycogen phosphorylase (glycogenolysis) (Fig. 6). Thus, the suppressive effect of PhK-γ mutation corresponds to the observation that the fat body, the main organ for glycogen storage, does not degenerate in the rescued larvae. This rescue led us to a recognition that carbohydrate metabolism might be affected in the adgf-a mutant larvae.

The striking sensitivity of the adgf-a mutant to dietary restriction strongly supports this conclusion. We used a pure yeast diet, which is rich in proteins but poor in carbohydrates (Buch et al., 2008). The use of this diet led to enhanced larval lethality of the adgf-a mutant, whereas the wild-type larvae developed normally on this food. Supplementing the pure yeast diet with 5% sucrose gave larvae resembling the adgf-a mutant phenotype on the cornmeal-based fly food. An increase of sucrose to 10% then leads to a significantly higher mutant survival rate, with a greatly improved morphology of the mutant larvae and most of them not showing any sign of fat body degeneration. Thus, although lowering carbohydrates enhances the mutant phenotype, increasing the amount of carbohydrate in the diet most probably compensates the loss of energy reserves or allows the saving of more stores than in the restrictive conditions. Saving stores of energy is crucial for the larvae to continue in their development. The survival of the adgf-a mutant larvae is thus critically dependent on the amount of carbohydrate in the diet.

Finally, an analysis of carbohydrate concentrations in the hemolymph demonstrated that the balance between energy storage and release was indeed shifted in the adgf-a mutant towards release because the mutant larvae presented doubled levels of glucose in the hemolymph compared with the wild type. The elevated levels of extracellular adenosine in the adgf-a mutant are thus associated with hemolymph hyperglycemia.

Extracellular adenosine was not lowered in the mutant larvae rescued on a 10%-sucrose diet compared with mutants on a 5%-sucrose diet, demonstrating that the rescue by increased sugar is not due to lowering of the extracellular adenosine level. The level of glucose in the hemolymph was also similarly increased, most probably because the elevated adenosine increases the circulating glucose to these levels. Therefore, it is plausible that increasing the amount of carbohydrates in the diet allows the saving of more stores, while still keeping the hemolymph glucose levels high. This in turn is expressed as a thick fat body of the rescued larvae on 10% sucrose. The rescue by diets higher in sucrose further demonstrates that it is not hyperglycemia (at this level) that kills the adgf-a mutants.

The measurement of carbohydrate stores demonstrated that the accumulation of both glycogen and trehalose during feeding period is much slower in the adgf-a mutant larvae than in heterozygous siblings most probably because the elevated adenosine forces the circulating glucose levels to be kept high. This might have fatal consequences for larvae on a diet poor in carbohydrates that does not allow larvae to save enough stores to proceed normally in development, as seen with the adgf-a mutant on a pure yeast diet. Slowing down the glycogen breakdown (by mutation in PhK-γ, as seen in Fig. 4B) does not help much in this situation; EP779 increased the survival of adgf-a poorly on a pure yeast diet. However, it does help larvae on a 5%-sucrose diet, which seems to provide adgf-a with just the minimum amount of carbohydrate, sufficient to save enough stores, at least for some larvae (about a third of them) to proceed in development. Probably any situations requiring a mobilization of the energy from the glycogen stores shifts the thin balance on 5% sucrose so it does not allow the adgf-a mutants to pupate. In these conditions, slowing down the glycogen breakdown helps the adgf-a mutants, as demonstrated by a significant increase in pupariation rate of the EP779; adgf-a mutants on 5% sucrose.

The main storage organ in insects is the fat body, a counterpart of mammalian liver and adipose tissue. The deposition of stores as well as the stimulation of their release is intensively studied in many insect models, which show that the basic regulation of carbohydrate metabolism is similar to that in higher organisms (Hudson et al., 1993). As shown in Fig. 6, glycogen synthesis and breakdown is regulated by glycogen synthase and glycogen phosphorylase. The energy-demanding processes in insects stimulate the release of energy from the fat body stores through the action of AKHs (reviewed in Gäde and Auerswald, 2003). AKH, a counterpart of mammalian glucagon, binds to G-protein-coupled receptors and, in the case of carbohydrate metabolism, activates glycogen phosphorylase via cAMP production and Ca²⁺ (binding to G_s and G_q subunits). This includes regulation by Ca²⁺-dependent PhK.

Extracellular adenosine can affect carbohydrate metabolism in the fat body by two possible mechanisms (Fig. 6): signaling through AdoR, and intracellularly by transport via adenosine transporters (Thorn and Jarvis, 1996) and a conversion to AMP by adenosine kinase inside the cell (Van De Wiele et al., 2002). Glycogen phosphorylase is, in addition to phosphorylation by PhK, also activated by allosteric effectors including AMP (Tsitsanou et al., 1999). AMP can also inhibit glycogen synthase through the action of AMP-activated protein kinase (AMPK) (Towler and Hardie, 2007). However, we show here that the blockage of AdoR signaling almost completely rescues the mutant larvae on a pure yeast diet. This is supported by resetting the glucose level in the hemolymph of the adoR adgf-a double mutants back to normal levels. These results clearly demonstrate the role of AdoR signaling in the regulation of carbohydrate metabolism in Drosophila. A similar role for adenosine signaling through adenosine receptors stimulating the glucose release was described for mammalian liver cells (Gonzalez-Benitez et al., 2002), suggesting that this role of extracellular adenosine and the mechanism of action are evolutionary conserved from flies to mammals.

Because AdoR is expressed in the Drosophila ring gland (Dolezelova et al., 2007) where the AKH-producing cells are located (Lee and Park, 2004; Isabel et al., 2005), we speculated that the effect of adenosine could be triggered by AKH release in response to AdoR signaling in the ring gland. However, the ablation of AKH-producing cells did not rescue the adgf-a mutant, which demonstrates that AKH is not involved in the observed effect of adenosine on carbohydrate metabolism. Previous analysis of AdoR expression by in situ hybridization did not reveal its expression in the fat body (Dolezelova et al., 2007). However, using a more sensitive method of tissue-specific RT-PCR, we show here that AdoR is indeed expressed in the fat body. Although we have not presented a definite proof, the clear suppressive effect of adoR mutation and AdoR expression in the fat body suggest that the direct signaling of extracellular adenosine to the fat body via AdoR might lead to the increase of glucose in the hemolymph. It was previously shown that Drosophila AdoR is, similarly to the AKH receptor, associated with G_s and G_q proteins, leading to formation of cAMP and Ca²⁺ release (Dolezelova et al., 2007). This leads to the conclusion that AdoR could serve as an additional receptor along with the AKH receptor for activating glycogenolysis via G-protein/PhK stimulation in the fat body (Fig. 6).

Glycogenolysis is not the only way to increase glucose in the hemolymph, especially during feeding. The mutation in PhK-γ does not decrease the glucose level in the adgf-a mutant whereas the adoR mutation does. The extracellular adenosine thus instructs the organism to keep hemolymph glucose (originating most probably from dietary sucrose) levels high, while lowering glycogen storage. Therefore, there might be an additional AdoR signaling effect, besides that on glycogenolysis, that influences how much of the carbohydrate obtained from food will be used for storage and how much will be kept in the hemolymph for immediate use (Fig. 6).

Extracellular adenosine can be then seen as a true anti-insulin hormone. This agrees with a concept proffered by the most recent work of Piña’s group, that adenosine can play an even stronger anti-insulin hormone-like role in the mammalian system under certain conditions than classical anti-insulin hormones such as adrenaline or glucagon (Cortés et al., 2009). It is important to stress that although our study focuses solely on carbohydrate metabolism, adenosine might have a similar effect on lipid metabolism.

The observation that accumulated extracellular adenosine can lead to death of the organism through an impairment to the process of saving energy stores might help to better understand certain pathologies. The immune response is energetically costly (Derting and Compton, 2003; Freitak et al., 2003) and a progressive loss of energy stores – also called wasting – is observed during certain chronic infections in both flies and humans (Dionne et al., 2006; Lazzaro and Galac, 2006). In spite of a potentially high biomedical importance, the network of the molecular signals leading to wasting is largely unexplored. Results of this work suggest that extracellular adenosine could be one of those signals.

Adenosine is well known as a stress hormone and it is produced during the immune response (Hasko et al., 2002); therefore it could play an important hormonal role in energy allocation during stress conditions such as infection. Our previous work suggests a connection between adenosine signaling and Toll signaling, which is an important regulator of the immune response in flies (Dolezal et al., 2005). A requirement to use virtually sterile conditions to obtain reproducible results in this work (see Methods) suggests a high sensitivity of the adgf-a mutants to infectious stress. Microarray analysis in the fly model for wasting also shows changes in the expression of the adenosine-regulating enzymes (Dionne et al., 2006). It is important to further investigate this intriguing role for extracellular adenosine, and our work presented here demonstrates that Drosophila can be now utilized as a useful, genetically tractable model to investigate the stress-hormone-like roles of extracellular adenosine in energy allocation in different conditions in vivo.

We have shown that (1) increased extracellular adenosine due to adenosine deaminase deficiency leads to high levels of glucose (hyperglycemia) in the hemolymph during Drosophila larval development. (2) This effect is carried out by adenosine signaling through the adenosine receptor. (3) Adenosine works in this regard as an anti-insulin hormone in parallel to AKH, a counterpart of mammalian glucagon. This role and mechanism seems to be well conserved between flies and mammals, making Drosophila a good model for these effects. (4) The hyperglycemia associated with the extensive adenosine signaling further leads to the impairment in building carbohydrate stores during the larval period and to high sensitivity to sugar restrictions in the diet. This can be partially suppressed by lowering the glycogen breakdown by a mutation in PhK. In suboptimal conditions or with diet restrictions, the described effects can lead to death of the organism. Because adenosine signaling is associated with the immune response and the response to stress in general, our results make extracellular adenosine a good candidate for a signal involved in the wasting syndrome, which has an impact on various pathologies.

The adgf-a mutant (chr. III) was prepared by crossing w; adgf-a^Karel/TM3 Ser Act-GFP with w; adgf-a^Gerda/TM3 Ser Act-GFP and selection was based on the lack of GFP in w; adgf-a^Karel/adgf-a^Gerda larvae. Both parent lines carried duplications of the ADGF-A2, ADGF-A, ADGF-B cluster interrupted by a mini-white marker in which only the ADGF-A gene was mutated on both sides of each duplication (Dolezal et al., 2003). Homozygotes for each duplication as well as transheterozygotes adgf-a^Karel/adgf-a^Gerda were functionally verified to be a loss-of-function mutant just for the adgf-a gene (Dolezal et al., 2003). Because these two lines were produced independently and always kept separately, their combination was used as the adgf-a mutant to avoid the effects of any side mutations. This combination also gave us the advantage of the mini-white marker presence, making certain crosses easier. Additional lines that were mutant for the ADGF-A gene without the mini-white marker (adgf-a^Am2and adgf-a^Fr3) (Dolezal et al., 2003) were used only when specified in the text. Other strains used in this work included: adgf-a2 adgf-b and adgf-c adgf-d (double mutants on chr. III) (Dolezal et al., 2003); adoR (mutant in the adenosine receptor on chr. III) (Dolezal et al., 2005); adoR adgf-a^Karel and adoR adgf-a^Gerda recombinant double mutants; P{EP}PhKgamma[EP779] (on chr. X, BDGP disruption project); UAS-Grim (insertion on chr. II, gift of John Nambu, University of Massachusetts, Amherst, MA); AKH-Gal4 (pAKH02 insertion on chr. III, gift of Jan Veenstra, CNRS, University of Bordeaux, France).

A Bloomington deficiency kit for chromosome X and II (a set available in 2005, generated by the Bloomington Deletion Project, the DrosDel Project and Exelixis) was used for the screen. Deletions were crossed with the adgf-a mutation as shown in supplementary material Fig. S1. The adgf-a homozygous progeny were scored for numbers of pupae and adult flies and for the presence of melanotic capsules, in three consecutive vials.

Flies were grown on standard cornmeal medium (8% cornmeal, 5% sugar, 4% yeast, 1% agar, 0.16% methylparaben). To count viability and observe the development, flies were allowed to lay eggs for 4 hours on agar juice plates supplemented with fresh yeast paste. Embryos of ∼20 hours old were collected by brush, washed in water to clean the yeast, and washed for 30 seconds in 96% ethanol to eliminate infectious agents. The ethanol-treated embryos were transferred to sterile agar plates with the desired diet to complete embryonic development. The early feeding between hatching to first-instar larvae and the transfer to vials seemed to have an important impact on the results of feeding experiments; therefore, the washed embryos were always transferred to the same diet as was in the vial. The first-instar larvae of desired genotype scored by GFP expression were collected with a sterile needle into vials (20–30 individuals per vial) supplemented with 80 μl of penicillin/streptomycin solution (Sigma; 10000 U penicillin/10 mg streptomycin per ml). The washing of embryos and treatment with penicillin/streptomycin appeared to be crucial for obtaining consistent data because the adgf-a mutant seems to be very sensitive to any environmental stress. The development from larva to adults was observed and scored for each individual vial. All counting experiments were done at least in triplicates and repeated at least twice at different times.

For the dietary restriction experiments, the diets were prepared as follows: dry yeast and agar were dissolved in distilled water to achieve 8 and 1.2% w/v, respectively. The solution was boiled for 10-15 minutes to inactivate live yeast while maintaining the total volume. At the end, sucrose was added to make 5 or 10% w/v (for the pure yeast diet, no sucrose was added) and briefly boiled. The suspension was cooled down to ∼60°C, methylparaben (Sigma) was added to 0.16% and the final diet was poured into the vials. A few hours before collecting the first-instar larvae, 80 μl of penicillin/streptomycin solution was added on the surface of each vial and allowed to dry. Agar plates for washed embryos contained the same diet with 4% agar.

During experiments we noticed that the length of cooking is important because cooking for more than 30 minutes led to a decrease in the adgf-a mutant survival (not the wild-type control). It seems that certain carbohydrate constituents (maybe also other important nutrients) present in the yeast and/or cornmeal become unavailable with longer cooking or boiling in a microwave oven. Therefore, we used strictly 30 minutes at 90°C using a thermomixer and all comparing experiments were done using a diet from the same cooking batch.

We crossed flies carrying UAS-Grim with flies carrying AKH-Gal4 to ablate AKH-producing cells. In the rescue experiments, AKH-Gal4 was recombined with chromosome carrying adgf-a^Karel to obtain UAS-Grim/+; AKH-Gal4 adgf-a^Karel/adgf-a^Gerda. The presence or absence of AKH-producing cells was determined by wholemount immunostaining of the third-instar larvae using a primary rabbit anti-AKH antibody (Isabel et al., 2005), secondary anti-rabbit Cy3.5 (Rockland) and DAPI staining.

Oregon-R and yw larvae of the late third-instar (118 hours after egg laying) were dissected in Ringer solution and the fat body separated from the rest of the larval body. Total RNA was then isolated from both fat body tissue and from the remaining carcass using Tri Reagent (MRC), following the manufacturer’s instructions. Isolated RNA was treated with Turbo DNase (Ambion) and reversely transcribed into the first strand cDNA using Superscript III (Invitrogen). Control reactions (RT-) were prepared simultaneously without the addition of Superscript. The following primers were used for PCR amplification of Actin: Actin5CF 5′-TACCCCATTGAGCACGGTAT-3′ and Actin5CR 5′-GGTCATCTTCTCACGGTTGG-3′; for amplification of AdoR: Probe2F 5′-AGATCTCGAGCGAGAAATGTGGAAC-3′ and Probe2R 5′-CTCCTGAGCCTGGCCCATG-3′. AdoR was amplified by 34 cycles without the additional final elongation step.

Hemolymph was collected from third-instar pre-wandering larvae. Some 10-15 larvae were teared onto a parafilm-covered microscopic slide on ice in 1 μl of inhibition solution (1 mM thiourea, 20 mM EDTA). Then, 3 μl of the hemolyph were transferred by automatic pipette into 60 μl of cooled extraction medium [MeCN/water 8:2 (v/v) with 20 mM ammonium acetate, pH 7.5-8.1, spiked with the ¹³C-labeled standards: 0.5 μM ¹³C₅-adenosine and 20 μM ¹³C₆-glucose], mixed and centrifuged (1600 g for 2 minutes at 4°C). To 50 μl of the supernatant, 50 μl of MeCN was added and the mixture stored at −20°C until analysis using liquid chromatography coupled with tandem mass spectrometry. HPLC analysis was carried out on a quaternary Rheos Allegro UPLC system (Thermo, San Jose, CA) equipped with an HTC PAL autosampler system (CTC Analytics, Switzerland) cooled at 5°C during analysis. Separation of the analytes was performed on a Luna hydrophilic interaction liquid chromatography column (150 mm × 2.0 mm × 3 μm; Phenomenex, Torrance, CA) at room temperature with an injection volume of 5 μl. Gradient elution at a flow rate of 250 μl/minute was used, with solvent A [MeCN/water 95:5 (v/v) with 20 mM ammonium acetate, pH 8.1] and solvent B [MeCN/water 5:95 (v/v) with 20 mM ammonium acetate, pH 8.1] going from 90:10 (A/B) to 20:80 (A/B) in 10 minutes, followed by equilibration at 90:10 (A/B) for 8 minutes. The detection was carried out using a LTQ XL linear ion trap mass spectrometer (Thermo) equipped with an electrospray ion source operated simultaneously in the positive and negative ion mode, setting the sprayer voltages at 4.5 and −3.5 kV, respectively, and the ion source temperature at 350°C. Positive ions were detected after the collision-induced decomposition in the MS/MS scan mode by monitoring the transitions m/z 368.2 →136.1 and 373.2→136.1 for adenosine and ¹³C₅-adenosine, respectively, and in the selective ion monitoring (SIM) scan mode by monitoring of the [M-H]⁻, [M+Cl]⁻ and [M+CH₃COO]⁻ adduct ions of glucose, ¹³C₆-glucose and trehalose at m/z 179.1, 215.1, 239.2; m/z 185.1, 221.1, 245.1 and m/z 341.2, 377.2 and 401.3, respectively. Metabolite levels in hemolymph were calculated from the calibration data obtained by the standard addition method, by means of the previously measured hemolymph samples spiked with known concentrations of native standards against ¹³C₅ adenosine (i.e. adenosine) and 13C₆-glucose (glucose and trehalose).

Larvae of desired age were washed from the diet with cold PBS. Some10–15 larvae were crushed in 200 μl of PBS on ice, then immediately heated to 70°C for 7 minutes. Before heating, an aliquot for the protein measurement was taken. Samples were stored at −20°C (protein aliquots at −70°C). The thawed samples were centrifuged at 15,000 g for 7 minutes. The supernatant was transferred to a new tube. Glucose was measured with the kit GAGO-20 (Sigma) in 96-well plates. Enzyme solution (100 μl) was added to 45 μl of the sample and incubated for 30 minutes at 37°C. The reaction was stopped with 100 μl of 12 N H₂SO₄ and measured at 540 nm within 10 minutes of stopping the reaction. Glycogen and trehalose were first converted to glucose with amyloglucosidase (Fluka-Sigma 10115) or trehalase (Sigma T8778) in 30 μl of reaction mixture per well and measured with the kit GAGO-20. For glycogen, 5 μl of 100 U/ml solution of amyloglucosidase was added to 4 μl of sample and filled to 30 μl with PBS. Then, 100 μl of GAGO-20 enzyme solution were added and the reaction incubated at 37°C for 40–50 minutes. For trehalose, 3 μl of the sample was mixed with 2 μl PBS and 25 μl trehalase buffer (5 mM Tris pH 6.6, 137 mM NaCl, 2.7 mM KCl and 14.8 mU/ml trehalase) (Teleman et al., 2005), incubated for 20 hours at 37°C, and then measured with GAGO-20. Proteins were measured by the Bradford assay.

For starvation experiments, larvae were collected at 108 hours after laying eggs and were starved for either 2 or 4 hours on filter paper moistened with PBS.

Circulating hemocytes were counted as previously described (Dolezal et al., 2005).

The following statistical tools were used: two-sample assuming equal variances t-test, ANOVA Tukey’s HSD (honestly significant difference) test pro, and multiple comparison among regressions Tukey’s HSD test (Zar, 1996).

We thank the Bloomington Stock Center, John Nambu and Jan Veenstra for fly strains and antibodies. We especially thank Katerina Kucerova and Pavla Kruzberska for major technical assistance and Zdenek Fric for help with statistical analysis. This work was financially supported by the Grant Agency of the Academy of Sciences of the Czech Republic (KJB501410602), by the Grant Agency of the Czech Republic (204/09/1463 and 204/09/H058) and by the Ministry of Education, Youth and Sports of the Czech Republic (MSM6007665801).