Natural variation in food acquisition mediated via a Drosophila cGMP-dependent protein kinase

Animals must feed to survive, but foraging often involves a delicate balance of benefits and risks. The benefits of nutrient acquisition are weighted against elevated exposure to threats from predators, parasites,exhaustion or toxins, for example. Many risk-mitigation adaptations, which help to increase the foraging benefit-to-risk ratio, have evolved among animals. These include behavioral strategies, such as foraging at times when predators are less active or abundant, or when food is less well defended. Other adaptations include altered metabolic or physiological traits. Some animals, for example, may gorge themselves to rapidly increase fat supplies when food is plentiful and then deplete these reserves in times of famine. Alternatively, other animals, such as reptiles, adjust their metabolism or gut absorption levels to maintain body mass in response to food scarcity(Stubbs and Tolkamp, 2006; Wang et al., 2006).

Within a species, individuals may exhibit polymorphic and/or plastic responses to fluctuating food availability. Naturally varying genes involved in plastic responses to changes in food availability are difficult to identify, often due to the subtlety of their influence. Nevertheless,within-species genetic polymorphisms affecting foraging behaviors have been identified and characterized in the fruit fly, Drosophila melanogaster (Osborne et al.,1997), and the nematode, Caenorhabditis elegans(de Bono and Bargmann, 1998). Here we establish a novel role for the D. melanogaster gene, foraging (for), which influences foraging locomotory behaviors, in the mediation of food acquisition strategies and plastic responses to food availability.

The fly gene, for, encodes a cGMP-dependent protein kinase (PKG),and natural allelic variation in for affects rover- or sitter-type foraging locomotion (de Belle et al.,1989). Larvae with the wild-type rover (for^R)allele travel further when feeding and move more between food patches than those with only sitter (for^s) alleles(Osborne et al., 1997; de Belle et al., 1989; Sokolowski et al., 1983; Sokolowski, 1980). Well-fed rovers have higher PKG enzyme activities compared to sitters(Osborne et al., 1997; de Belle et al., 1989), and transgenic expression of for in sitters is sufficient to induce rover-like elevations of PKG activity and foraging locomotion(Osborne et al., 1997).

We speculated that the increased foraging locomotion of rovers may be energetically costly (Berrigan and Lighton,1993), and thus, we hypothesized that the rover variants would require more energy and thereby differ in food acquisition compared to the sitter variants. Food acquisition was assessed by measuring rates of both food intake and nutrient absorption. Intriguingly, rover allelic variants have lower food intake than sitters despite having similar sizes and metabolic and developmental rates. This apparent paradox is clarified by our further observation that rovers may counterbalance their reduced intake with increased rates of nutrient absorption.

Despite different foraging styles, rovers and sitters occur at stable frequencies in natural populations (70% rovers: 30% sitters)(Sokolowski et al., 1997). However, selection experiments have demonstrated a selective advantage of the rover allele (for^R) compared to the sitter allele(for^s) when rover and sitter larvae are reared together for many generations in crowded conditions(Sokolowski et al., 1997). We postulated that this rover advantage may stem in part from an increased ability to tolerate or adjust to the more depleted medium found in crowded culture conditions. Thus, we hypothesized (1) that rovers will have enhanced survival compared to sitters in nutrient-restricted conditions, and (2) that this advantage may arise from differences in nutrient uptake. Indeed, our results indicate that, when reared in environments with relatively low nutrient levels, rover larvae have higher survivorship and faster development than sitter larvae. Moreover, changes in for expression can induce corrective behavioral modifications in response to food. That is, when food is scarce both rover and sitter larvae express similarly reduced PKG activity and food intake is elevated to a common maximal level. Rover larvae, however,retain their relatively elevated nutrient absorption efficiency, perhaps contributing to their survival advantage under such conditions.

The mechanisms by which changes in gene function and expression are naturally selected to confer adaptive behavioral responses to varying environments is a central question in behavioral genetics. This work uses the D. melanogaster rover/sitter polymorphism to investigate this question by addressing the role of for in mediating plasticity in food acquisition strategies.

The rover and sitter strains used in these experiments are homozygous for the for^R or for^s alleles on chromosome-2 and share co-isogenic third chromosomes from the for^R strain (de Belle and Sokolowski, 1987). The X-chromosomes shared by these strains were mostly from the for^R background. for^s2 is a sitter mutant generated on the rover genetic background and does not differ from for^R aside from the induced mutation within the for gene(Osborne et al., 1997; de Belle et al., 1989). In order to increase the expression of for, a DNA fragment encoding the complete forT2 amino acid sequence(Kalderon and Rubin, 1989) was subcloned into the transformation vector, pUAST, and transformed into w¹ embryos using standard techniques(Spradling and Rubin, 1982). The sitter second chromosome (containing the for locus) was substituted into the transgenic strain, creating w¹;for^s;UASforT2. These were then crossed to w¹;for^s;hsGAL4, resulting in an increase in expression of forT2 in an otherwise for^s genetic background. We relied on leaky expression of the hs promoter (e.g. Bainton et al., 2005);experiments were done at 23°C and no heat shock was applied.

Flies (Drosophila melanogaster) were maintained in 170 ml plastic culture bottles with 40 ml of standard culture medium at 25±1°C and a 12 h:12 h L:D photocycle. Standard culture medium contained 50 g Baker's yeast, 100 g sucrose, 16 g agar, 0.1 g KPO₄, 8 g sodium potassium tartrate, 0.5 g NaCl, 0.5 g MgCl₂ and 0.5 g Fe₂(SO₄)₃ per litre of tapwater. Larvae were reared from egg-hatch to mid-third instar (96±2 h post-hatch) at densities of 100 larvae per 35 ml of medium in 100 mm×15 mm Petri dishes.

Whole body measurement. Larvae were removed from food plates,washed in distilled water and groups of 10 were placed into circular wells (86 mm in diameter and 0.5 mm deep) previously filled with yeast paste (2:1 water:yeast) mixed with 0.08% Brilliant Blue R dye (Sigma, Mississauga, ON,Canada). The wells were then covered with 9 cm Petri plate lids. Larvae remained on this dyed yeast paste for varying amounts of time depending on the experiment. They were then boiled for 10 s, aligned on a microscope slide,placed under a dissecting microscope (Zeiss, Toronto, ON, Canada) and imaged using Northern Eclipse software (Empix Imaging, Mississauga, ON, Canada). Food intake was measured as the number of pixels (square pixels were used for quantification) in the image colored by the dye relative to the total number of pixels in the whole larval body, taken as a percentage. Image J software was used (ImageJ V 1.28, 2002 and ImageJ V 1.32j, 2004) for the digital quantification. Thirty larvae/strain/condition combinations were assayed. Initially food intake was measured after feeding larvae for 10, 15 and 20 min. Further experiments were done using the 15 min feeding period, which was representative of strain differences. All larvae were staged to mid-third instar prior to testing food intake(Demerec, 1994). When larvae were reared in high-quality food conditions this was 96±2 h post-hatch. When larvae were reared in 25% food this was 144–168 h post-hatch, and when larvae were reared in 15% food this was 168–192 h post-hatch,depending on humidity in the rearing incubator.

Fructose–agarose and glucose–agarose intake.Fructose– agarose and glucose–agarose food intake experiments were performed as above with the exception of food substrate. For fructose–agarose food intake experiments, 100 mm×15 mm Petri plates containing 1% agarose, 2 mol l^–1 fructose and 0.5%Carmine dye (Sigma) that were prepared 4 h prior to the test. For glucose–agarose food intake experiments, 45 mm×10 mm plates were made 24 h prior to testing with 2.3% agarose, 10% glucose and 0.5% Carmine dye. Small cuts were made in glucose–agarose plates to aid in feeding on a solid substrate. Larvae were left on sugar–agarose food substrate for 15 min before food intake was measured.

Gut measurement. Larvae were reared as above and fed dyed yeast paste for 15 min. Guts were dissected in phosphate-buffered saline (PBS) and gently elongated on a Sylgard™ base for imaging. Gut food content was quantified by calculating the amount of dye-colored pixels as a percentage of the total number of pixels in the gut. Twenty larval guts/strain/condition combinations were measured.

Methods for spectrophotometric analysis of food intake were modified from Edgecomb et al. (Edgecomb et al.,1994). Briefly, groups of 50 larvae were reared, as above, and fed yeast paste with 0.16% Erioglaucine dye (aka FD&C Blue No. 1, Sigma) for 15 min. After feeding, each group of larvae was washed 3× in distilled water, placed in 1.5 ml tubes and immediately frozen in liquid nitrogen. Larvae were then homogenized in 250 μl distilled water, centrifuged at 13 g for 10 min, and 225 μl of supernatant was transferred to a new 1.5 ml tube containing 50 μl 100% ethanol. Tubes were vortexed for 30 s and re-centrifuged for 10 min. 250 μl of supernatant was placed in a new tube, which was centrifuged at 13 g for 5 min. Then, 200 μl supernatant was placed in a 96-well crystal plate and the OD⁶³³read (SpectraMax Plus 384, Molecular Devices, Sunnyvale, CA, USA). Nine groups of 50 larvae per strain were analyzed.

Mouthhook contractions were measured for 15 min per larva on a glucose–agarose substrate described above. Contractions were measured when larvae were scraping the glucose–agarose substrate in and/or along a crack in the substrate. Individual larvae were immediately boiled following mouthook measurements and food intake was measured.

Length and width measurements were obtained from the digital images using ImageJ software. To establish dry mass, groups of ten larvae were placed in 1.5 ml tubes and dried for 48 h in a desiccator, after which masses were determined for 30 groups. All larvae were staged to mid-third instar prior to measurement.

Larvae were reared in food qualities defined as 100%, 75%, 50%, 25% or 15%until mid-third instar (Demerec,1994) for food intake tests and size measurement. Food quality was manipulated by decreasing the proportion of both yeast and sucrose in the standard culture medium (see recipe above) while keeping all other constituents constant.

Twenty newly hatched larvae were placed in a 9.5 mm×25 mm diameter glass vial with 10 ml of food medium (100% 75%, 50%, 25% or 15% quality). Animals were grown at 25°C on a 12 h:12 h L:D cycle with lights on at 08:00 h. Eclosed adult flies in each vial were counted and removed once every 24 h for 26 days.

Third-instar larvae were removed from food plates, washed gently and stuck to a clear microscope slide using double-sided stick tape, ventral side facing up. The number of gut contractions in the anterior midgut, acidic region, and posterior midgut were measured for 2 min. This number was divided by two to yield the number of gut contractions per minute.

The anterior midgut was defined as the region of the gut immediately following the proventriculus. The acidic midgut was defined as the narrow region of midgut following the wide foregut region and is characterized by a low pH marked by a color change from blue to yellow after ingestion of Bromophenol Blue. The posterior midgut was defined as the muscular area of the midgut immediately following the acidic portion and preceding the hindgut.

Third-instar larvae were removed from food plates, washed gently, then placed on a 100 mm×15 mm Petri dish with 5 ml yeast paste with 0.08%Fast Green FCF dye for 30 min. Larvae that had food in their gut were selected. Larvae were then placed on an agarose substrate for 3 h. The number of excretion spots were counted every 15 min and then added for a total number after 3 h. Total concentration of excretion was measured by soaking feces–agarose in 1 ml distilled water for 12 h then measuring concentration of dye at 625 nm in a spectrophotometer.

PKG enzyme assays were performed on 96 h post-hatch whole larval homogenates. Ten whole larvae were homogenized in 25 mmol l^–1Tris (pH 7.4), 1 mmol l^–1 EDTA, 2 mmol l^–1EGTA, 5 mmol l^–1 β-mercaptoethanol, 0.05% Triton X-100 and protease inhibitor cocktail (Roche Diagnostics, Laval, QC, Canada) and microcentrifuged for 5 min. The supernatant was removed and total protein levels were quantified. Supernatants containing equal amounts of total protein were analysed for cGMP-dependent protein kinase (PKG) activity. The reaction mixture contained (at final concentration): 40 mmol l^–1Tris-HCl (pH 7.4), 20 mmol l^–1 magnesium acetate, 0.2 mmol l^–1 [g³²P]ATP (500–1000 c.p.m. pmol^–1) (Amersham Pharmacia Biotech, Baie D'Urfe, QC,Canada), 113 mg ml^–1 heptapeptide (RKRSRAE), 3 mmol l^–1 cGMP (Promega, Burlington, ON, Canada) and a highly specific inhibitor of cAMP-dependent protein kinase (5-24, Calbiochem, San Diego, CA, USA). The reaction mixtures were incubated at 30°C for 10 min,followed by termination of the reaction by spotting 70 μl of the reaction mix onto Whatman P-81 filters, which were then soaked with 75 mmol l^–1 H₃PO₄ for 5 min and washed three times with 75 mmol l^–1 H₃PO₄ to remove any unreacted [γ³²P]ATP. Filters were rinsed with 100%ethanol and air dried before quantification. For quantification of PKG activity, counts were taken in a Wallac 1409 Liquid Scintillation Counter(Perkin Elmer, Woodbridge, ON, Canada) using universal scintillation cocktail(ICN). Specific activity of PKG was expressed as pmol of ³²P incorporated into the substrate min^–1 mg^–1protein.

Respiration rates were measured by indirect calorimetry(Gibbs et al., 2003). Groups of ten mid-third-instar larvae were placed in 5 ml glass–aluminum respirometry chambers containing a strip of medium, approximately 1 cm×2 cm× 0.2 cm. Larvae exhibited apparently normal behavior, eating and crawling in and around the food. The chambers were placed in a Sable Systems(Las Vegas, Nevada, USA) TR-2 respirometer, and CO₂-free air was pumped through the chamber at 100 ml min^–1. Placement of the larvae in the respirometer was staggered, so that they had been in place approximately 60 min before data were collected. Rates of CO₂release were measured over a 15 min period with a Li-Cor LI-6262 infrared CO₂ sensor. Data acquisition and analysis were done using Datacan V software (Sable Systems, Las Vegas, NV, USA). Control readings from chambers containing medium alone indicated no CO₂ release from microbial contaminants.

Larvae were removed from food plates, washed in distilled water and placed in groups of 70 in 9 cm×1 mm circular wells. Wells were filled with dead yeast paste (2:1 water:yeast, autoclaved 20 min) mixed with 0.08% Brilliant Blue R dye (Sigma) and 2 μCi ml^{–1 14}C-6-glucose (specific activity 58 μCi mmol l^–1, Amersham Biosciences) or 2μCi ml^–1l-[U-¹⁴C]leucine (specific activity 318 μCi mmol l^–1, Amersham Biosciences). After 15 min of feeding, larvae were removed and washed gently with a constant stream of 50 ml distilled water. Groups of ten larvae were then placed into 1.5 ml tubes, frozen in liquid nitrogen and stored at –80°C. For absorption experiments [protocol modified from Riha and Luckinbill(Riha and Luckinbill, 1996)]larvae were purged of the radio-labeled yeast paste prior to collection by placing them on unlabeled, undyed, heat-killed yeast paste for 3 h, the time it took for larvae to have no visible dye remaining in their guts (data not shown). We found no differences in the rate of passage of food through the rover or sitter guts (data not shown). Larvae were then carefully removed from the yeast paste, washed, placed in groups of ten into 1.5 ml tubes, and frozen in liquid nitrogen, as above.

Twenty-four hours later, larvae were removed from –80°C, placed in scintillation vials in groups of ten and solubilized at 70°C for 10 min in 200 μl Solvable (PerkinElmer Life Sciences, Woodbridge, ON, Canada) with 100 μl perchloric acid (Sigma). Next, 100 μl H₂O₂was added and samples were vortexed for 30 s. We then added 10 ml scintillation fluid, and samples were vortexed again for 30 s and shaken for 2 h, then left at room temperature for 24 h. Samples were then vortexed again,left for 2 h at room temperature, and the amount of ¹⁴C in each vial calculated using counts observed over 60 s per sample in a scintillation counter (Wallac 1409 Liquid Scintillation Counter). Sample sizes were ten larvae/vial with six vials/strain.

Prior to calculating the specific activity (fmol) of intake and absorption of ¹⁴C-6-glucose or l-[U-¹⁴C]leucine per larva, a conversion factor taking into account the specific activity of the radio-labeled substance was calculated. First the amount of specific activity(fmol) per 1 μl of radio-labeled substance was calculated by dividing the concentration supplied in the vial (mCi ml^–1) by the specific activity of the batch (mCi mmol l^–1). Next, the number of counts per minute (c.p.m.) in 1 μl of radioactive substance was divided by the number of fmol per 1 μl of radio-labeled substance to give the specific activity conversion factor. Each experimental c.p.m./larva was then divided by this conversion factor to reveal the specific activity of the sample(fmol).

Western blots were carried out as described(Belay et al., 2007) with the following changes. Protein was extracted from fly heads of 3–7 day-old flies reared at 23°C. 20 μg of protein extract was electrophoretically separated on SDS–10% polyacrylamide gels. To control for equal protein loading and transfer, blots were probed with a 1:5000 diluted monoclonal anti-β-actin antibody (Sigma).

The y-axis in all figures represents mean ± s.e.m. unless otherwise stated. SAS or JMP/IN 5.1 was used for all statistical analyses (SAS Institute Inc., Cary, NC, USA). Two-way and one-way analyses of variances(ANOVA) were performed followed by Student–Neuman–Keuls pairwise post-hoc comparisons using P<0.05 as significant. Non-parametric Kruskal–Wallis tests were performed followed by Wilcoxon two-group comparisons using P<0.05 as significant when sample size was smaller than six.

We first asked whether variation in for elicits differences in food intake and absorption in well-fed rovers and sitters. To determine if the effects we saw were specific to for, we also measured food intake and absorption in the sitter mutant (for^s2), which was generated on a rover genetic background(Pereira and Sokolowski,1993).

To measure food intake, well-fed third instar larvae were placed in food, a yeast paste colored with blue dye, and the amount of dye ingested was quantified. The amount of dye in the gut seen through the cuticle of each larva relative to the larval body area was used to estimate food intake(Fig. 1A). We quantified food intake in the natural rover (for^R) and sitter(for^s) allelic variants as well as the sitter mutant(for^s2) to establish the connection between for-PKG and food intake. Well-fed rovers (for^R)had significantly lower food intake than well-fed sitters after 10, 15 and 20 min on dyed yeast paste (Fig. 1B,C).

These data were confirmed using several assays including spectrophotometric quantification of dye (Fig. 1D), and visual quantification of dye in dissected guts(Fig. 2A). Quantification of the rate at which the gut filled with food calculated between 5 and 30 min of feeding on a yeast paste substrate demonstrated that sitter(for^s and for^s2) larvae fill their gut at a higher rate than rover (for^R) larvae(Fig. 2B). After approximately 30 min of feeding, the midguts of all larvae were filled with blue yeast paste, thus limiting further quantification of food intake. for^R larvae also showed decreased food intake compared to for^s and for^s2 larvae on glucose–agarose and fructose–agarose substrates, suggesting that the decreased food intake of for^R larvae is a general response to all food types (Fig. 2C).

Traditionally, mouthhook movements have been used to quantify food intake in larvae (Sewell et al.,1975; Wu et al.,2005). However, our observations of larval behavior suggested that larvae use their mouthhooks for both food ingestion and locomotion. We found no significant correlation between mouthhook movements and amount of dye ingested when larvae fed on a glucose–agarose substrate(Fig. 2D). In summary, our results clearly show that for affects the amount of food consumed in well-fed larvae, where variants of for associated with higher PKG activity confer lower food intake than variants of for associated with lower PKG activity.

Rovers were originally distinguished from sitters by their increased foraging locomotion (Sokolowski,1980). We confirmed that after 15 min exposure to food, rovers travel significantly farther on yeast than do sitters (P<0.0001)(Fig. 3A). This difference is expressed in the presence of a nutritive (yeast) but not a non-nutritive(agar) substrate (Fig. 3A). Thus, it was unclear how this increased locomotion could be sustained by lower food intake especially since, when food is plentiful, rover and sitter larvae develop at similar rates, reach similar sizes (length, width, mass) and have respiration rates that do not differ (Table 1, Fig. 4A).

Intriguingly, despite rovers showing increased locomotion on food, the similarity in metabolic rate between the natural variants suggests similar energy output. Notably, there was no significant behavioral difference observed in mouthhook extensions (Fig. 2D), which could be used for both locomotion and feeding. This suggests that active feeding in a smaller area by sitter larvae equalizes the total activity and therefore metabolic demand between the two natural variants. Thus for D. melanogaster larvae, reaching out and grabbing a mouthful of food seems to require as much energy as moving to a new location.

This was our first indication that even though rovers ingested less food than sitters, they do not suffer a fitness deficit relative to sitters. Thus,we hypothesized that rovers may absorb their food more efficiently than sitters.

We hypothesized that rovers may offset their reduced food intake with enhanced glucose absorption. We tested this using ¹⁴C-6-glucose-labeled food to compare ingestion and absorption(Riha and Luckinbill, 1996; Carvalho et al., 2005) (see Materials and methods). We measured the total amount of ¹⁴C ingested and the total amount of ¹⁴C absorbed.

As described above, for^R larvae have lower food intake than sitter larvae, as measured by total ¹⁴C uptake levels with ¹⁴C-6-glucose (Fig. 3B). Notably, chasing the radio-labeled medium with unlabeled food resulted in significantly higher ¹⁴C counts in for^R animals relative to for^s(Fig. 3B). Thus, for^R larvae absorbed approximately 50% of the ¹⁴C-6-glucose ingested whereas for^s and for^s2 larvae absorbed approximately 15% of the ¹⁴C-6-glucose ingested. Rovers retain more of the radioactive label after clearing their guts with unlabeled medium, indicating higher glucose absorption in rovers than in sitter and sitter mutant larvae.

To determine if rovers and sitters differed in amino acid absorption we repeated the above experiment with l-[U-¹⁴C]leucine(Fig. 3C). Leucine was chosen because it is one of the most abundant amino acids in yeast, a major food source of D. melanogaster larvae(Yamada and Sgarbiera, 2005). No absorption differences were found with l-[U-¹⁴C]leucine in well-fed larvae(Fig. 3C). As for glucose,rovers had lower intake of radiolabelled leucine than sitters. This suggested that the for-mediated differences in absorption in well-fed larvae may be specific to carbohydrates. It is not known how rovers absorb more glucose in their guts than sitters. Measures of excretion concentration, gross gut morphology and gut contraction rate did not show obvious differences between for^R, for^s or for^s2(Fig. 3D,E). Together, these results provide evidence that the rover/sitter natural polymorphism is a food-related phenomenon involving food intake and absorption.

Our results demonstrate that for affects the nutrient acquisition strategies of well-fed larvae, where rovers have decreased food intake and higher glucose absorption compared to sitters. We hypothesized that these differences in energy acquisition could affect larval responses to changes in food availability. Because the opportunity for larval food deprivation in nature is high (Atkinson, 1979)we investigated whether larvae exhibited plastic responses to food deprivation and if so whether this plasticity was mediated by for.

In order to determine appropriate food deprivation conditions we reared larvae from egg-hatch to pupation on media containing 100%, 75%, 50%, 25% or 15% of the yeast and sucrose content of standard medium (see Materials and methods). We hypothesized that the differences in food acquisition between variants of for affect larval development and survivorship. Larval development and survivorship of for^R, for^s and for^s2 did not differ at 100%,75% or 50%, suggesting that these food levels were not limiting. However, all strains were developmentally delayed when reared on 25% and 15%, and survivorship to eclosion also tended to be reduced(Fig. 4A). As a result, we used 25% and 15% as our food deprived conditions. Within-strain comparisons showed that all strains developed more slowly at 25% and 15%, and had significantly lower survival to eclosion at 15%. The magnitude of these changes was greatest in for^s and the for^s2 sitter mutant larvae. Between-strain survivorship analysis revealed that differences were not significant at 100%, 75%, 50% or 25% food levels, but at 15% rovers differed significantly from for^s and for^s2. Between-strain comparisons of developmental delay revealed that rovers differed from both for^s and for^s2 at 15% and rovers differed from for^s2 at 100%, 50% and 25% food levels, but not at 75%. These results suggest variation in for affects larval survivorship under low food conditions.

When larvae are reared to third instar under food deprivation conditions and then placed on a dyed yeast paste, for^R,for^s and for^s2 larvae exhibit plasticity in food intake, and this plasticity is mediated by for (see below). The food intake of all strains was significantly elevated when larvae were reared to mid-third instar on media containing 25% or 15% food compared to those reared on 100% food (Fig. 4B; two-way ANOVA for food level, F_(4,435)=18.94). Rovers (for^R)ingested less yeast paste than sitters when (for^s and for^s2) when raised on high levels of food (100%, 75%,50%), but not when reared on low levels of food (25%, 15%) whether they were feeding for 15 min (Fig. 4B) or 5 min (Fig. 4C). When for^R, for^s and for^s2 larvae reared in food-deprived conditions (25% and 15%) reached mid-third instar,their sizes did not significantly differ from each other, despite being developmentally delayed and smaller than larvae reared in abundant food (50%,75% and 100%) (Fig. 4D); for all food intake measures, larvae were staged to mid-third instar (see Materials and methods).

Glucose absorption differences persisted when larvae were reared under food-deprivation conditions (25% or 15%) demonstrating the persistence of increased glucose absorption by rovers compared to sitters and sitter mutants(Fig. 4E). As expected, in 15%food, for^R, for^s and for^s2larvae ingested similar amounts of ¹⁴C-labeled media in 15 min. However, for^R still absorbed more than for^s and for^s2 larvae. Since both rovers and sitters ingest food at the same rate when reared under poor food conditions, these results also suggest that rovers are able to absorb more nutrients per unit food ingested, and differences in absorption are not due to the amount of food in the gut.

Intriguingly, the absorption efficiency, defined as the percentage of ¹⁴C-absorbed/¹⁴C-ingested remained at about 50% for for^R larvae and 17% for for^s and for^s2 larvae, regardless of the quality or dilution of food in which the larvae are reared. Although it looks as if the amount of ¹⁴C absorbed increased from 100% to 15% food in for^R larvae, their total food ingestion also increases,resulting in an unchanged absorption efficiency.

We also found a nearly significant difference in absorption of leucine between rovers and sitters when reared under 15% food(Fig. 4F), suggesting that larvae may alter their leucine absorption in response to food deprivation but further experiments are required to assess this hypothesis. Overall, both rovers and sitters ingest food at the same rate after being reared under food deprivation conditions; however, rovers absorb more glucose than sitters. This may contribute to their enhanced fitness under 25% and 15% food deprivation conditions.

From the data above, we conclude that food ingestion rates of D. melanogaster larvae are sensitive to the nutritional state of the individual. Larvae do not eat at maximal rates when well fed but may approach ceiling ingestion rates when placed on yeast paste after having been raised in the 25% and 15% food conditions provided here. Furthermore, we find that rover food intake is not fixed at lower levels than sitters; rather, the differential varies depending on the degree of food deprivation. Since maximal rates of food intake do not appear to differ between the variants, the greater plasticity of rovers can be explained by their lower intake rates under well-fed conditions, which may be balanced by their higher glucose absorption under both well-fed and food-deprived conditions. These data may help explain why long-term selection at high larval density favors rovers(Sokolowski et al., 1997).

As the larval cells in which FOR are expressed are not yet known, we chose to manipulate for by ubiquitously increasing expression of for in sitters. To this end, we employed the transgenic sitter strain carrying a T2 transgene (UASforT2), which was expressed ubiquitously in a sitter genetic background using the leaky expression of a heat-shock promoter (hsGAL4). Such expression of forT2 in a sitter genetic background has been previously shown to result in rover-like foraging locomotion (Osborne et al.,1997) and increases the amount of protein produced by forT2 (Fig. 5A).

When sitters expressing forT2 were examined, rover-like food intake and glucose absorption patterns were observed(Fig. 5B–D). Consistently, expression of forT2 in sitters also significantly decreased food intake and increased the glucose absorption rate. Assessment of the survivorship to eclosion of hs-GAL4, UASforT2 larvae over a range of food levels demonstrated that variation in for also accounts for differences between rover and sitter survivorship and development rates(Fig. 6A). Expression of forT2 was sufficient to increase developmental rate and survivorship of food-deprived sitters raised on low nutrient media (25% or 15%). These data provide further proof that variation in for levels can account for the distinct rover and sitter food acquisition types.

Our goal in the present study was to test the hypothesis that forplays a role in food-related phenotypes. Follow-up investigations will address the cellular mechanisms through which PKG acts to mediate this plasticity. To further establish the connection between for-PKG and plasticity in food intake, we assayed PKG activity in rovers and sitters reared on 100%, 25%and 15% food levels. When food was not limited (100%), rovers ingest less and have higher PKG activity than sitters or for^s2 sitter mutants (Fig. 6B). In contrast,when for^R, for^s and for^s2larvae were grown under food deprivation conditions (25% or 15%), food intake levels increased to a similar level. Concomitantly, as food deprivation increased, PKG enzyme activities dropped to comparable levels, reducing rover–sitter differences in food intake(Fig. 6B,C). Rovers(for^R) had significantly higher PKG activity than sitters(for^s and for^s2) when reared on high food levels, but this difference diminished when larvae were food deprived. Thus, food intake and PKG enzyme activity are negatively correlated,suggesting that for is sensitive to changes in nutritional state. Furthermore, sitter larvae expressing a forT2 transgene had rover-like reduced food intake when food was plentiful, but not when reared in lower food levels (Fig. 6C). Together these data suggest that the regulation of for responds to food deprivation, and that changes in its expression can induce plasticity in food intake.

Interestingly, unlike food intake, the increased rate of glucose absorption in rovers did not vary in a food-dependent manner. Regardless of food level,higher levels of glucose absorption were found in rovers and sitters expressing the forT2 transgene compared with sitters and the sitter mutant. Elevated glucose absorption in rovers might partially account for their increased survivorship compared to sitters at 25% and 15%, despite similar PKG enzyme activity levels.

Our data do not address whether for acts throughout development or in an immediate nature to mediate food acquisition and plasticity in food intake. Thus, several possibilities could explain why glucose absorption does not vary with food level in a similar way as food intake and PKG activity. For example, food deprivation may not downregulate for activity in cell types important for glucose absorption, or the effects of PKG on glucose absorption may be fixed early in development. Future studies on specific expression of for and targeting manipulation of for in the cells in which for is expressed may elucidate both the developmental timescale and cellular mechanism through which for acts to mediate these responses.

The implications of this study are twofold. Firstly we demonstrate that natural variation in PKG affects food acquisition, and secondly we show for's role in plastic compensatory food intake responses to food deprivation. The rover/sitter system provides an enticing model for further investigations into the effects of `thrifty' genes that regulate efficiency of nutrient homeostasis (Zimmet and Thomas,2003; Shmulewitz et al.,2006). Alleles of such genes may promote individual fitness under restrictive food regimes (Neel,1962) just as for-PKG regulates normal individual differences in food intake. The role of for in food-specific behaviors is conserved in several species including honeybees, nematodes and harvester ants (Ben-Shahar et al.,2002; Fujiwara et al.,2002; Ingram et al.,2005). Future studies will determine whether similar gene-by-environment interactions prevail in the for orthologs identified in other species.

Intriguingly, a phylogeny constructed from available for-PKG protein sequences suggests that despite broad taxonomic breadth (nematodes to humans), there may be a widespread conserved association between PKG and food-related behaviors (Fitzpatrick and Sokolowski, 2004). If the role of for in the plasticity of food intake and regulation of food absorption is conserved in other animals including humans, then it may play an important role in regulation of energy homeostasis essential in maintaining a healthy body mass. The human homolog of for, cGK1, plays an important role in intestinal muscle function and thus may affect passage of intestinal content and ability to absorb food(Hofmann, 2005). Changes in PKG are also associated with disorders such as obesity and diabetes. For example, high expression of cGK1 has been associated with obesity,and a reduction of PKG has been associated with diabetes and high glucose concentrations (Wang et al.,2002; Wang et al.,2003; Wang et al.,2004; Engeli et al.,2004; Su et al.,2003; Zanetti et al.,2005; Chang et al.,2004). Although a polymorphism in cGKI has not yet been found between obese and healthy mass populations(Zakharkin et al., 2005), the evidence from D. melanogaster larvae suggests that cGKI may help sustain a healthy mass by maintaining a balance between energy input and output. Future research will help resolve whether a natural polymorphism in for also contributes to individual differences in energy balance in humans and across diverse taxa.

Little is known of the genetic or molecular basis for phenotypic plasticity, as genetic variability has intentionally been removed or reduced during the establishment of model organisms used in genetic analyses(DeWitt and Scheiner, 2004). Our findings provide a rare example of a naturally varying gene that affects plasticity providing a unique opportunity to investigate both the genetic and functional aspects of plasticity.

We would like to thank J. Levine and the members of the Sokolowski lab for their support and comments on the manuscript. We would also like to thank two anonymous reviewers for their very helpful comments in revising this manuscript. Funding was provided by NIDDK grants 5R01DK070141-03 to M.B.S. and NSF grant IOB-0514402 to A.G.G. The authors declare no competing financial interests.