• Open Access

DILP-producing median neurosecretory cells in the Drosophila brain mediate the response of lifespan to nutrition

Authors

Errata

This article is corrected by:

  1. Errata: Corrigendum Volume 9, Issue 5, 930, Article first published online: 16 September 2010

  • Present address: Y. Driege, VIB Department for Molecular Biomedical Research, Ghent University, UGent-VIB Research Building FSVM, Technologiepark 927, 9052 GENT, BELGIUM.

Linda Partridge, Institute of Healthy Ageing and GEE, University College London, Gower Sreet, London, WC1E 6BT, UK. Tel.: +44 20 7679 2983; fax: +44 20 7679 7096; e-mail: l.partridge@ucl.ac.uk

Summary

Dietary restriction extends lifespan in diverse organisms, but the gene regulatory mechanisms and tissues mediating the increased survival are still unclear. Studies in worms and flies have revealed a number of candidate mechanisms, including the target of rapamycin and insulin/IGF-like signalling (IIS) pathways and suggested a specific role for the nervous system in mediating the response. A pair of sensory neurons in Caenorhabditis elegans has been found to specifically mediate DR lifespan extension, but a neuronal focus in the Drosophila nervous system has not yet been identified. We have previously shown that reducing IIS via the partial ablation of median neurosecretory cells in the Drosophila adult brain, which produce three of the seven fly insulin-like peptides, extends lifespan. Here, we show that these cells are required to mediate the response of lifespan to full feeding in a yeast dilution DR regime and that they appear to do so by mechanisms that involve both altered IIS and other endocrine effects. We also present evidence of an interaction between these mNSCs, nutrition and sleep, further emphasising the functional homology between the DILP-producing neurosecretory cells in the Drosophila brain and the hypothalamus of mammals in their roles as integration sites of many inputs for the control of lifespan and behaviour.

Introduction

Lifespan extension by a moderate reduction in food intake without malnutrition (dietary restriction or DR) is evolutionarily conserved in many organisms including yeast (Jiang et al., 2000; Lin et al., 2000; Kaeberlein et al., 2005a), invertebrates (Klass, 1977; Lakowski & Hekimi, 1998; Houthoofd et al., 2003; Kaeberlein et al., 2006; Chippindale et al., 1993; Mair et al., 2003; Magwere et al., 2004) and mammals (Masoro, 2005; Merry, 2002). Dietary restriction has also been shown to delay or prevent many age-related declines in function and diseases (Wan et al., 2009; Mattson & Wan, 2005; Wang et al., 2005a). However, due in part to the use of a variety of different techniques to restrict dietary intake in each organism, it is still unclear how DR extends lifespan and improves health during aging, and there is little consensus as to whether the mechanisms involved in diverse species are universal or distinct. Dietary restriction has been proposed to extend lifespan by a variety of mechanisms in a range of model organisms. The protein deacetylase SIR2, the target of rapamycin nutrient-sensing pathway and the insulin/IGF-like signalling (IIS) pathway have all been implicated in yeast, worm, fly and mouse (Lin et al., 2000; Kaeberlein et al., 2005a; Rogina & Helfand, 2004; Kaeberlein et al., 2005b; Kapahi et al., 2004; Hansen et al., 2007; Clancy et al., 2002; Giannakou et al., 2008; Greer et al., 2007; Bjedov et al., 2010).

In addition to these genes and pathways, the tissues responding to changes in dietary intake to control longevity have been investigated, and recent evidence suggests that the central nervous system plays a crucial role in worms and flies (Bishop & Guarente, 2007b; Libert et al., 2007; Smith et al., 2008). In worms, DR-mediated lifespan extension required the activity of the transcription factor skn-1 acting in a pair of neurons in the head, the ASIs (Bishop & Guarente, 2007b) and has been shown to act via both decreased food consumption and food sensing (Smith et al., 2008). In flies, food odour was sufficient to reduce DR-induced lifespan extension, and pan-neuronal mutation of a chemoreceptor extended lifespan and disrupted the normal response of lifespan to DR (Libert et al., 2007), although the specific neurons mediating the response have not yet been identified. Such a noncell autonomous neuronal mechanism of DR-induced lifespan extension in worms and flies raises the possibility of a similar mechanism in mammals, and the hypothalamus has been suggested to be the crucial regulator in the mammalian brain (Bishop & Guarente, 2007a).

We have previously shown that partial ablation of insulin-like peptide-producing median neurosecretory cells (mNSCs) in the pars intercerebralis region of the adult Drosophila brain resulted in reduced levels of dilp2, 3 and 5 transcripts and extension of lifespan, as well as a reduction in fecundity and increased resistance to starvation and oxidative stress (Broughton et al., 2005). These cells are also the site of action of the stress responsive p53 and JNK signalling pathways for the determination of lifespan. Expression of a dominant negative form of p53 in the mNSCs was sufficient to extend lifespan, which correlated with lowered levels of dilp2 transcript in the mNSCs and lowered PI3K activity in the periphery (Bauer et al., 2007). Similarly, upregulation of JNK signalling specifically in the DILP-producing mNSCs extended lifespan, probably via the regulation of dilp expression (Wang et al., 2005b). Furthermore, the mNSCs respond to nutrients and short neuropeptide F, the fly orthologue of neuropeptide Y, to regulate dilp expression and IIS in peripheral tissues (Ikeya et al., 2002; Lee et al., 2008). Interestingly, gene expression patterns and the cell lineages giving rise to the mNSCs are very similar to those giving rise to the beta cells of mammals, and there are similarities in development between the mNSCs and the mammalian anterior pituitary and hypothalamic axis (Wang et al., 2007). The mNSCs, therefore, appear to be an integration site for many signals to control insulin-like ligand secretion and peripheral IIS in the fly similarly to the hypothalamus in mammals.

Here, we show that the DILP-producing mNSCs are required to mediate nearly all of the response of lifespan and some of the response of fecundity to yeast levels, thus identifying these neurosecretory cells as a key neuronal tissue for future research to resolve the role of IIS and other mechanisms that mediate responses to DR in Drosophila.

Results

DILP-producing median neurosecretory cells are required to mediate the response of lifespan to full feeding in a regime of DR by yeast dilution

To determine whether the DILP-producing mNSCs in the Drosophila brain are involved in the control of DR-mediated extension of lifespan and reduction of fecundity, we measured the responses of these two traits in flies lacking these cells to an optimised DR regime. DILP-producing mNSC-ablated flies were produced by dilp2GAL4-driven expression of UAS-reaper, as described previously (Broughton et al., 2005).

Dietary restriction in Drosophila is usually achieved by diluting the food medium, with lifespan increasing to a peak at an intermediate food concentration. At lower food levels, lifespan declines because of starvation, resulting in an overall tent-shaped response of lifespan to level of food intake. In Drosophila, the increase in lifespan in response to DR does not depend on calories consumed and instead has been shown to be because of the dilution of specific nutrients in the food, principally the yeast, with the sugar component playing only a minor role, and higher levels of sugar intake inhibiting fecundity (Mair et al., 2005; Bass et al., 2007). In this study, we therefore used a DR regime where we diluted only the yeast in the food medium whilst sugar was maintained at a constant concentration (Bass et al., 2007).

In two independent experiments, lifespan of control flies showed the typical tent-shaped response to food dilution. Lifespan peaked at 0.5 food dilution, or in one case 1.0 dilution, and then declined steadily at higher food concentrations, giving lifespan increases of between 12.5% and 19.5% in response to DR (Fig. 1A,C and Table 1 for statistical analysis, Appendix S1 and S2). In contrast, the decline in lifespan in response to increasing food concentration was almost completely disrupted by mNSC ablation, with no significant decrease in one replicate experiment, and a 5.3% decrease in a second replicate (Fig. 1A,C and Table 1 for statistical analysis, Appendix S1 and S2). The mNSCs are thus required for most of the decrease in lifespan in response to full feeding. Interestingly, the maximal lifespan of the ablated flies at the peak of the DR response was higher than that of the controls in both experiments. Together, these data suggest that mNSC ablation extends lifespan by mechanisms that are in part nonoverlapping with those that produce the response of lifespan to DR.

Figure 1.

 The effect of DR by yeast dilution on lifespan and fecundity of DILP-producing mNSC-ablated females and controls. (A and C) Median lifespans of d2GAL/UAS-rpr females and their controls (d2GAL/+ and UAS-rpr/+) across a yeast dilution series in two independent experiments. (A) Experiment 1. d2GAL/UAS-rpr on: 0.1 × Y N = 111; 0.5 × Y N = 74; 1.0 × Y N = 71; 1.5 × Y N = 72; 2.0 × Y N = 74. d2GAL/+ on: 0.1 × Y N = 108; 0.5 × Y N = 87; 1.0 × Y N = 76; 1.5 × Y N = 96; 2.0 × Y N = 89. UAS-rpr/+ on: 0.1 × Y N = 81; 0.5 × Y N = 61; 1.0 × Y N = 65; 1.5 × Y N = 66; 2.0 × Y N = 66. (C) Experiment 2. d2GAL/UAS-rpr on: 0.1 × Y N = 94; 0.5 × Y N = 105; 1.0 × Y N = 97; 1.5 × Y N = 93; 2.0 × Y N = 86. d2GAL/+ on: 0.1 × Y N = 96; 0.5 × Y N = 122; 1.0 × Y N = 112; 1.5 × Y N = 109; 2.0 × Y N = 111. UAS-rpr/+ on: 0.1 × Y N = 113; 0.5 × Y N = 102; 1.0 × Y N = 111; 1.5 × Y N = 80; 2.0 × Y N = 119. Survival curves for each genotype were compared between the yeast dilution giving the peak DR lifespan (0.5 or 1.0 × Y) and 2.0 × Yeast food using nonparametric log rank tests and P values calculated (median lifespans, the percentage DR-induced increase in lifespan above 2.0 × Y food and statistical analyses are given in Table 1). (B and D). Fecundity of d2GAL/UAS-rpr females and their controls (d2GAL/+ and UAS-rpr/+) from the survival experiments in (A) and (C): (B) fecundity in experiment 1; (D) fecundity in experiment 2. Data were analysed by anova with genotype and food dilution as the main effects. At each food dilution, planned comparisons of mean eggs laid by genotype were performed by Tukey HSD or Student’s t-test as appropriate, and * indicates significant difference to controls, P < 0.05. Mean number of eggs laid/female, the % increase in fecundity between the peak food for lifespan and 2.0 × Y food, and statistical analyses by food dilution are given in Table 2.

Table 1.   Median lifespans, % DR increase in median lifespan and statistical analysis of dilp2GAL/rpr, dilp2GAL/+ and UAS-rpr/+ female flies across a yeast dilution series in two independent replicate experiments. For each genotype, survival curves between the peak* food dilution and 2.0 × Y were compared using nonparametric log rank tests and P values calculated
Genotype/Replicate experiment Median Lifespan (Days)% increase in median lifespanP value, peak* vs. 2.0 × Y
0.1 × Y0.5 × Y1.0 × Y1.5 × Y2.0 × Y
  1. *Food dilution at which median lifespan peaked.

dilp2GAL/+1206764605619.50.003
2247469646219.4<0.0001
UAS-rpr/+1206063605612.50.0023
2227369666414.1<0.0001
dilp2GAL/121697777744.10.7851
UAS-rpr224807979765.30.0021

The fecundity of control females increased in response to increasing food concentration across the whole range, with significantly increased egg laying at each interval (Fig. 1B,D). mNSC ablation has previously been shown to reduce egg laying on standard food (Broughton et al., 2005), and in this study, it similarly resulted in decreased egg laying compared to controls at each food concentration in the DR range (Fig. 1B,D). Although the response of egg laying to DR was present, it was slightly but significantly reduced in the ablated flies (Table 2 for statistical analysis), suggesting that the mNSCs also play a role in mediating the effect of DR on fecundity.

Table 2.   Mean number of eggs laid/female on day 7 and statistical analysis of dilp2GAL/rpr, dilp2GAL/+ and UAS-rpr/+ flies from the replicate survival experiments 1 + 2 shown in Fig. 1 and Table 1. For each genotype individually, planned comparisons of mean number of eggs laid/female/day between peak* and 2.0 × Y food treatments were performed using Student’s t-test
Genotype/Replicate experiment Fecundity (eggs/female/day)% increase in fecundity, peak* vs. 2.0 × YP value, peak* vs. 2.0 × Y
0.1 × Y0.5 × Y1.0 × Y1.5 × Y2.0 × Y
  1. *Food dilution at which median lifespan peaked.

dilp2GAL/+11.6912.9715.1118.8524.071000.0031
25.445.958.0811.4414.55150<0.0001
UAS-rpr/+1NDNDNDNDNDNDND
24.786.258.5711.7915.11140<0.0001
dilp2GAL/10.686.8312.0414.4415.14250.0818
UAS-rpr23.855.946.739.6611.32900.0008

Previous work has demonstrated that with the DR regime we used, flies do not compensate for dilution of the yeast in the food medium by increasing their feeding rates on DR food (Mair et al., 2005; Wong et al., 2009). However, mNSC ablation could have reduced feeding behaviour, and thus affected the responses of lifespan and fecundity to DR or even have resulted in compensation in feeding rate. We therefore measured the feeding behaviour of 7-day-old, once-mated, ablated and control female flies by observation of proboscis extension during a 60-minute period of undisturbed feeding (Fig. 2A). As described in Wong et al. (2009), this observational analysis method was calibrated by direct quantification of blue dye food intake (Fig. 2B, Supplementary Fig. S1). In the full fed to DR range (0.5 × Y–2.0 × Y), no significant differences in feeding behaviour were observed between genotypes or between foods (Fig. 2A). Thus, the attenuated lifespan and fecundity responses of the ablated flies to the full-fed condition (2.0 × Yeast) were likely not because of a reduced feeding rate on this food compared to controls.

Figure 2.

 Feeding rates of mNSC-ablated and control flies on Sugar/Yeast food dilutions. Flies were reared at standard density on 1.0 × Y food. After a 48-h mating period on 1.0 × Y food, females of the indicated genotype were transferred to 0.1 × Y/0.5 × S, 0.1 × Y/0.1 × S, 0.5 × Y, 1.0 × Y or 2.0 × Y foods prior to analysis. (A) Undisturbed feeding behaviour: 6-day-old flies of each genotype were transferred to fresh food of the appropriate yeast dilution the evening before the assay. Feeding was measured during a 60-minute period the next morning by observation of proboscis extension behaviour. The proportion of flies feeding during this period is presented as a proportion of feeding events/possible feeding events ± SEM. (B) Calibration of feeding observations by direct quantification of food consumption: 7-day-old flies were transferred to blue dye containing fresh food of the appropriate yeast dilution. Observations of proboscis extension behaviour were performed over a 30-minute period, followed by quantification of food consumed by colour spectrophotometry. The blue dye food consumption data are presented as mean μg of food per mg of fly, ± SEM. The calibration analysis demonstrating that the proboscis extension observation method in (A) is an accurate indicator of food intake in female Drosophila under each food condition is shown in Supplementary Fig. S1. For both experiments, N = 10 vials of five flies per genotype, per food.

Another factor which could potentially contribute to differences between genotypes in the response of lifespan and fecundity to DR is mating rate. Mating increases egg laying and can shorten female lifespan (Fowler & Partridge, 1989). However, the cost of mating on lifespan has only been seen under repeated mating conditions throughout life and not under the early ‘once-mated’ conditions used in the present study (Chapman et al., 1995), and we directly compared survival of virgin and once-mated d2GAL/+ control females and found no difference in median or maximal lifespan (Supplementary Fig. S2). It is very unlikely therefore that the differences observed here in the response of lifespan to DR were because of differences in mating rates during the short 48-h mating period that females were subjected to prior to survival analysis. However, to further support this conclusion, we measured mating rates under these conditions and found no differences between genotypes (Supplementary Table S1). These data further suggest that the differential response of fecundity to DR of the ablated flies was similarly not because of differences in prior mating rate.

DILP5 levels correlate with the DR lifespan response

To begin to examine the mechanism by which the mNSCs respond to nutrient levels to regulate lifespan, we asked whether our DR regime affected transcript levels of the insulin-like peptides produced by these cells in control flies. The cell ablation because of d2GAL-driven expression of UAS-rpr is partial, resulting in a variable number of unablated mNSCs and low residual levels of dilp 2, 3 and 5 transcripts in adult fly heads (Broughton et al., 2005). We therefore also measured whether DR affected dilp levels in mNSC-ablated flies. Relative message abundance of dilps 2, 3 and 5 was measured in heads of flies maintained for 48 h on 0.1 ×, 0.5 × and 2.0 × Yeast media. There was no significant effect of yeast concentration on dilp2 and dilp3 transcript levels in control flies or on the residual transcripts of these dilps in the ablated flies (Fig. 3A,B). In contrast, dilp5 transcript levels increased significantly with increasing yeast concentration in control flies, consistent with previous studies showing that this dilp is nutrient responsive (Ikeya et al., 2002; Min et al., 2008). Residual dilp5 levels in ablated flies in the three foods were not significantly different (Fig. 3C). Thus, disruption of dilp5 nutrient responsiveness by mNSC ablation correlated with a disruption of the DR lifespan response.

Figure 3.

 The effect of DR by yeast dilution on transcription of dilps 2, 3 and 5 in adult heads of mNSC-ablated and control flies. Relative mRNA abundance of dilps 2, 3 and 5 from adult heads of flies of the indicated genotypes maintained on 0.1 × Yeast, 0.5 × Yeast, 2.0 × Yeast and 0.1 × Sugar/Yeast food dilutions was measured by quantitative RT-PCR and normalised to the abundance of actin5C: (A) relative abundance of dilp2 in each genotype; (B) relative abundance of dilp3; (C) relative abundance of dilp5. Data are shown as means of four independent experiments (N = 4) ± SEM. For each genotype, anovas were performed and expression levels of dilps 2, 3 or 5 were compared between the three food dilutions using Tukey HSD. *indicates significant difference to the 2.0 × Y expression level, P < 0.05.

To determine whether the observed changes in dilp5 transcript in control flies reflected changes in DILP5 protein levels, we next examined DILP5 protein under the same DR conditions in control flies. DILP5 could not be detected in Western blots, probably because of the low abundance of the protein, but we were able to detect it in the mNSCs by immunohistochemistry of whole mount brains. In d2GAL/+ flies after 48 h maintained on very low yeast food (0.1 × Yeast), DILP5 was virtually undetectable in 3/6 brains and was observed at very low levels in 2/6 brains (Figs 4 B,C and S3). However, after 48 h on 0.5 × Yeast and 2.0 × Yeast foods, it was detected in all brains examined (Figs 4B,C and S3). Using mean fluorescence levels in the mNSCs of each brain as a measure of DILP5 quantity, we determined that there was a significant effect of food on DILP5 protein (Fig. 4 C). Thus, in normal flies, DILP5 protein levels mirrored the observed changes in dilp5 transcript and lifespan under DR conditions.

Figure 4.

 The effect of DR by yeast dilution on DILP protein. (A) Western blot analysis of DILP2 levels in protein extract from female heads of wDah flies following 6 and 24 h of the indicated food treatment. (B-E) Immunohistochemical analysis of DILP5 and DILP3 protein in control d2GAL/+ 7 day old female brains following 48 h treatment with 0.1 × Y, 0.5 × Y and 2.0 × Y foods. Representative images of DILP expression from the analysis of brains examined at the same confocal microscope settings are shown. (B) (i) DILP5 protein was virtually undetectable following 0.1 × Y food treatment in 3/6 brains examined and was detectable at very low levels in 2/6 brains (see Supplementary Fig. S3 for images of all brains examined). (ii) DILP5 protein was detectable in 6/6 brains after 0.5 × Y treatment (see Supplementary Fig. S3). (iii) DILP5 protein was detected in 6/6 brains after 2.0 × Y treatment (see Supplementary Fig. S3). (C) Quantification of DILP5 levels using Image J performed on samples shown in (B) and in Supplementary Fig. S3. Data are shown as mean expression ± SEM, and *indicates significant difference to 2.0 × Y level by Tukey HSD, P < 0.05. (D) DILP3 was detectable at similar levels in all brains at all food treatments. (i) 0.1 × Y, DILP3 was detected in 5/5 brains. (ii) 0.5 × Y, DILP3 was detected in 3/3 brains. (iii) 2.0 × Y, DILP3 was detected in 3/3 brains. (E) Quantification of DILP3 levels using Image J. Data are shown as mean expression ± SEM.

dilp2, 3 and 5 transcript levels appear to be differentially regulated by nutrients and genetic manipulations (Broughton et al., 2005; Hwangbo et al., 2004; Min et al., 2008; Wang et al., 2005b; Bauer et al., 2007; Ikeya et al., 2002). Although dilp2 and 3 transcripts were not nutrient responsive in adults in this study, it was possible that their respective protein levels could be altered by DR. We therefore measured the effect of nutrients on DILP2 and 3 protein in control flies. Because of the great abundance of dilp2 transcript in the mNSCs (Broughton et al., 2005) DILP2 protein can be detected by Western blot (Broughton et al., 2008), but similarly to DILP5, DILP3 protein could only be measured in the mNSCs by immunohistochemistry of whole mount brains. Similarly to their transcripts, differences in levels of DILP2 and 3 protein in response to nutrient levels could not be detected (Fig. 4A,D,E).

mNSC ablation causes behavioural changes under low sugar conditions

We previously showed that the mNSC-ablated flies were resistant to complete starvation (Broughton et al., 2005, 2008), so it was surprising that these flies did not live longer than controls on the starvation side of the DR tent (at 0.1 × Yeast) (Fig. 1A,C). The DR food used in this study contains a constant sugar concentration of 50 g L−1 (Bass et al., 2007) across the range of yeast dilutions. We therefore tested the effect on lifespan of reducing the sugar concentration to 0.1 × (10 g L−1) in the low yeast food (0.1 × Yeast). In two independent experiments, the ablated flies were significantly longer lived than controls when maintained on this 0.1 × Yeast/0.1 × Sugar food (Fig. 5A,B; Appendix S3 and S4). Therefore, when yeast is at starvation levels (0.1 × or 10 g L−1), mNSC ablation extends lifespan when sugar is at 10 g L−1 but not when it is present in the media at 50 g L−1. The residual mNSCs in the ablated flies could have responded to the lower sugar concentration in the 0.1 × Yeast/0.1 × Sugar food to further reduce expression of dilp5 compared to that at 0.1 × Yeast/0.5 × Sugar. However, this is unlikely because transcript levels of dilps 2, 3 and 5 in each genotype were not significantly different between 0.1 × Yeast/0.1 × Sugar food and 0.1 × Yeast/0.5 × Sugar food. Although these data suggest that the lifespan extension of the d2GAL/UAS-rpr flies compared to controls under low sugar conditions was not because of further reductions in dilp expression, the already very low level of dilp expression in the ablated flies precluded any detection of a further reduction under low sugar conditions (Fig. 3).

Figure 5.

 Survival, activity and sleep behaviour of mNSC-ablated and control flies maintained on low sugar and yeast foods. (A and B) Median lifespans of d2GAL/UAS-rpr females and their controls on 0.1 × Yeast/0.1 × Sugar food in two independent experiments. Survival curves by genotype were compared using nonparametric log rank tests and P values calculated. Data are presented as median lifespan, and * indicates significant difference to both controls, P < 0.0001. (A) Median lifespans in Experiment 1: d2GAL/UAS-rpr = 22, N = 72; d2GAL/+ = 18, N = 78; UAS-rpr/+ = 18, N = 72. (B) Median lifespans in Experiment 2: d2GAL/UAS-rpr = 27, N = 116; d2GAL/+ = 22, N = 121; UAS-rpr/+ = 22, N = 101. (C) Mean total activity levels of d2GAL/UAS-rpr and control females on 0.1 × Yeast/0.5 × Sugar, 0.5 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.1 × Sugar foods. (D) Mean quantity of night sleep (minutes/12 h) ± SEM of d2GAL/UAS-rpr and control females on 0.1 × Yeast/0.5 × Sugar, 0.5 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.1 × Sugar foods. N = 16 for each genotype on each food. (E) Mean quantity of day sleep (minutes/12 h) ± SEM of d2GAL/UAS-rpr and control females on 0.1 × Yeast/0.5 × Sugar, 0.5 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.1 × Sugar foods. N = 16 for each genotype on each food. Data were analysed by anova and planned comparisons of means performed using Tukey HSD, and *indicates significant differences (P < 0.05) between indicated genotype/food treatments.

Although the ablated flies were found to display similar feeding behaviour to controls on foods in the DR range (Fig. 2), it was possible that they would respond differently at starvation level foods. To determine whether feeding behaviour was affected by mNSC ablation, we measured observed feeding rate and food consumed of ablated and control flies on 0.1 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.5 × Sugar foods. In agreement with other studies (Skorupa et al., 2008), the very low nutrient content (most likely the low yeast concentration) in both foods resulted in increased feeding compared to flies feeding on foods in the fully fed to DR range in all genotypes (Figs 2B and S1). This compensatory increase in feeding on 0.1 × Yeast/0.1 × Sugar food was attenuated by mNSC ablation (Fig. 2B) suggesting that the extended lifespan of these flies on this food was not because of a compensatory increase in feeding. These data further suggest that the mNSCs are involved in the modulation of feeding behaviour in response to low nutrient levels.

The dilp-producing mNSCs have been suggested to be involved in controlling both nutrient-independent basal locomotor activity and starvation-induced hyperlocomotion (Mattaliano et al., 2007), raising an alternative possibility that the resistance of the ablated flies to the low yeast and sugar food could have a behavioural basis dependent on the mNSCs but independent of IIS. Normal flies respond to complete starvation by increasing their activity levels and this hyperlocomotory response requires the expression of dARC in the DILP-producing mNSCs (Mattaliano et al., 2007). Flies mutant for dARC do not undergo starvation induced hyperlocomotion and are long-lived, presumably because they do not use up stored resources as quickly as controls. To determine whether our 0.1 × Yeast/0.1 × Sugar food elicited a hyperlocomotory response, we measured total locomotor activity of flies maintained on 0.1 × Yeast/0.1 × Sugar (equivalent to 1% sugar, 1% yeast) compared to the same flies on 0.1 × Yeast/0.5 × Sugar food (equivalent to 5% sugar, 1% yeast) and 0.5 × Yeast/0.1 × Sugar food (equivalent to 1% sugar, 5% yeast). We found that control flies did not differ significantly in total activity on the three foods (Fig. 5C) showing that the reduction in sugar and yeast concentrations in our 0.1 × Yeast/0.1 × Sugar food was not sufficient to elicit the behavioural hyperlocomotory starvation response. The mNSC-ablated flies actually showed an increase in total activity on the 0.1 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.5 × Sugar foods compared to the 0.5 × Yeast/0.1 × Sugar food (P < 0.0001) further supporting the conclusion that a lack of a hyperlocomotory starvation response was not the basis of the extended lifespan of the mNSC-ablated flies on the 0.1 × Yeast/0.1 × Sugar food. Although the activity phenotype was very variable in control genotypes, these data showing a trend towards a yeast dependent modulation of activity in the ablated flies raise the possibility that under low sugar conditions the DILP-producing mNSCs may normally act to suppress the negative effect of yeast on activity.

In addition to their starvation-induced hyperlocomotion phenotype, dARC mutant flies display altered light/dark basal activity levels resulting from a change in the distribution of sleep between night and day (Mattaliano et al., 2007). We therefore next examined the effect of reducing sugar concentration in the presence of high and low levels of yeast on the amount of day and night sleep in control and ablated flies (Fig. 5 D,E). Day sleep in all genotypes responded similarly to the three foods, showing an increased amount of sleep on 0.5 × Yeast/0.1 × Sugar food compared to 0.1 × Yeast/0.1 × Sugar and 0.1 × Yeast/0.5 × Sugar foods (P < 0.05). This suggests that flies respond to yeast levels to control day sleep and that the DILP-producing mNSCs are not required for this response (Fig. 5D). In contrast, the quantity of night sleep in control flies did not differ between foods suggesting that night sleep is not nutrient sensitive (Fig. 5E). However, the mNSC-ablated flies showed a significant decrease in night sleep only on 0.1 × Yeast/0.1 × Sugar food (P < 0.05). These data suggest that day and night sleep are differentially regulated by nutrient levels and that during the night the mNSCs may normally act to inhibit the negative effect of low nutrient levels on sleep. In terms of explaining the extended lifespan of ablated flies on 0.1 × Yeast/0.1 × Sugar food, it is possible that increased wakefulness during the night could allow more feeding opportunities. However, this is unlikely because these flies actually showed decreased consumption of this food in the day time feeding assay (Fig. 2).

Thus, the mNSC-ablated flies on 0.1 × Yeast/0.1 × Sugar food are more active, display less night time sleep and eat less than control flies. Although the basis for their longevity on this food remains to be determined, taken together, the data presented here suggest that the mNSCs are involved in the control of the response of lifespan, feeding and night time sleep to low nutrient intake.

Discussion

The DILP-producing mNSCs in adult flies are required for most of the response of lifespan to DR by yeast dilution, indicating that these neurosecretory cells act noncell autonomously to regulate lifespan in response to changes in nutrition. A similar neuroendocrine mechanism has been indentified in worms (Bishop & Guarente, 2007b) and suggested in the mouse (see [Bishop & Guarente, 2007a] for review). The data presented here support the evolutionary conservation of the neuroendocrine control of DR-induced longevity and provide further evidence that the mNSCs act as an integration centre for many signals in a similar way to the hypothalamus of mammals.

mNSC ablation extends lifespan significantly at 1.0 × Y food (the DR condition) but lifespan does not decline with increasing yeast levels (the full fed 2.0 × Y condition). We interpret these data to suggest that the mNSCs are required to mediate the lifespan shortening effect of high yeast levels. However, it should be noted that an alternative interpretation of the data is that the mNSC-ablated flies are already at a physiological maximum for lifespan at 2.0 × Y food that cannot be increased by DR. Such an interpretation assumes that under the environmental conditions used, flies are incapable of living longer with any manipulation. We know that lowered temperature can increase fly lifespan to much greater values than has ever been seen with DR or reduced IIS (Mair et al., 2003). Furthermore, larger reductions in IIS because of ubiquitous expression of a dominant negative insulin receptor can result in greater extensions of lifespan than mNSC ablation at 1.0 × Y food in the same wDah genetic background (Ikeya et al., 2009). This strongly suggests that the level of IIS down-regulation in the mNSC-ablated flies is not at a physiological maximum for lifespan at 1.0 × Y food and that there is no ceiling or floor effect in the data.

The response of DILP5 transcript and protein in the mNSCs to DR by yeast dilution in control flies correlated with DR-induced lifespan extension, and mNSC ablation disrupted both the dilp5 nutrient responsiveness and the DR lifespan response. These data raise the possibility that lifespan extension by DR is normally mediated at least in part via the regulation of DILP5 in the mNSCs. Such a role is consistent with the higher peak lifespan of the mNSC-ablated flies under DR conditions, because the ablation reduces total dilp levels to a greater degree than does DR. It should be noted that the mNSC ablation in the d2GAL/UAS-rpr genotype is partial (Broughton et al., 2005), so it is likely that the few residual mNSCs can respond to DR by altering DILP5 protein levels, although to a much lesser extent than in control flies. This would be consistent with the greatly reduced but not completely absent DR lifespan response observed here in the ablated flies. However, this could not be confirmed because we were unable to measure changes in DILP protein levels in the ablated flies because of the variability in the number of cells ablated between individual flies, the low abundance of the protein and the low sensitivity of the immunohistochemistry quantification method.

Min et al. (2008) recently reported that the effect on lifespan of DR was independent of IIS. This conclusion was in part based on the finding that although a reduction in dilp5 transcript was associated with DR lifespan extension, blocking the nutrient responsiveness of this dilp with RNAi did not block the DR lifespan response. However, the dilpRNAi construct used in the Min et al. (2008) study reduced levels of all three dilps to 50% of control levels, and the effect of DR on DILP5 protein in these flies was not reported. Given the nutrient responsiveness of DILP5 protein presented here and the high level of dilp5 transcript remaining in the dilpRNAi flies, DILP5 protein levels may in fact change in the Min et al. flies to mediate the observed DR lifespan response. Experimental testing of a role of dilp5 in the response of lifespan to DR will require the complete and specific knock-down of dilp5 by mutation or RNAi. However, the potential for compensatory regulation of dilp expression (Broughton et al., 2008) may complicate the interpretation of such experiments.

Despite the correlation between DILP5 and lifespan in response to DR, it is of course equally possible that the mNSCs respond to DR by controlling an as yet unknown, IIS-independent endocrine factor to which the IIS-dependent lifespan extension of the mNSC ablation is additive. However, other studies suggesting a modulatory role of IIS in extension of lifespan by DR (Clancy et al., 2002; Giannakou et al., 2008) raise the possibility that DILP5 may at least play a partial role. Flies mutant for the insulin receptor substrate chico show a right-shifted DR response (Clancy et al., 2002). This has been suggested to indicate that chico modulates the DR response. More recently, Giannakou et al. (2008) showed that although the key downstream effecter of IIS in flies, dFOXO, was not required for a DR lifespan response, its over-expression right-shifted the response, again suggesting that IIS normally plays a role in DR-mediated lifespan extension. The data presented here support a model whereby the DILP-producing mNSCs are central to the mediation of DR lifespan extension by both IIS-dependent (via the regulation of DILP5) and independent (via any unknown endocrine factor) mechanisms.

A very recent article by Ja et al. (2009) suggests that increasing food concentration causes reduced feeding and therefore reduced water intake, such that providing access to water can rescue the DR lifespan extension. This is certainly an issue that needs to be addressed in all DR studies in Drosophila. The DR regime used here was that optimised by Bass et al. (2007), which determined that the detrimental effect of high food on lifespan was not because of reduced water availability. The addition of a fresh water source did not rescue the effect of this high nutrient food on lifespan or effect fecundity (Bass et al., 2007). Ja et al. (2009) suggest that the contradictory results in Bass et al. (2007) and Ja et al. (2009) may be explained by differences in water surface area in the two studies. However, a recent study supports the sufficiency of the water supplementation method used in Bass et al. (2007). Grandison et al. (2009) found that although this water supplementation method again did not rescue the detrimental effect of specific nutrients on lifespan, it was sufficient to rescue the toxic effect of high salt levels in food. Together, these data strongly suggest that the lifespan shortening effect of the high nutrient food in control flies in our study was not because of an inability to access sufficient water.

On the starvation side of the lifespan/food dilution relationship, the mNSC-ablated flies displayed the same median lifespan as controls when on 0.1 × Yeast/0.5 × Sugar food, which was surprising given the resistance of these flies to complete starvation (Broughton et al., 2005, 2008). This prompted us to examine the effect on survival of reducing both yeast and sugar to the 0.1 × level. The enhanced survival of the ablated flies compared to controls on this 0.1 × Yeast/0.1 × Sugar food was not because of further reductions in dilp levels or a reduction in total activity. One explanation could be that low nutrient levels elicit the mobilisation of stored energy and the increased trehalose and glycogen stores in the ablated flies (Broughton et al., 2005) contributed to their enhanced survival. However, our finding that the ablated flies displayed a change in night sleep behaviour on 0.1 × Yeast/0.1 × Sugar food suggests that a behavioural change may play a role in their enhanced survival on this food. Moreover, whether or not the increased night wakefulness resulted in increased feeding opportunities and subsequently enhanced survival under low nutrient conditions, these data indicate that the mNSCs are involved in more than simply producing insulin-like peptides. Interestingly, activation of EGFR and ERK by rhomboid signalling in a subset of neurosecretory cells in the pars intercerebralis region of the brain has been shown to regulate night sleep (Foltenyi et al., 2007). It is possible that these cells are within the DILP-producing subset of cells in the pars intercerebralis or interact with them. In mammals, the hypothalamus links and coordinates the regulation of sleep and nutrient intake (Alvarenga et al., 2005; Vanitallie, 2006), and long-term food restriction has been found to alter sleep patterns in rats (Alvarenga et al., 2005) and food restricting anorexic humans (Lauer & Krieg, 2004) for review). The data presented here indicating an interaction between sleep, nutrition, feeding and the mNSCs in Drosophila further emphasises the functional homology between these DILP-producing cells and the hypothalamus of mammals as an integration site of many inputs for the control of lifespan and behaviour.

This study identifies the DILP-producing mNSCs as a key neuronal tissue required to mediate the response of lifespan to DR in Drosophila. Much remains to be determined about, for instance, the dietary components to which the mNSCs respond, the putative sensory inputs to these cells, the signalling pathways required for the DR response and the endocrine factor(s) released by them. These DILP-producing median neurosecretory cells will thus be an important focus of future studies to elucidate the IIS-dependent and independent molecular and neuronal mechanisms mediating DR lifespan extension.

Experimental procedures

Fly stocks and maintenance

The control whiteDahomey stock was derived by backcrossing w1118 into the outbred wild-type Dahomey background. UAS-reaper and dilp2-GAL were backcrossed into wDah. Stocks were maintained and experiments conducted at 25 °C on a 12 h : 12 h light/dark cycle at constant humidity. Flies for lifespan analysis were raised on standard sugar/yeast medium (Bass et al., 2007) before transfer to the appropriate DR food, as described in the following text.

Lifespan

Procedures for lifespan studies were as described in (Clancy et al., 2001) and (Mair et al., 2003). Lifespan was measured in once-mated female flies kept at 10/vial on dietary restriction food medium (Bass et al., 2007) and transferred to new food three times a week. Deaths were scored 5–6 times in every 7 days.

DR regime (Food dilution)

Dietary restriction experiments were performed across a yeast dilution series (xY), using an optimised regime with sugar at a constant 50 g L−1 concentration as described in (Bass et al., 2007).

Feeding assays

Feeding rate in 7-day-old, once-mated females was measured using a behavioural observation method (proboscis extension onto solid food) calibrated by direct quantification of food consumption (brilliant blue dye quantification), as described in Wong et al. (2009).

Quantitative RT-PCR

dilp transcript levels were measured as in (Broughton et al., 2005), except SYBR Green master mix (Applied Biosystems, Warrington, UK) was used and four independent head RNA extractions performed per genotype.

Antibody generation

Polyclonal antibodies against DILP3 were raised in rabbits against peptides CLRYCAAKPRT and GRKLPETLSKLC. Polyclonal antibodies against DILP5 were raised in rats against peptides DFRGVVDSCCRKS and DMLRVACPNGFNSMFA. Immunisation package AS-DOUB-LX was used (Eurogentec, Liege, Belgium). The peptides were coupled to keyhole limpet haemocyanin and both peptides were injected together into each of two rabbits or rats.

Western blots

Western blot analysis of DILP2 was performed as described in (Broughton et al., 2008) on extracts of control fly heads of the indicated feeding regime.

Immunohistochemistry and confocal microscopy

Immunohistochemical analysis of DILP3 and DILP5 protein in whole mount brains of 7-day-old females following 48-h feeding on the indicated DR food was performed as described in (Lee et al., 2000). Anti-DILP3 primary antibody was used at a dilution of 1 : 20 followed by a Fluorophor 488 labelled anti-rabbit secondary antibody (Molecular Probes) at 1 : 500 dilution. Anti-DILP5 primary antibody was used at a dilution of 1 : 50 followed by a Fluorophor 488 labelled anti-rat secondary antibody (Molecular Probes) at 1 : 500 dilution. Confocal imaging of Fluorophor 488 fluorescence was carried out on a Zeiss LSM 510 microscope using the same settings for each sample. Confocal image stacks were converted to projections using Zeiss LSM 510 software. Relative quantifications of DILP3 and 5 levels in the mNSCs were performed in Image J (NIH, Bethesda, MD, USA) by taking mean fluorescence over a defined area encompassing the mNSC cluster in each brain image examined.

Fecundity

Fecundity of females in lifespan experiments was measured as described in (Broughton et al., 2005). Data are reported as the mean number of eggs laid per day per female ± SEM over a 2-day period.

Locomotor activity and sleep data

Individual 7-day-old female flies were placed in 65 mm × 5 mm glass tubes (Trikinetics, Waltham, MA) containing either 0.1 × Yeast/0.5 × Sugar, 0.5 × Yeast/0.5 × Sugar or 0.1 × Yeast/0.1 × Sugar food. Flies were entrained for 2 days at 25 °C on a 12 : 12 light/dark cycle before the collection of 4 days of locomotor activity in 5-min bins using DAM System monitors (Trikinetics, Waltham, MA). Sleep was defined as 5 min of inactivity, as described in (Shaw et al., 2000), and total day and night sleep was calculated using BeFLY! Analysis Tools v7.23 (Ed Green) in Excel.

Statistical analyses

Statistical analyses were performed using JMP (version 7) software (SAS Institute Inc., Cary, NC, USA). Lifespan data were subjected to survival analysis (Log Rank tests) and presented as median lifespan. Mating rates were compared by chi-square analysis. Other data (fecundity, QPCR, activity, sleep and DILP levels) were tested for normality using the Shapiro-Wilk W test on studentised residuals (Sokal & Rohlf, 1998) and found to be normally distributed. One-way analyses of variance (anova) were performed and planned comparisons of means were made using Tukey–Kramer HSD or Student’s t-test.

Acknowledgments

We thank the Wellcome Trust and the BBSRC for financial support. We are grateful to Bambos Kyriacou and Ed Green for the BeFLY! Analysis software, and Matt Piper for critically reading the manuscript.

Ancillary