Conditional gene expression systems such as promoters regulated by tetracycline or doxycycline (DOX) have several advantages for testing the effects of genes on aging and lifespan [1,2,3]. In the 'tet-on' system, feeding DOX to Drosophila melanogaster causes high levels of transgene expression in all tissues. By waiting until the young adult stage to administer DOX, all of pre-adult development is identical between control and experimental groups, and any difference in lifespan must be due to changes in the adult. Subtle differences in the genetic background of Drosophila strains can have significant effects on lifespan [4,5,6]. With the tet-on system, control and experimental animals have identical genetic backgrounds, and therefore any differences in lifespan must be due to DOX administration and transgene expression. DOX itself, and overexpression of control genes such as Escherichia coli lacZ, have no detectable effects on lifespan [1]. In contrast, several genes have been identified for which DOX-regulated overexpression has negative effects on lifespan, and overexpression of the dGMII gene, encoding α-mannosidase II, was associated with slightly increased lifespan [7].

Expression of antisense RNA has long been known to be able to inhibit gene expression in Drosophila and other organisms [8,9,10,11,12]. Double-strand RNA (dsRNA) formed by hybridization of sense and antisense sequences is thought to initiate a pathway in which homologous RNA sequences are destroyed. This phenomenon has been referred to as post transcriptional gene silencing (PTGS) or RNA interference (RNAi). Expression of inverted-repeat constructs, where the transcript is expected to fold into a dsRNA hairpin, has been shown to be an efficient initiator of RNAi [13,14,15]. In experiments reported here, the tet-on system was used to drive expression of inverted-repeat constructs in Drosophila to determine if a conditional system for RNAi could be created.

In natural populations, lifespan and reproductive period are thought to co-evolve [16]. The duration of the reproductive period is subject to natural selection, and lifespan is altered as a consequence. Significant experimental support for this model has come from laboratory selection experiments with D. melanogaster [17,18,19,20,21]. Genetically heterogeneous Drosophila populations have been selected over hundreds of generations for late-life reproduction. Such selection results in strains with increased lifespan relative to controls. Selection is thought to function by altering the frequency of gene alleles present in the starting population. Selected strains exhibit a number of correlated phenotypes. The correlated phenotypes can vary somewhat depending on the starting strains used; however, increased stress-resistance appears to be common to all or most long-lived strains [22,23,24,25,26]. In one well studied set of five replicate control (B) and long-lived (O) strains, lifespan was doubled and was correlated with increased stress-resistance and increased glycogen and lipid stores [27,28].

Starch-gel electrophoresis has been used to assay for changes in the frequency of enzyme electrophoretic alleles in the O and B lines. A change in the frequency of Cu/Zn-superoxide dismutase (Cu/ZnSOD) alleles was observed, with the more active allele enriched in the long-lived O strains [29]. Consistent with this observation, overexpression of Cu/ZnSOD in transgenic Drosophila has been shown to be sufficient to cause significant increases in lifespan [30,31]. The most dramatic change in allele frequency in the O and B lines was observed for the gene for phosphogluconate mutase (Pgm) [32]. Pgm exists in three electrophoretic forms: fast (Pgm^F), medium (Pgm^M) and slow (Pgm^S). Pgm^M predominates in the B strains, whereas Pgm^Sallele frequency is on average ten-fold higher in the long-lived O strains relative to the B controls. When O strains were taken off selection for several generations (back-selection), lifespan decreased and Pgm^S allele frequency was reduced to levels more like those in B strains. These results suggest that Pgm or a closely linked gene is responding to selection, and make Pgm^S a candidate lifespan regulator.

Pgm encodes the enzyme phosphoglucomutase, which inter-converts glucose 1-phosphate and glucose 6-phosphate. Its activity is therefore important for both glycolysis and glycogen synthesis, and the altered Pgm allele frequency might therefore be relevant to the increased glycogen and lipid stores of the O strains. If the increased Pgm^S allele frequency is contributing to the unique phenotypes of the long-lived O strains, this might be because Pgm^S has increased or decreased enzyme activity, or enzyme activity has been altered in some way (perhaps by a change in its regulation). One way to begin to test these models experimentally is to engineer transgenic flies with increased or decreased Pgm expression and assay for effects on lifespan.

Multiple strains were generated that were homozygous for each of the F, M and S electrophoretic alleles of Pgm, by appropriate crosses to a third chromosome balancer stock. Pgm enzyme activity and lifespan varied greatly across the strains with no correlation with Pgm allele (data not shown). This result was expected, as differences in genetic background between such purified chromosome strains has profound effects on lifespan and the activities of various enzymes [5,6,33]. The Pgm coding-region sequences were cloned and sequenced from the Pgm^F, Pgm^M and Pgm^S homozygous strains and from the Canton-S wild-type strain. Amino-acid substitutions were identified that predicted pI values for the alleles that correlated with their mobility on starch gels (Figure 1). The substitution of T for A in Pgm^S creates the amino-acid sequence TTK (in the single-letter amino-acid code) which is a potential phosphorylation site for protein kinase C (PKC) that is absent in Pgm^M and Pgm^F.

If Pgm^S contributes to the increased lifespan of the O lines, this might be due to increased enzyme activity, decreased enzyme activity, or alteration in enzyme activity in some other way such as in its regulation or subcellular localization. The tet-on system was used to test whether simply increasing or decreasing Pgm expression would be sufficient to alter lifespan. A P-element transformation vector called USC1.0 was generated, which had unique EcoRI and PstI sites downstream of the tet-on doxycycline-regulated promoter (Figure 2a). USC1.0 also contains the mini-white⁺ transformation marker gene. The Pgm^S cDNA sequences were cloned into USC1.0 to create the PGMsense construct (Figure 2b), which should allow overexpression of the PGM^S enzyme. A fragment of Pgm^S coding region of approximately 1 kilobase (kb) (from +711 to +1,794) was cloned into USC1.0 in an inverted-repeat orientation to create construct PGMinvrpt (Figure 2c), which should potentially cause conditional RNAi. For each PGM construct five independent transgenic lines were generated using standard methods. As a control, two trangenic lines were generated containing the USC1.0 vector in which no transgene is expressed.

Expression of the transgenic constructs and the endogenous Pgm gene was assayed by northern blot, with or without DOX (Figure 3a). The Pgm inverted-repeat sequences were used as probes and should hybridize to both transgenes and the endogenous Pgm transcript. In the sense-construct lines, feeding DOX resulted in efficient expression of the transgene at levels five- to ten-fold greater than the endogenous transcript. DOX-induced expression of the inverted-repeat construct had two effects detected by northern blot. First, DOX produced a smear of hybridization resulting from the inverted-repeat transcript (Figure 3a). The smear pattern may have resulted from incomplete denaturation of the hairpin structure and/or instability of the inverted-repeat transcript. Second, the level of the endogenous Pgm transcript was found to be reduced by DOX feeding.

The northern blots were also hybridized with a Pgm-specific probe corresponding to Pgm coding-region sequences (+1 to +700) located outside the region used to create the inverted repeat. This probe is therefore specific for the endogenous Pgm transcript. Using this probe the smear of hybridization derived from the inverted-repeat construct was no longer detected, and the endogenous Pgm RNA levels were more readily observed and quantitated (Figure 3b). DOX induced expression of the inverted-repeat construct was found to cause a reduction of endogenous Pgm RNA levels. The decrease varied across the five transgenic lines, with reductions ranging from 1.3- to 24-fold. The inverted-repeat construct therefore appears to function as expected to induce RNAi and cause decreased expression of the endogenous Pgm gene. In the control strains where no transgene is expressed, DOX administration was found to have no effect on Pgm transcript levels (Figure 3c).

The indirect PGM activity assay was used to assay for changes in Pgm gene expression at the protein level. The sense construct yielded conditional overexpression of PGM activity in each transgenic line, with increases ranging from two- to five-fold (Table 1). Potential reductions in PGM activity were more difficult to assay because of the inherent limitations of the assay. The PGM activity assay is indirect and is coupled to the conversion of NADP⁺ to NADPH. The extract will contain other activities capable of converting NADP⁺ to NADPH, thereby creating a significant background activity. Decreases in PGM activity could not be reliably measured because of this unavoidable background. The PGMinvrpt construct yielded decreases in PGM activity of as much as 50%; however, the effect was quite variable from experiment to experiment (Table 1). In the control strains where no transgene is expressed, DOX administration was found to have no consistent effect on PGM levels (Table 1).

To determine whether altered Pgm gene expression could affect lifespan, mean lifespan was assayed in multiple PGMsense and PGMinvrpt transgenic lines with or without DOX treatment. The percentage change in mean lifespan caused by DOX and transgene expression is presented for each line. Lifespan was observed to vary across transgenic lines and replicate experiments with changes ranging from -7% to +6%, with one outlier at +16% (Table 2). However, no change in lifespan was observed that was consistent across the sense or inverted-repeat lines or in multiple experiments. The lifespan assays were repeated at 29°C, and again no consistent changes were observed (Table 3). We conclude that under these conditions, altered Pgm expression does not have a detectable effect on adult Drosophila lifespan.

The phenomenon of RNAi (or PTGS) is of great interest for two reasons [9,10,11,12]. First, it represents an evolutionarily conserved pathway that probably has important functions in the regulation of gene expression and the control of transposable elements. Second, it provides a means for researchers to test gene functions by experimentally downregulating expression of specific genes under well controlled conditions. In Drosophila, RNAi can be initiated by injection of dsRNA into embryos, and this has allowed identification of novel phenotypes for genes during development [34]. However, injection has limited potential application in the adult, as it is unlikely that the RNAi could be targeted to all tissues, or to specific tissues, and the trauma of injection is likely to have negative effects on phenotypes such as lifespan. Transgenic constructs have been used to cause RNAi in Drosophila, by using various promoters to drive expression of inverted-repeat fragments of gene coding regions [14,15]. Using the conditional tet-on promoter system to drive expression of an inverted-repeat provided conditional RNAi of the Pgm gene. This conditional system should have many applications, both in the study of the RNAi pathway itself and in the design of experiments where RNAi is used a research tool. Conditional RNAi should be particularly useful to study phenotypes in the adult such as aging and lifespan. Conditional gene-expression systems have previously been used to identify positive regulators of lifespan, such as Cu/ZnSOD and dGMII [7,31]. In addition, mutations have been used to identify several genes in Drosophila that act as negative regulators of lifespan, such as mth, INDY, InR and chico [35,36,37,38]. For mth and InR the null phenotype of the genes is lethality. Therefore it is only rare hypomorphic alleles or allele combinations that allow the animals to survive to adult stage where increased lifespan can be observed. Conditional RNAi should provide a powerful means to study these genes and other negative regulators of aging and lifespan. Inactivation or downregulation of the genes specifically in the adult using tet-on RNAi may allow increased lifespan without confounding effects on development. Lam and Thummel have recently reported the use of a heat-shock gene promoter to drive expression of dsRNA, and the efficient conditional inhibition of gene expression during larval and pupal stages of Drosophila development [39].

Genetic selection of Drosophila populations for late-life reproduction caused the correlated phenotype of increased lifespan and increased the frequency of the slow electrophoretic allele, Pgm^S, of the Pgm gene. If Pgm^S contributes to the increased lifespan phenotype of the selected strains, it might be because Pgm^S has increased activity, decreased activity or activity that is altered in some way such as through its regulation or subcellular localization. The Pgm alleles were cloned and sequenced and were found to differ by amino-acid substitutions consistent with the relative electrophoretic mobilities of the encoded enzymes.

A recent study reported the sequencing of several fast, medium and slow electrophoretic alleles of Pgm from a wild population of D. melanogaster [40]. The sequences reported in that study (Verrelli alleles) are comparable to those reported here (Ives alleles) as follows: the medium Verrelli allele used for alignment is identical in amino-acid sequence to the Ives medium allele. The Verrelli fast alleles are identical to the Ives fast allele in each of the three amino-acid positions at which the Ives fast differs from the Ives medium (wild-type) allele. There are a number of differences between the Verrelli slow alleles and the Ives slow allele. The Verrelli slow alleles are not all identical to each other. The Ives slow allele differs from the Ives medium (wild-type) amino-acid sequence at two positions (amino-acid positions 6 (E → G) and 9 (A → T)). The amino-acid change in the Ives slow allele at position 6 (E → G) is not reflected in any of the Verrelli slow alleles. The second amino-acid change in the Ives slow allele (at position 9) is shared by two of the Verrelli slow alleles. In addition, the Verrelli slow alleles have ammo-acid substitutions that are not shared by the Ives slow allele. The data are consistent with the conclusion that Pgm is highly polymorphic and subject to selection in natural populations [40,41]. Consistent with this idea, Pgm haplotypes and glycogen content have been found to vary among flies at different latitudes [42].

The tet-on promoter system allowed a test of the hypothesis that increasing or decreasing Pgm gene expression would affect lifespan. PGM enzyme levels were increased two- to five-fold using a sense transgenic construct, and Pgm RNA levels were decreased 1.3- to 24-fold by driving expression of an inverted-repeat construct. No changes in lifespan were detected that were consistent across multiple transgenic lines or replicate experiments. The results indicate that simply increasing or decreasing Pgm gene expression does not significantly affect the lifespan of adult Drosophila, at least under the conditions tested.

The Pgm^S allele used in these experiments was cloned from the Ives strain, which is the progenitor of the long-lived selected strains and their controls. The high degree of polymorphism of Pgm makes it possible that the slow electrophoretic allele in the Ives strain could consist of multiple DNA sequence forms. The failure to observe effects on lifespan could therefore conceivably have resulted from overexpression of the wrong DNA sequence form of Pgm^S. However, Pgm^S was cloned and sequenced from four independent lines derived from the Ives strain and the clones had the same DNA sequence. The data therefore suggest that most, if not all, of the Pgm^S alleles in these strains have the same sequence, making sequence heterogeneity an unlikely explanation of the present results.

The results suggest several possibilities for the relationship of Pgm allele frequency to the increased lifespan of the selected strains. The first is that Pgm^S might in fact contribute to the increased lifespan of the selected strains as a result of an increased or decreased activity, but that the effect on lifespan is too small to be detected in these assays. The lifespan assay appears to have a variability with control strains of approximately -5% to +5%, as seen in the data presented here and elsewhere (G.L., D. Bhole and J.T., unpublished results). It is therefore unlikely that an effect much smaller than 10% could be identified. The second possibility is that Pgm allele frequencies are altered because of selection for an unknown linked gene, and this possibility cannot be ruled out at this time. The experiments presented here suggest that tet-on regulated overexpression and RNAi of genes near Pgm may be a promising approach.

Third, an interesting model is that Pgm^S contributes to increased lifespan of the selected strains because of an alteration in its regulation and/or subcellular localization. The sequence analysis of the Pgm alleles reveals a novel potential site in Pgm^S for phosphorylation by PKC; however, the potential significance of this change is unknown. The Pgm^S allele was used for the overexpression experiments presented here, and presumably the overexpressed enzyme should be subject to any regulation or localization characteristic of Pgm^S. However, the Pgm^S enzyme was overexpressed in a 'wild-type' Pgm^M background, and it may be that any phenotypic consequences of Pgm^S allele function are not apparent when Pgm^M is also present. Finally, another possible explanation for the absence of an effect on longevity is the method of assay. The selected O stocks were handled and assayed with males and females together, repeatedly mating during adult life. The present assay used segregated sexes, which is a different environment. This was done because the segregated-sexes assay is sensitive to even small changes in lifespan, and has been used successfully in the past to detect the effects of other genes on lifespan. It is possible that an effect of PGM on lifespan might be identified with different lifespan assay conditions. However, even if this explanation were to be true, the data still indicate that PGM does not have a simple strong effect on longevity. Further experiments will be required to distinguish between these interesting possibilities for the link between Pgm allele frequency and lifespan.