• Open Access

The demography of slow aging in male and female Drosophila mutant for the insulin-receptor substrate homologue chico


Marc Tatar, Tel.: +1 401 863 3455; e-mail: Marc_Tatar@Brown.edu


Hypomorphic mutants affecting the Drosophila insulin/IGF signal pathway are reported to increase longevity in females but not in males. To understand this sex-difference, we conducted a large-scale demographic study with three new isogenic strains of alleles at chico, the insulin-receptor substrate homologue. We verify that female dwarf homozygotes (ch1/ch1) and normal-sized heterozygotes (ch1/+) are long-lived, as originally reported. We find for the first time that male heterozygotes are long-lived relative to wildtype, by about 50%. The life span of male ch1/ch1 is similar to that of wildtype but these dwarf males age at a slow demographic rate. The levels of demographic frailty and of age-independent mortality are elevated in ch1/ch1 males, counteracting the effect of slow aging upon life expectancy. Mortality deceleration occurs amongst the oldest-old wildtype adults, as seen in many organisms. Remarkably, in similarly sized cohorts of male and female ch1/ch1 and of male ch1/+ mortality deceleration is absent. Mortality deceleration is a phenotype of chico.


Mutations that affect the insulin/IGF signal pathway increase adult survival in Caenorhabditis elegans and in Drosophila melanogaster. As originally reported in flies, mutants of genes in this pathway increase longevity in females but not in males (Clancy et al., 2001; Tatar et al., 2001). This pattern raises the question, are the genes of the insulin/IGF signal pathway sex-limited in their effects upon aging? Alternatively, these loci might reduce aging in males and females but differentially induce counterbalancing age-independent mortality. Both cases can produce mutant females with extended longevity and mutant males with ordinary life spans. To distinguish between these alternatives we present an intensive demographic analysis of chico, the D. melanogaster gene encoding the insulin-receptor substrate homologue (Bohni et al., 1999).

Insulin/IGF signalling appears to be a conserved regulatory system of aging (Kenyon, 2001). In Drosophila, mutants of the insulin-like receptor (InR) and of the insulin-receptor substrate homologue (chico) extend female longevity by 36–85% (Clancy et al., 2001; Tatar et al., 2001). In males these same alleles produce no net improvement in mean longevity, although InR heteroallelic mutants increase life expectancy measured at age 20 days. In C. elegans, hypomorphic mutants affecting insulin/IGF signalling extend life span in hermaphrodites and in males (Johnson, 1990; Kenyon et al., 1993; Gems & Riddle, 2000). In both species, aging is modified by insulin through cell-nonautonomous signals (Apfeld & Kenyon, 1998; Tatar et al., 2001), presumably in conjunction with secondary hormones and gonadal endocrine feedback (Arantes-Oliveira et al., 2002). This being so, sex differences in longevity might be expected since endocrine integration of reproductive and somatic physiology is inherently sex-specific. Alternatively, the insulin/IGF mutants may similarly modulate aging in males and females but still produce differences in longevity if they have sex-limited effects upon development and growth. Growth and body size are sexually dimorphic in D. melanogaster where males mature quickly and with relatively small size. Since insulin/IGF regulates cell growth and proliferation (Chen et al., 1996; Stocker & Hafen, 2000), male development and subsequent adult age-independent viability may be especially vulnerable to insulin hypomorphism.

These alternatives can be resolved, in part, through parametric analysis of mortality. Here we apply this approach to chico genotypes. Cohorts with extensive demographic and genetic replication are used to estimate age-specific mortality for all genotypes within each sex. A new system to control the effect of genetic backgrounds among chico alleles is introduced. We replicate the original finding of Clancy et al. (2001) –ch1 extends female survival both as a heterozygote and as a homozygote. Furthermore, we now see that male heterozygotes outlive wildtype. From the parametric analysis of mortality we find that male and female mutants age at equally slow rates, that ch1 has a countervailing, age-independent impact upon adult mortality, and that mortality deceleration at advanced ages is a phenotype of chico.


Non-parametric survival analysis and life tables

Life tables were contemporaneously constructed for each chico genotype, ch1/ch1, ch1/+, +/+. All genotypes segregated as sibs from inter se crosses among three independent backcross strains (de-2, de-4, de-5) and from self-crosses within strains. Thus, each genotype at chico was represented in six independent, highly co-isogenic backgrounds.

Survival among the F1 cohorts within each genotype is strikingly consistent (Fig. 1A,B). Wildtype adults are uniformly the shortest lived. Among heterozygotes, five of the six ch1/+ cohorts have similar survival proportions; the de-4 × de-4 cross is shorter lived. Survival is also consistent among dwarf ch1/ch1 cohorts within both sexes. Survival of female ch1/ch1 is superior at median and late ages but similar to that of ch1/+ at early adulthood. At early ages the survival of male ch1/ch1 is similar to wildtype but superior at later ages.

Figure 1.

Cumulative adult survival of chico genotype cohorts in (A) females and (B) males. Each line represents the combined offspring of replicate cages from within each cross. The six F1 background replicate cohorts are denoted by symbols, the genotypes at chico are categorized by heavy, thin and broken lines.

We combine data among the F1 replicates to statistically evaluate life tables of each genotype (Supplementary material). Female survival (Fig. 2A) and life expectancy (Table 1) is greatest in ch1/ch1, intermediate in ch1/+ and least in +/+. In males, overall survival is highest in ch1/+ (Fig. 2B). Life expectancy of male +/+ and ch1/ch1 are similar and less than seen in heterozygote (Table 1).

Figure 2.

Cumulative adult survival of chico genotype cohorts in (A) females and (B) males with the F1 background replicate cohorts within each genotype combined. The genotypes at chico are categorized by heavy, thin and broken lines.

Table 1.  Mortality and survival statistics cohorts of chico with best-fit mortality models. Parameters of the nested Logistic–Makeham–Gompertz model are frailty (λ), rate of demographic aging (γ), age-independent mortality (c) and mortality deceleration (s). Estimates are provided for parameters that correspond to the best-fit model. Within sex, parameters are compared among all pairs of cohorts; differences in superscripts within columns indicate significance between parameters based on the log-likelihood test (P < 0.001). Adult life expectancy is estimated from eclosion; superscripts indicate significant differences within each sex for overall survival (log-rank test, P < 0.001 in all cases).
 ModelλγcsInitial cohort sizeAdult life expectancy (days)
 +/+Logistic0.00286a0.137a 0.467a205727.1a
 ch1/+Logistic0.00175b0.110b 0.493a208436.9b
 ch1/ch1Gompertz–Makeham0.00119b0.0770c0.00606 193342.7c
 +/+Logistic0.00250a0.117a 0.413149431.1a
 ch1/+Gompertz0.00163b0.0723b  185946.7b
 ch1/ch1Gompertz–Makeham0.00461c0.0612b0.00439 161932.8c

Parametric mortality analysis

Differences in the shape of survival functions in Figs 1 and 2 suggest that patterns of mortality are altered by mutation of chico. We use the combined F1 data to assess variables of the nested Logistic–Makeham–Gompertz model (Pletcher, 1999). In the simplest case, the Gompertz model assumes that mortality rate µx at age x is µx = λeγx. The parameter λ represents demographic frailty or baseline mortality (Sacher, 1977; Vaupel et al., 1979). The parameter γ represents the rate of change in mortality with age, the demographic rate of aging. Two mortality features can be added to this model. When change in mortality decelerates at advanced ages, the Logistic model applies, s is non-zero and µx = λeγx[1 + (λs/γ)(eγx − 1)]−1. When cohorts experience age-independent mortality, denoted as c, the Gompertz–Makeham model applies and µx = c + λeγx. When both cases hold, the Logistic–Makeham model applies as µx = c+λeγx[1 + (λs/γ)(eγx − 1)]−1.

Deaths through day 4 are left censored to remove incidental mortality associated with initiation of the demography cages. We test for improved goodness-of-fit relative to the simple Gompertz model by comparing the log-likelihood when s and c are fixed at zero relative to when s, c, or s and c are freely estimated. The more parameterized model provides a better fit if twice the difference in log-likelihood is greater than 3.84 (log-likelihood test, d.f. = 1). Table 1 summarizes model choice, parameter estimates and statistics for genotypes within each sex.

Mortality among wildtype and heterozygote females is best described as Logistic. Mortality at older ages presents a strong plateau near µx = 0.38 (Fig. 3). To compare parameters among genotypes, we estimate all variables based on the most complex model of the pair and then assess the log-likelihood under this case relative to the case when the tested variable is estimated as common to both cohorts (Table 1). Following this procedure, s does not differ between wildtype and heterozygote females but λ and γ are reduced in the heterozygote. The demographic rate of aging (γ) is further reduced in dwarf females. Mortality deceleration (s) is not significant in dwarf females. Since similar sized cohorts are studied for all genotypes, the absence of s in ch1/ch1 (or in ch1/+ males) is not likely to be an artefact of insufficient demographic power. Dwarf females also present a significant value for the Makeham term (c), unlike wildtype and heterozygote females.

Figure 3.

Mortality rate of chico genotype cohorts in (A) females and (B) males with the F1 background replicate cohorts within each genotype combined. The genotypes at chico are categorized by heavy, thin and broken lines.

Males show a similar progression from Logistic to Gompertz–Makeham with increasing dose of the ch1 allele, although ch1/+ is best described by the simple Gompertz. The dwarf male, as with the female, has Gompertz–Makeham mortality. As in females, the demographic rate of aging (γ) in both ch1/+ and ch1/ch1 is reduced relative to wildtype. Dwarf males have significant age-independent mortality (c) and the largest observed value of demographic frailty (λ). Αs a result, although mortality accelerates with a slow rate in male ch1/ch1, this genotype is not long-lived.


As reported by Clancy et al. (2001), the ch1 allele extends adult life span. We independently replicate this result in a new genetic background. Our strains have useful features for the demographic analysis of a single locus mutation. First, the replicate backcross lines have a total of 54 meiotic events to reduce linkage disequilibrium about the chico locus. Second, crosses among isogenic lines produce F1 offspring with reproducible backgrounds. Finally, all genotypes segregate within sibships and alleles within each stock perpetually retain background similarity since we do not rely upon balancer chromosomes.

We found remarkable agreement among the replicate life tables within genotypes but note that a strain can occasionally produce a stray result (i.e. de-4 × de-4). Overall, we verify that females are long-lived as dwarf homozygotes (57% increase relative to wildtype) and as normal size heterozygotes (36% increase relative to wildtype). As reported by Clancy et al. (2001), small body size is not a necessary condition for extended longevity mutants affecting insulin/IGF signalling in D. melanogaster. Among males, we find that heterozygotes are 50% longer lived than both dwarf and wildtype. The data of Clancy et al. (2001) suggested this pattern but inference was limited by cohort size. We resolve these issues and show that ch1 has a clear survival advantage for males as well as for females.

How ch1 differentially affects life span of males and females might be understood from the perspective of mortality patterns (Vaupel, 1986). Life expectancy and the level of mortality acceleration are similarly ranked across genotypes in females. Overall, mortality accelerates the slowest and maintains the lowest level in the ch1/ch1 cohort. Female heterozygotes have an intermediate value for γ and share with ch1/ch1 a low level of frailty (λ). In contrast to females, frailty (λ) modifies the rank order of life expectancy across male genotypes. Dwarf males, like females, have a Makeham–Gompertz mortality trajectory. The Makeham term (c) represents a degree of age-independent mortality that is superimposed upon a slow rate of demographic aging. In males the frailty of ch1/ch1 is relatively large. As a result, the absolute value of early adult mortality dominates the cohort and reduces life expectancy at eclosion. Wildtype and heterozygote males are free of age-independent mortality and have frailty levels similar to the levels estimated for females. The absence of extended longevity in slowly aging ch1/ch1 males is the outcome of sex-by-genotype expression of elevated frailty (λ) combined with age-independent mortality (c).

These data imply that ch1 is pleiotropic for demographic traits. The allele has dominant beneficial effects upon senescence and recessive deleterious effects upon age-independent mortality. A single allele of ch1 reduces the demographic rate of aging and the initial mortality rate but incurs no age-independent mortality; life span is extended in both sexes. Two alleles of ch1 can further reduce the rate of demographic aging but also increase age-independent mortality factors and, in males, frailty. The increase in age-independent mortality may result from strong hypomorphism of insulin/IGF function during development that retards growth and delays pupation. Altered development may be directly deleterious if it produces anomalies in morphology or physiology of adults. As well, slow growth could be indirectly detrimental if it retards escape from a deteriorating larval environment, which in turn may reduce adult fitness. Since males are intrinsically small in size and even more so as dwarfs, they may be especially sensitive to either of these mechanisms.

The data reveal intriguing genotypic differences among the patterns of old-age mortality. Mortality deceleration is a common feature of many animal life tables, including D. melanogaster (Vaupel et al., 1998). Cohort heterogeneity with the selective loss of frail individuals can generate mortality deceleration (Vaupel & Yashin, 1985). Alternatively, personal rates of senescence may decline with advancing age. It is difficult to empirically distinguish among these models but genetic approaches may provide a useful tool. Quantitative genetic analysis of D. melanogaster reveals heritable variance for the mortality deceleration parameter s of the Logistic model (Promislow et al., 1996). Here we find an explicit gene that influences the parameter s. Among females, mortality decelerates in wildtype and heterozygote genotypes. In contrast, mortality among the old dwarf females (ch1/ch1) accelerates beyond this level and with no apparent attenuation. Among males, late age mortality strongly decelerates in the wildtype cohort. Male heterozygotes and homozygotes accelerate beyond this plateau. These data indicate that mortality deceleration is a demographic phenotype of chico.

If the ch1 allele decreases variance in frailty, the loss of late-age mortality deceleration in ch1 cohorts may result from reduced opportunities for demographic selection. In this case, sex differences may arise if ch1 more strongly reduces variance for frailty in males. Alternatively, the individual rate of senescence may decelerate within wildtype individuals but continue to accelerate in ch1/ch1 individuals even as this genotype slows the overall rate of aging. In this case, sex differences for deceleration may occur if a single ch1 allele in males fully expresses the slow-aging and the loss-of-deceleration phenotypes while a single dose of ch1 in females weakly affects aging and has no impact on deceleration. The genetic features of our chico strains offer a new experimental model to study these alternative mechanisms of mortality deceleration.

Experimental procedures

Strains segregating alleles of chico

The original strain of ch1 P{ry+} was provided by E. Hafen (Zurich), the same source of ch1 used in Clancy et al. (2001). We crossed ch1 into an isofemale line with markers cn/cn; ry506/ry506. Females of ch1, cn/cn; ry/ry were recovered through progeny testing and backcrossed to males of the cn/cn; ry506/ry506 line. Recombinant offspring (cinnabar) were retained. Eighteen backcross generations were used to isogenize ch1 in three independent lines (denoted de-2, de-4 and de-5). Presently, segregation of chico alleles is maintained by propagation of heterozygotes (normal-size, cinnabar). For demography, segregating genotypes among sibs were identified as: ch1/+ normal-size, cinnabar; ch1/ch1, dwarf, cinnabar; +/+, normal-size, apricot.


Life tables were generated for F1 offspring of direct (de-2 × de-2, de-4 × de-4, de-5 × de-5) and inter se (de-2 × de-4, de-2 × de-5, de-4 × de-5) crosses. Reciprocal crosses within genotype were combined. Similar egg densities were laid up for all crosses. Larvae developed in standard cornmeal–dextrose–agar–yeast media supplemented with live yeast. From each cross, sibs that emerged within an 8-h period were sorted by genotype under light CO2 and placed in three demography cages with about 200 flies per cage, mixed sex. Demography cages were 1-L clear food service containers with a screened lid, a gasket-covered opening for access of an aspiration pipe and a short connector tube the width of a shell food vial. Food vials were fixed to the connector tube by a plastic sleeve. With these cages, flies remained in place day-to-day while fresh food (standard diet plus excess live yeast) was provisioned on alternate days. Dead flies were aspirated from cages when food was changed. Three replicate cages were started for each genotype. Cohorts of all genotypes were synchronously initiated and maintained at 25 °C, 40% r.h and a 12-h light cycle.

Survival was calculated as the proportion remaining alive at day x relative to the number of adults N0 initiating the cohort. Data of replicate cages were combined. N0 was estimated from the extinct cohort method as inline image where dx was the number dead in the 2-day period ending at age x. Adult life expectancy was estimated from eclosion (adult age 0-days) and calculated by the actuarial method with the census interval of 2 days.

Parameter estimates and hypothesis tests for Gompertz-family mortality models were executed with WinModest (Pletcher, 1999) where a series of nested models can be systematically evaluated with maximum likelihood methods.


Support provided by the Ellison Medical Foundation and the U.S. National Institutes of Health (AG-16632).