Mapping Drosophila insulin receptor structure to the regulation of aging through analysis of amino acid substitutions

Genetic manipulations of the Drosophila insulin/IGF signaling system slow aging, but it remains unknown how the insulin/IGF receptor acts to modulate lifespan or differentiate this control from that of growth, reproduction and metabolism. With homologous recombination we produced an allelic series of single amino acid substitutions in the fly insulin receptor (InR). Based on emerging biochemical and structural data, we map amino acid substitutions to receptor function to longevity and fecundity. We propose InR mutants generate bias in the process of asymmetric transphosphorylation when the receptor is activated. This induces specific kinase subdomains that modulate lifespan by additive processes, one involving survival costs of reproduction and the other involving reproduction-independent systems of longevity assurance. We identify a mutant in the kinase insert domain that robustly extends lifespan without affecting growth or reproduction, suggesting this element controls aging through unique mechanisms of longevity assurance.

The breadth of insulin/IGF function, however, also challenges our ability to understand how it affects aging (37)(38)(39)(40). Insulin/IGF signaling simultaneously regulates many traits including growth, metabolism, cell proliferation, differentiation, Dauer/diapause, and reproduction. Furthermore, these traits are mediated by a single insulin-like receptor in C elegans and Drosophila while these invertebrates produce many insulin-like ligands (41)(42)(43). The insulinlike receptor is a molecular switch-board taking many incoming calls and routing each to distinct signaling destinations. Broadly, understanding such pleiotropy is a longstanding problem in receptor tyrosine kinase biology (44,45). For instance, how does mammalian IR and IGFR differentially control metabolism relative to growth (46,47), how does the fibroblast growth factor receptor (FGFR) modulate mitogenesis relative to glucose homeostasis (48), or how does the c-Kit receptor differentially activate progenitors of hematopoietic cells relative to mast cells (49)?
One approach to this problem studies insulin/IGF receptor amino acid variants in cultured cells, in polymorphic or mutant animals, and even in humans with inherited insulin and IGF resistance (50)(51)(52)(53)(54). Insight is developed by relating the amino acid substitutions to knowledge of receptor structure. This strategy was illustrated by work with C. elegans daf-2 (55,56). About two dozen daf-2 variants extend lifespan, and variously also effect Dauer, fertility, and growth.
Classes of these traits were defined by where substitutions fell in the ectodomain relative to the kinase domain, but understanding how the substitutions control outcomes was limited by the known structural data of the time.
Structural analysis has advanced in recent years and we can now productively apply this approach with Drosophila. The Drosophila insulin-like receptor was initially identified in the lab of Rosen (57). Subsequently, EMS mutants screened over a deficiency produced alleles that were viable as trans-heterozygotes (58)(59)(60). Sequencing of the cytoplasmic portion of several alleles identified substitutions in the kinase domain (42). These genotypes variously reduced cell growth, fecundity and receptor kinase activity (42,60). In early work with this receptor, aging was slowed by one EMS-generated mutant (InR E19 ) when heterozygous over a P-element insertion allele (5). Here, we characterize how other, archival single amino acid InR EMS mutants effect aging using the method of Quantitative Complementation Testing (61,62). We validate and extend these results by independently generating the putative substitutions of the InR EMS alleles through homologous recombination gene replacement and evaluate lifespan, growth, development rate and fecundity. Finally, recent advances provide remarkable insights into the structure of activated human insulin receptors (63,64). We combine these data with our genetic analysis to generate new ways to understand how the insulin-like receptor modulates aging.
We will propose four interpretations. 1) Drosophila InR mediates aging in two independent ways. One effects reproduction that carries associated adult survival costs, the other effects mechanisms of longevity assurance independent of reproduction. These outcomes may explain why a variety of longevity-extending manipulations reduce reproduction, but others do not: they impact different programs regulated by insulin/IGF signaling. 2) Reducing the quantity of insulin receptors has little effect on aging. Longevity mutants of the insulin receptor simultaneously change the intensity and quality of signaling. 3) The genetic effects of InR alleles may be understood through the model of tyrosine kinase receptor activation by asymmetric transphosphorylation (45,64). In wildtype homodimeric receptors the direction of asymmetry is random. We propose particular InR mutants bias asymmetry whereby only the protomer from one allele of the InR heterodimer can be activated. In mutant heterozygotes, aging is slowed when the activated protomer carries an appropriate mutation. 4) Some mutants slow aging by altering the receptor's specificity for substrate adaptor proteins. We identify a dominant substitution in the kinase insert domain that robustly increases lifespan yet paradoxically produces normal growth and robust fecundity in the lab setting. We hypothesize the kinase insert domain modulates a unique SH2 binding site that regulates mechanisms of longevity assurance that are independent of reproduction.

Demographic quantitative complementation
We used Quantitative Complementation Testing (QTC) to screen InR EMS-alleles (Frasch series) for their ability to slow aging. QTC measures allelic effects as the phenotypic difference of a tested allele complemented to a standard hypomorph or deficiency relative to when complemented to a standard wildtype allele (61,62,(65)(66)(67). Figure 1 illustrates mortality for several InR alleles when complemented to InR E19 relative to TM3, InR + Sb (Fig. 1, Table S1).
Mortality rate increases exponentially for all cohorts after the 'left-hand boundary' (68). The natural wildtype InR NC442 is tested in each block i of demographic analysis to control for period effects (Table S2). Figure 1G, H (Block 3) illustrates modestly reduced mortality (b < 0) for InR NC442 / InR E19 relative to InR NC442 /TM3, InR + Sb -a pattern observed in all blocks. In contrast, mortality coefficients for EMS alleles ranged from -1.80 to 1.64 (block adjusted bi adj ) (Table S1).
To infer when an allele slows aging, we asked when does an EMS allele effect mortality more than expected relative to segregating variation among a sample of wildtype alleles? Accordingly, we applied QCT to a collection of 18 wildtype InR alleles extracted from a natural population (NC series) and evaluated each EMS mutant relative to this distribution (z-test) (Table S1).
Mortality coefficients for five EMS InR mutants did not significantly differ from wildtype alleles. One EMS mutant (InR 327 ) increased mortally. InR 74 and InR 211 significantly reduced mortality in males and females. The InR 353 allele reduced female mortality but this difference did not reach significance relative to the wildtype distribution. The transformation $ ! "#$ $ estimates fold change in mortality: InR 74 and InR 211 alleles reduced male and female mortality 2.6-to 6-fold, while InR 353 reduced female mortality about 2-fold (Fig 2A, B).  Table S1).

Body size quantitative complementation
To describe the allelic effects on body size, we quantified the ratio of each EMS allele complemented to InR E19 relative to TM3, InR + Sb. All EMS InR alleles significantly reduced body size (female), with relative effects ranging from 0.9 to less than 0.75. (Fig 2C). Among the EMS InR alleles, relative body size does not associate with proportional change in mortality ( Fig   2D).

Homologous recombination alleles
Quantitative Complementation Testing has limitations: it only measures recessive allelic effects; it confounds epistatic interactions from cosegregating second site mutations; alleles are not derived from the same wildtype InR or background and may contain unrecognized polymorphisms; and allelic effects are relative rather than absolute and in each case these relative measures include deleterious effects of the TM3 haplotype.
To resolve these issues we generated new, single amino acid substitutions from a common InR in a universal background, focusing on alleles that potentially slowed aging in QTC, or in one case an allele that did not reduce mortality but produced small adults (InR 246 ).
We used published and unpublished data on the putative substitution sites ( Except for the allele InR E19(HR) which resides in the Fibronectin III-1 ectodomain, the remaining substitutions are located in the intracellular kinase domain (Fig 3).
We evaluated viability, eclosion time, and adult size for all allele combinations ( Fig 4A, B, Fig S2).

Figure 2. Traits from Quantitative Complementation Test of InR among EMS mutant and wildtype accessions. A) Males, B)
Females; distribution of proportional hazard (fold risk of mortality) for nine EMS mutants (red bars) and 18 wildtype (NC series) accessions (black bars); estimated as exp(|b|). Labeled InR alleles have difference relative to distribution of wildtype hazard (Table S1). C) Distribution of relative body size (head capsule width) among mutant (red) and wildtype (black) females. All mutants smaller than the expected distribution of wildtype, ztest each allele, p < 0.0007. D) Absence of correlation between relative female mortality (b) and relative body size among InR alleles assessed in QCT (r 2 = 0.08, p = 0.85). No. lines InR 353 Arg1466 InR 211 Gly1598 Insulin receptors are preformed dimers assembled with protomers from both alleles ( Fig   4C). Homodimeric receptors of InR (HR) mutants appear to be nonfunctional because homozygotes of each mutant are inviable. Heterozygotes between two different mutants must produce functional heterodimers because these genotypes are viable, but these heterozygotes will also generate fewer functional dimers per cell. This raises the possibility that InR mutant heterozygotes extend longevity because their cells contain fewer functional receptors. To address this issue, we measured the survival of the hemizygote InR +(HR) /InR null(HR) . Hemizygotes produced 26% less InR mRNA and increased life expectancy between 2 to 4 days when assessed across replicate trials, but only significantly so in Trial 2 ( Fig 5, Table S3). We use this 2 to 4day difference as a benchmark to infer when an InR genotype extends longevity more than expected from quantitative loss of functional InR receptor dimers.
In QCT, the InR 246 allele did not slow aging more than expected relative to a sample of wildtype alleles, although it did reduce body size. We confirm these observations with : adults are small (Fig 4) but not long-lived (Fig 6G, H). We note, however, the survival curve of InR 246(HR) /InR E19(HR) crosses over that of wildtype, suggesting the mutant cohort has high age-independent mortality potentially coupled with reduced demographic aging.  Table S4).
Overall, the homologous recombination alleles validate inferences from QCT but now measured as absolute effects. We also describe a new heteroallelic genotype that extend lifespan, and document an exceptional longevity benefit conferred by InR 353(HR) that was obscured in QCT because this allele is dominant.

Fecundity
Drosophila insulin signaling modulates fecundity, and survival is a well-known cost of reproduction (69). We therefore determined how the InR alleles effect egg production (Fig 7A,   B). Wildtype, hemizygotes and InR E19(HR) /InR 246(HR) have similar egg production. Mutant heterozygotes reduce fecundity. In contrast, InR +(HR) /InR 353(HR) females produce more eggs than wildtype. This exceptional fecundity could arise because these females have more ovarioles or produce more eggs per ovariole. Accordingly, we measured the number of ovarioles for all genotypes and regressed this trait against daily egg production. Fecundity associates with ovariole number, but ovarioles from genotypes with high fecundity also produces more eggs per day because the observed slope < 1.0 implies fecundity increases faster than the number of ovarioles ( Fig 7C). We therefore regressed lifespan against ovariole egg production (eggs/day/ovariole) across genotypes, treating the InR 353(HR) allele as a covariate (Fig 7D). A striking pattern emerges: egg production per ovariole is negatively associated with life expectancy (although, N=8, p = 0.06), but independent of egg production, genotypes with one   confers an additive effect upon longevity assurance that is independent of reproductive tradeoffs.

DISCUSSION
The impact of InR on longevity was originally studied with EMS-generated alleles heterozygous with a p-element InR mutant (3,59). In particular, InR E19 / P{PZ}InR 05545 extended lifespan and this genotype reduced InR tyrosine kinase activity (3). InR mutants are generally said to slow aging because they diminish insulin signaling, but this explanation is incomplete because many InR mutants reduce growth and reproduction without benefits to survival (3).
Rather, InR may regulate aging through signals that are distinct from how it effects other traits.
Here we explore this idea based on how Drosophila InR alleles impact aging, growth and reproduction.

Alleles in the quantitative complementation test
We used quantitative complementation testing (QCT) to identify EMS InR mutants that reduced adult mortality. Among alleles complemented to InR E19 relative to TM3, InR + Sb, three alleles reduced relative mortality more than expected from a sample of naturally occurring wildtype alleles. We followed the same approach to quantify how mutant alleles effect growth.
Every allele significantly reduced adult size, and there was no association between relative size and relative mortality among genotypes.
QCT has caveats (62,67). Our archival InR mutants were not generated on a common third chromosome. Unknown second site mutations could affect traits we attribute to InR and we cannot rule out the impact of epistatic loci that segregate with InR + upon the TM3Sb balancer chromosome. We only estimate recessive effects of the mutants as they complement InR E19 relative to when complement TM3, InR + Sb. We cannot evaluate other mutant heterozygote combinations. The causal substitutions within InR were not actually known, and unknown polymorphisms within each InR locus might exist in the stocks. Finally, QCT measures relative rather than absolute allelic effects, and the balancer haplotype itself appears to be somewhat deleterious.
To address these issues, we reconstituted the putative InR substitutions using ends-out homologous recombination. These de novo alleles provide new, well-defined substitutions derived from a common wildtype progenitor in a controlled genetic background. We produced four homologous recombination (HR) alleles of the Frasch series (InR74 (HR) , InR 211(HR) ,

Figure 7. Fecundity, ovarioles and visual model. A)
Among female mean eggs laid from eclosion through age 15d for mutant heterozygotes, wildtype and hemizygote. B) Average daily fecundity. Eggs per day averaged over 16 days. ANOVA with Tukey HSD post hoc analysis: genotypes with the same letter show no significant difference in average egg production. C) Regression of mean ovariole number (total from both ovarioles) upon average fecundity among genotypes. Least squares linear regression: R 2 = 0.85, m = 0.27< 1.0 F= 34.1, p < 0.001. D) Relationship between lifespan and rate of ovariole egg production among genotypes treating InR353 as a covariate. Average median lifespan is the average of median lifespans observed for each genotype among all its independent cohorts and accessions. Eggs/day/ovariole is estimated from average daily fecundity/mean ovariole number of each genotype. Parameters by ANCOVA: b (fecundity/ovariole) = -9.15, t = -2.58, p = 0.061; b (InR353) = 6.23, t = 4.25, p = 0.013; b (fecundity/ovariole x InR353) = 1.01, t = 0.28, p = 0.79). E) Model for biased asymmetry modulates aging through InR domain specific function. Wildtype protomers (blue) each have the potential to asymmetrically act as either enzyme or substrate, activating catalysis by either tyrosine kinase (TK), producing wildtype phenotypes through the activation of just one TK. Protomers with mutation in tyrosine kinase activity (green) can act as substrate or enzyme protomers; when heterozygous with a wildtype protomer, enough dimers produce wildtype TK activation for normal phenotypes. The InR E19(HR) protomer (grey) can only act as enzyme and transphosphorylate its partner protomer (biased asymmetry). When this is a wildtype protomer, the TK is active and phenotypes are normal. When the partner protomer has a hypomorphic mutation in the TK, this mutant TK reduces growth and reproduction, relieving survival costs of reproduction (COR) to extend lifespan. The 353 protomer only acts as substrate (biased asymmetry) but retains normal TK catalytic activity when transphosphorylated, and thus normal growth and reproduction. The protomer alters its kinase insert domain (KID) function, which induces a program of longevity assurance independent of reproduction.  The InR E19 substitution Val810Asp occurs in a linker sequence between the L2 and FnIII-1 ectodomains, corresponding to Phe475 of human IR (Ebina) and to Phe495 of human IGF-1R ( Figure 3, Suppl fig 3). Recent cryo-EM analyses show insulin induces a hinge motion in this linker to swing the fibronectin domains inward, bringing together the intracellular domains of each protomer (63,71,72). This proximity permits asymmetric kinase transphosphorylation, a non-reciprocal process whereby the kinase domain of one protomer acts as enzyme to phosphorylate the opposing (substrate) activation loop (A-loop) (45,64,73 Val1384Met protomer produces enough catalytic activity to permit growth although with high age-independent adult mortality.

InR 211(HR) : kinase domain C-terminal lobe
The InR 211(HR) allele is normal when heterozygous with wildtype, lethal as a homozygote and extends lifespan when heterozygous with InR E19 . Unexpectedly, InR 211(HR) also complements Pro1466Ser. We can gain insights on this segment from human the fibroblast growth factor receptor FGFR3 where the homolgous C-terminal region stabilizes the asymmetric dimer interface required for transphosphorylation (64). C. elegans Pro1466 corresponds to P694 of FGFR3, and substitution to serine will disrupt a stabilizing hydrogen bond in this interface. The Drosophila G1598R substitution corresponds to G697 of FGFR3 and this change is likely to disrupt an interface hydrophobic pocket. Based on the FGFR3 structure, we propose that loss of stability at the dimer interface reduces the ability of the protomer with Gly1598Arg to act as the enzyme in A-loop transphosphorylation, but maintains its capacity to be transactivated. Like InR E19(HR) , the biased asymmetry of InR 211(HR) may explain why the allele is normal over wildtype while it complements mutations on opposing protomers to extend lifespan.
Drosophila Arg1466 corresponds to IR Arg1092 (Ebina). Relative to the kinase insert of human IR, the Drosophila KID contains 12 additional amino acids including a potential SH2 motif, Tyr(1477)-Leu-Asn. This motif occurs in mammalian IRS-2 where it specifies binding with Grb2 (85) and deletion of IRS-2 extends mouse lifespan (30,86). Drosophila Chico likewise contains SH2 binding motifs for Grb/Drk (Grb/Downstream-of-kinase), and a Y243A mutation of this Grb/Drk site extends lifespan (87,88). Drosophila InR itself contains adaptor protein binding sites. The C-terminal tail has SH3sites that recruit Dock to direct axon guidance, and SH2-sites that recruit Chico to regulate growth (89). At the InR N-terminal juxtamembrane domain, SH2B sites recruit Lnk to modulate interaction with Chico, and mutation of lnk extends lifespan (90)(91)(92)(93). Based on these facts, we propose the Drosophila InR kinase insert domain likewise recruits Drosophila Grb/Drk, and Arg1466Cys impedes this interaction to induce longevity assurance.

Synthesis: Drosophila InR modulates aging through distinct structure-defined mechanisms
The phenotypes of InR +(HR) / InR 353(HR) present a paradox: adults are highly fecundity (and large) and yet long-lived. This positive trait association is seen in other cases involving C.
elegans and Drosophila and challenges life-history theory that expects longevity to trade-off with reproduction (1,56,94,95). Here we propose a mechanistic explanation: InR regulates survival in distinct, independent ways ( Fig 7E).
The first involves reproductive costs (69). The alleles InR E19(HR) , InR 74(HR) , InR 211(HR) , InR 246(HR) slow oogenesis, and previous work shows Drosophila longevity is increased when germline stem cell activity is suppressed (96). We propose these InR alleles quantitatively reduce germline proliferation and thus proportionally relieve survival costs-of-reproduction. This explanation is consistent with how life expectancy is negatively associated with per-ovariole oogenesis when the InR 353(HR) allele is treated as a covariate. How these alleles might modulate oogenesis is unknown but we propose they affect the level of kinase activity in ovarioles and secondarily reduce non-autonomous signals regulating survival-reproduction trade-offs.
The second mechanism confers longevity assurance independent of reproductive costs.
We propose the kinase insert domain of InR 353(HR) recruits adaptor proteins that regulate longevity assurance -systems that increase robustness, homeostasis and survival. These effects are additive to effects of reproduction on survival. Thus, while InR +(HR) /InR 353(HR) has elevated per-ovariole production and may suffer associated survival costs, the genotype gains ~13 days of reproduction-independent longevity assurance to yield a net increase in lifespan. Mutation to block Grb/Drk (Y243) re presses Ras-Erk signaling and extends fly lifespan without reducing growth or fertility (87,88).

T his pathway mediates the ETS transcription factor
Anterior-open (Aop). In a similar way, the C. elegans daf-2(sa223) a llele is proposed to alter Ras signaling and is long-lived (55,56).
We hypothesize the proposed SH2 motif Tyr(1477)-L eu-Asn of Drosophila KID facilitates binding of Grb2/Drk. We reason this SH2 binding event is repressed by Arg1466Cys of InR 353(HR) . Reduced Grb2/Drk recruitment and activation slows aging by specific mechanisms of longevity assurance transduced by Ras, as does the Chico Y243 substitution. At the same time the InR 353(HR) protomer still activates PI3K-Akt-Foxo signaling to maintain growth and fecundity.
Other InR mutant alleles, we propose, inhibit PI3K signaling; their substitutions reduce growth and fecundity, which extends lifespan by lessening survival costs-of-reproduction.
We suggest the longevity benefits of Ras-Erk-Aop and PI3K-Akt-Foxo regulate different cellular and physiological systems. The proposed KID-Grb2/Drk network controls longevity through mechanisms that do not involve costs-of-reproduction. Our data support emerging ideas (45,89,101,102) w here protein tyrosine receptors segregate phenotypes through the action of specific sites within particular structural domains. was used in each block of demographic analysis to provide a common strain to normalize survival data across test periods.

Genotypes for Quantitative Complementation Analysis
We made reciprocal crosses between every InR EMS /TM3 and InR NC /TM3 with

Drosophila insulin receptor alignments with human IR and IGF1R
The annotation of Drosophila InR and human IR and IGF1R varies among sources (Fig 3, S1).
The human insulin receptor (IR) is referenced as the mature long isoform of Ebina (105), and the Human Genome Variation Society (HGVS, http://hgvs.or/rec.html) that includes an additional 27 a.a. localization sequence (Fig 3, Fig S1). The human IGF1R sequence is numbered as NCBI Reference Sequence: NP_000866.1. Drosophila InR numbering begins at the first translation initiation site in wDah wildtype (GenBank accession MT_563159), +45 amino acids relative to the TIS reported by Fernandez (59) (NCBI Reference Sequence: NM_079712.6), which also differs by an insertion/deletion in the L1 ectodomain (wDah lacks H144) and the substitution H146Q. This polymorphism segregates along latitudinal clines of Drosophila natural populations (52). See Supplemental Materials (Fig S3, S4) for confirmed alignments of the homologous recombination allele series.

Sequence of EMS InR alleles
The EMS-generated InR E19 allele was sequenced using (rare) homozygous F1 adults generated from an InR E19 /TM3 stock (64). This allele contained several polymorphisms compared to InR of alleles InR 353 and InR 211 were reported by Brogiolo (42). In the current work, each of these described substitutions were used to design site directed mutagenesis with homologous recombination to produce coisogenic, single amino acid substitution alleles.

InR homologous recombination allelic series
Details of fly stocks and culture, constructs, primers, cloning and work flow for homologous recombination are in Supplemental Materials. In brief, targeting arms for ends-out homologous recombination of InR (cloned from wDah) (106) were cloned into the pW25.2 vector (Drosophila Genomics Resource Center, Bloomington, IN) and injected into w 1118 embryos by Genetic Services, Inc. (Sudbury, MA). Targeting arms contained single nucleotide substitutions to specify the desired amino acid replacements (Fig 3), while the native wDah sequence provided both arms to produce InR +(HR) . Flies with the white + eye color marker mapped to the first or second chromosomes were used for homologous recombination. These lines were crossed to flies expressing flipase and the restriction enzyme I-Sce1 to execute excision and linearization of the transgenic construct. Replicate, independent (accession) candidate insertion lines were selected based on white + mapped to chromosome 3. Gene replacement and mutant validation were confirmed by sequencing. Strains with gene replacement strains were crossed to flies expressing cre recombinase to remove the white + marker. The white + marker was retained to produce InR null(HR) . paraformaldehyde. The number of ovarioles per ovary was recorded as the sum over both ovaries from 15 females of each genotype.

Phenotypes of InR homologous recombination allelic series
Author contributions: R.Y., M.P. and N.C-J conducted the experiments. R.Y., M.P. and M.T. designed the experiments. M.T. and R.Y. analyzed the data and wrote the paper.