{"paper_id":"09656093-04ec-4c78-b7bf-edddd36e56eb","body_text":"Parallel adaptive responses to postponed reproduction increase lifespan and immune \ndefense \nKaren Alma Gamboa-Santarosa†a, Giovanni A. Crestani*†b, Alejandro Morana, Dolly Modhaa, \nHannah S. Dugob, Mansour Abdolic, Molly K. Burke‡b, and Parvin Shahrestani‡a \n† These authors have contributed equally to the work. \n‡ These authors have contributed equally to the work. \n* Corresponding author. \nAffiliations: \na Department of Biological Science, California State University, Fullerton \n800 North State College Blvd. Fullerton, CA 92831, USA b Department of Integrative Biology, \nOregon State University 2403 Cordley Hall, 2701 SW Campus Way. Corvallis, OR 97331, USA \nc Department of Mathematics, California State University, Fullerton 800 North State College \nBlvd. Fullerton, CA 92831, USA \nCorresponding author email: gacrestani@gmail.com \nCorresponding author mailing address: 2403 Cordley Hall, 2701 SW Campus Way. Corvallis, \nOR 97331, USA  \n1 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nABSTRACT \nExperimental evolution studies with Drosophila melanogaster have long played a role in the \neffort to dissect the genetic basis of aging and longevity. While selection for postponed \nreproduction reliably extends lifespan, additional phenotypic consequences and the genomic \nbases of this adaptation remain unclear. Here, we leveraged the highly replicated Drosophila \nExperimental Evolution Population (DEEP) system to further investigate the relationship \nbetween longevity and other life-history traits. Derived from the same wild population, two \ntenfold-replicated treatments, each with populations derived from two different ancestral \nbackgrounds: a control treatment in a 14-day generation cycle was maintained for 56 \ngenerations, and an experimental treatment in a 70-day cycle maintained for 20 generations. \nExperimental populations evolved to have longer lifespan, delayed development time, increased \nfecundity, greater stress resistance, and stronger immune defense. Pooled-population genomic \ndata reveal highly convergent allele frequency shifts within treatments, and point to 300 \ncandidate genes underlying differentiated phenotypes. Candidate genes are enriched for \nfunctional categories involving neural development and morphogenesis rather than canonical \naging or immune defense pathways. These results recapitulate that selection for postponed \nreproduction drives broad physiological changes and a highly polygenic adaptive response, with \nan unprecedented level of experimental replication. \nKEYWORDS \nExperimental evolution, Drosophila melanogaster, Beauveria bassiana, genomics, aging, \nimmune defense \n2 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nINTRODUCTION \nNatural populations harbor substantial genetic variation for lifespan and aging, defined as the \npost-reproductive decline in survival and fertility. Evolutionary theory provides a theoretical \nframework for understanding why lifespan is limited and variable: the force of natural selection \ndeclines with age (Hamilton, 1966), so alleles with late-acting deleterious effects may \naccumulate in populations (Medawar, 1952), while alleles with early-life benefits can be favored \ndespite later costs (Williams, 1957). These processes generate segregating genetic variation for \nlifespan, and predict that longevity can evolve. Indeed, classic selection experiments with \nDrosophila melanogaster and other species demonstrate that selection for postponed \nreproduction consistently produces longer-lived populations (reviewed by McHugh and Burke, \n2022; Rose et al., 2004), and this postponed longevity is accompanied by correlated shifts in \nother life-history traits. \nPostponed longevity has been repeatedly evolved in Drosophila across laboratories (e.g. \nLuckinbill et al., 1984; Partridge and Fowler, 1992; Remolina et al., 2012; Rose, 1984; Zwaan et \nal., 1995). Within tens of generations, evolved populations exhibit altered reproductive schedules \nand improved stress-resistance phenotypes in addition to longer lifespans (reviewed by Burke \nand Rose, 2009). More recently, Evolve and Resequence, or E&R (Long et al., 2015; Turner et \nal., 2011) strategies have been applied to characterize the genetic basis of these traits. E&R \nstudies consistently show that the genetic architecture of aging-related traits is highly polygenic \nand involves subtle frequency shifts at many loci (Burke et al., 2010; Carnes et al., 2015; Fabian \net al., 2018; Graves et al., 2017; Hoedjes et al., 2019). However, candidate genes differ across \nstudies, with little overlap and few canonical longevity genes implicated by any of them \n(explored by Fabian et al., 2018; Hoedjes et al., 2019). This inconsistency likely reflects \n3 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nvariation in founder populations, experimental design, and selection regimes, but could also stem \nfrom low statistical power. \nMost E&R longevity studies employ modest replication (3-5 populations), small population sizes \n(Ne<1000), and relatively short experimental durations. Simulations and empirical work suggest \nthat replication is the critical parameter for linking phenotypes to genomic loci (Baldwin-Brown \net al., 2014; Burke, Liti, et al., 2014; Kofler & Schlötterer, 2014). Underpowered designs obscure \nsubtle but replicable genomic responses, limiting the ability to distinguish true targets of \nselection from background noise. Thus, the lack of overlap across studies could be systematic of \nthe high false positive and false negative rate associated with insufficient replication, or it could \nimplicate genuine biological contingency. Highly replicated designs provide a powerful route for \nevaluating the repeatability of evolution and for testing whether selection for postponed \nreproduction consistently targets particular physiological processes. \nOne process of growing interest in the context of experimentally evolved longevity is immune \ndefense. While most E&R longevity studies have not implicated canonical aging genes, Fabian et \nal., 2018, reported differentiation at loci associated with immune defense between long-lived and \ncontrol populations. This observation raises the possibility that immune defense, or the response \nafter the exposure to a pathogen, may be an underappreciated axis of adaptation to postponed \nreproduction. Conceptually, this link is plausible. Immune activation can confer substantial \nearly-life benefits with pathogenic challenge, yet chronic or excessive immune signalling can \naccelerate senescence, a pattern broadly consistent with antagonistic pleiotropy (Schwenke et al., \n2016, Franceschi et al., 2000). Across taxonomic groups, mounting or maintaining immune \nresponses is energetically costly, and investment in immunity is often coupled with shifts in \n4 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nreproductive allocation, stress physiology, and somatic maintenance. These life-history trade-offs \nposition immune defense as both a potential early-life enhancer of fitness and also a late-life \nliability, aligning neatly with the evolutionary theory of aging. With this in mind, investigators \nhave begun to examine the relationship between immune defense and longevity in Drosophila \nthrough the lens of experimental evolution, though the extent to which each trait evolves as a \ncorrelated response to selection for the other therefore remains unclear. When immune defense is \nthe focal trait under direct selection, enhanced immunity appears to come at the cost of \nreproductive output and lifespan (Shahrestani et al., 2021). Alternatively, when age of \nreproduction is the focal trait under direct selection, improvements to longevity and reproduction \nlead to correlated improvements in immune defense without apparent costs (Bagheri et al., \n2025), and differentiated loci implicate genes related to immune function (Burke et al. 2014; \nFabian et al. 2018). These mixed outcomes underscore the complexity of interactions between \nimmunity and aging, and highlight the need for more powerful, highly-replicated tests. \nHere, we leveraged the Drosophila Experimental Evolution Population (DEEP) system (Bagheri \net al., 2025) to conduct an E&R experiment with unprecedented replication. The DEEP system \ndescends from Rose’s (1984) classic founding populations and includes tenfold replicated \ntreatments of postponed reproduction (“O-type”) versus baseline controls (“B-type”). Our goals \nwere to: i) characterize the phenotypic consequences of selection for postponed reproduction, \nwith an emphasis on immune defense; and ii) describe the genomic architecture of this \nadaptation by comparing allele frequencies across the first 20 generations of O-type selection. \nWe predicted a polygenic basis, with convergence across replicate populations and ancestral \nbackgrounds (Graves et al., 2017). We observed longer lifespan, increased fecundity and body \nweight, longer development times, greater stress resistance, and notably stronger immune \n5 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \ndefense. Genomic analyses revealed highly polygenic, convergent adaptation, implicating genes \nnot linked to canonical aging pathways or immune defense, but to neural development and \nmorphogenesis.  \n6 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nMETHODS \nDEEP Experimental Evolution System \nAll populations in the Drosophila Experimental Evolution Population (DEEP) System were \ngenerated from wild D. melanogaster sampled from an apple orchard in South Amherst, MA (the \n“IV”/Ives population) that have been maintained in the laboratory under controlled conditions for \nnearly 50 years (Ives, 1970). This ancestral IV population is maintained on discrete 14-day \ngeneration cycles and was used to establish five “baseline” populations (B1-5) and five long-lived \npopulations (O1-5) via experimental evolution for postponed reproduction (Rose, 1984). The B1-5 \npopulations are maintained on 14-day generation cycles, while in the O1-5 populations, the age of \nfirst reproduction was progressively postponed until generation cycles could be routinely kept at \n70 days. In 2008, a new set of populations was created from the O1-5 treatment; their regime was \nreverted to the ancestral 14-day generation cycles, and they were named BO1-5 (Burke et al., \n2016). The B1-5 and BO1-5 populations rapidly converged phenotypically (Burke et al., 2016) and \nat the genomic level (Graves et al. 2017). \nIn October 2018, B1-5 and BO1-5 populations were transferred from Dr. Michael Rose’s laboratory \nat the University of California, Irvine (UCI), to Dr. Parvin Shahrestani’s laboratory at the \nCalifornia State University, Fullerton (CSUF). At CSUF, the B1-5 and BO1-5 populations are \nmaintained under nearly-identical conditions to those at UCI, including a banana-molasses diet, \n24-hour light cycle, constant 25°C temperature, cage enclosures, and large census population \nsizes (>1,000). All 20 populations are reared in vials for 12 days before we dump each vial in its \nown cage. Six generations after this transition, we derived two treatments called nB1-5 and OB1-5 \nfrom each B1-5 population, and two treatments called nBO1-5 and OBO1-5 from each BO1-5 \n7 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \npopulation (Figure 1). The control treatments nBO1-5 and nB1-5 remained as “B-type”(on 14-day \ngeneration cycles), with the “n” designator standing for “new” and being added to indicate that \nthese populations are maintained outside the ancestral UCI environment. In these populations, \nflies were transferred from vials into cages on day 12 from egg, given yeasted food plates on day \n13, and had their eggs collected from them on day 14. On the other hand, the “O” designator \nsignals experimental populations selected for delayed reproduction, culminating in a 70-day \ngeneration cycle. They are referred to as the “O-type” populations. The transition from B-type to \nO-type was done gradually to minimize the chance of a population crash (i.e. population sizes \nwere kept at >1,000 individuals each generation), and to maximize the genetic diversity \npreserved. This gradual increase involved two generations of 21-day cycles, two generations of \n28-days cycles, five generations of 35-day cycles, two generations of 42-day cycles, two \ngenerations of 56-day cycles, two generations of 63-day cycles, and finally 70-day cycles for \neach subsequent generation. \nThe 20 populations (nBO1-5, OBO1-5, nB1-5, and OB1-5) constitute a valuable resource for \nexploring Drosophila life-history traits. With five-fold replication for each ancestry within each \nregime (B-derived vs. BO-derived), and ten-fold replication for each longevity regime (B-type \nvs. O-type), this system offers a rich set of possible comparisons. Figure 1 summarizes these \npopulations, their ancestry, and their current maintenance regime. \nPhenotyping \nWe surveyed life-history, stress resistance, and immune phenotypes in all 20 experimental \npopulations after approximately 20 generations of O-type selection. Longevity, development \ntime, fecundity, and body weight were assayed at exactly generation 20, and immune defense \n8 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nand starvation/desiccation resistance were assayed at generation 22. For all phenotype assays, \nflies were sampled from each population and cycled through two 14-day generations prior to the \nassay to minimize parental or environmental effects. Adult flies were assayed at day 14 from egg \nexcept where otherwise indicated. We also performed, with identical methods, a longevity and a \ndevelopment time assay at an intermediate timepoint (generation 12 of O-type selection), and \ndata are presented at Supplementary Figure 1. We summarize each assay procedure below. \nAdditional details and expanded protocols can be found in Walsh (2022). \nAge-specific mortality rate & longevity \nTo quantify age-specific mortality and lifespan, we established three replicate cages for each \nexperimental population, each initiated with around 500 flies (60 cages total). Dead flies were \nsexed, counted, and removed from cages four times a week until all flies died in all cages. For a \ngiven day d, mortality rate was calculated with the following equation:  \nmd = ln(Dd/Ad-1) (Eqn 1) \nwhere Dd is the number of flies that died on that day and Ad-1 is the number of flies alive at the \nend of the previous day. Mean population longevity was calculated by averaging mean cage \nlongevity across the three cages representing each experimental population. \nDevelopment time \nFollowing a pre-laying period, adults from each population were provided fresh yeasted food \nplates (one per cage) and allowed three hours to lay eggs. Batches of 60-80 eggs were transferred \nfrom each food plate into each of five vials of fresh banana medium (100 total vials across all \npopulations). Vials were monitored throughout their pupal development, and adult eclosion was \n9 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nrecorded beginning from the emergence of the first fly from its pupal case in each vial. Vials \nwere checked every 6 hours until all eggs eclosed. Mean development time per vial was \ncalculated and then averaged across all five vials to determine mean population-level \ndevelopment time. \nFecundity \nFor each population, we established 10 vials with a removable cap containing yeasted charcoal \nmedium (to facilitate egg counting), into which we transferred four male and four female flies. \nFlies were anesthetized daily to replace the charcoal medium, and the number of eggs on each \nday-old charcoal plug was recorded at the same time daily. Vials retained four male and four \nfemale flies throughout the experiment. Dead flies were sexed and replaced (using CO2 \nanesthesia) with a fly of the same sex sourced from a designated \"Replacement\" (R) replicate \nvial, which would always be the highest numbered replicate vial (e.g. nB1-vial10 would be \nrelabeled nB1-vial10R after being used to refill another vial). Once an R-vial was depleted, the \nnext highest-numbered replicate (e.g. CB1-vial9) became the new source. Egg counts from \nR-vials were included in the analysis only for the day before their first use as a source; \nsubsequent egg counts were excluded. All R-vials were maintained under the same daily \nhandling and feeding conditions as experimental vials. \nBody weight \nBody weight was measured for flies aged 17 days from egg. For each population, 50 females and \n50 males were sorted into groups of 10 same-sex individuals and weighed in pre-weighed \nmicrocentrifuge tubes. Mean vial weight was calculated for each sex and then averaged across \nreplicate vials to obtain mean population weight. Additional measurements of dry weight and \n10 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nwater loss from related assays are reported in Walsh (2022). \nStarvation and Desiccation Resistance \nTo assay starvation resistance, we collected 20 females and 20 males at age 15 days from egg. \nGroups of five same-sex individuals were anesthetized and transferred to vials (80 vials total), \nwhich were capped with a cotton ball soaked with 5 mL of DI water beneath a foam-plastic plug \ncut in half. We sealed vials with two layers of parafilm to conserve internal humidity. This setup \ndenied flies access to food, while providing enough moisture to prevent death from dehydration. \nVials were monitored every four hours and deaths recorded as they occurred. An identical \nprotocol was used for desiccation resistance assays, replacing wet cotton balls with 3 g of \ndesiccant per group. Vials were monitored every hour and deaths recorded as they occurred. For \nboth assays, we calculated mean population resistance by averaging mean vial resistance across \nall replicate vials of a given population. \nImmune Defense \nTo quantify immune defense we exposed flies to the fungal entomopathogen Beauveria bassiana \nstrain GHA (Bioworks, Inc., Victor NY , lot number TGA1-96-06B). Flies were sampled \nrandomly from rearing vials and divided into two replicate infected and two replicate control \ncages, with 100 to 200 flies per cage (estimated by volume during sampling), for a total of 80 \ncages. Infected cages were sprayed with a solution of 0.3 g of fungal spores suspended in 25 mL \nof a 0.03% Silwet surfactant solution, which was shaken by hand and a wrist-action shaker for 15 \nminutes. The control cages were sprayed with a solution of DI water and surfactant. For each \nspray, anesthetized flies were spread out on Petri dish lids and placed on ice to keep them \nimmobile. 5mL of the appropriate spray was applied using a custom-built spray tower \n11 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n(Shahrestani et al., 2021) designed to introduce controlled, even doses to the fly cuticle. Sprayed \nflies were transferred to population cages and maintained at 100% humidity for 24 hours to \npromote fungal germination and cuticle penetration. The humidity was then lowered to ~60% for \nthe remainder of the assay, and fly mortality was monitored daily for 12 consecutive days. \nInfected and control death percentages were calculated by averaging across the two replicate \ncages per population. \nPhenotypic Statistical Analysis \nAll statistical analyses were conducted in R version 4.5.0 (R Core Team, 2025). Data \nmanipulation used the packages readxl v1.4.5 (Wickham & Bryan, 2015) and tidyverse v2.0 \n(Wickham et al., 2019), and ggplot2 v3.5.2 (Wickham, 2016) was used for visualization. \nSurvival analyses were performed using survival v3.8-3 (Therneau et al., 2024) and survminer \nv0.5.1 (Kassambara et al., 2025). For mortality rate, we used Cox Proportional Hazard regression \n(Cox, 1972) via survival::coxph, modeling survival time as a function of “Regime”, “Sex”, and \n“Ancestry”. Proportionality assumptions were tested using survival::cox.zph, and all parameters \nfailed the test. Hence, the model was adjusted to allow the effect (hazard ratio) of the coefficients \nto change over experimental time. Differences in population mean values for other phenotypes \nwere analyzed using linear models or generalized linear models when normality assumptions \nwere violated. Fixed effects included “Regime”, “Sex”, and “Ancestry”, with “Infection” added \nfor immune defense. We considered the interaction terms “Infected:Regime” when analyzing \nImmune defense data, and “Ancestry:Regime” when analyzing all phenotypes. Welch’s t-tests for \npopulation mean comparisons (Figure 2 and Supplementary Figure 2) were conducted using the \npackage ggpubr v0.6.1 (Kassambara, 2016). Full statistical analysis output tables are available as \nSupplementary Table 1. \n12 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nSequencing and Bioinformatic Processing \nLibrary preparation \nLive flies were sampled from each of the 20 populations on day 14 from egg at generations 01 \nand 20 of O-type selection, and cryopreserved in ethanol at -80°C. Genomic DNA from 100 \nfemale flies per replicate per treatment per timepoint (40 pools total) was extracted using Qiagen \n(Puregene) reagents, following the vendor’s protocol for bulk extraction. We prepared \nsequencing libraries using a high-throughput tagmentation protocol (modified Illumina Nextera \nlibrary preparation protocol), pooling the 40 uniquely barcoded libraries into a single multiplex. \nThis multiplexed library was run on three lanes of a NextSeq2000 (P2, PE150) at the Oregon \nState University Center for Quantitative Life Sciences. Low-coverage samples were identified, \nre-pooled (from the same original group of 100 females), and re-sequenced on an additional two \nlanes to obtain higher and more consistent coverage across all populations.  \nGenomic Pipeline \nWe modified an existing GATK-based Pool-SEQ pipeline (Phillips et al., 2020) to identify SNPs \nand small INDELs relative to the D. melanogaster genome (v. 6.51), using Nextflow (Di \nTommaso et al., 2017) to increase efficiency and reproducibility. We aligned reads using BWA \nMEM v0.7.18 (Vasimuddin et al., 2019) with the default program parameters and called variants \nwith GATK v4.2.0.0 (Auwera & O’Connor, 2020), following best practices (Poplin et al., 2018). \nThen, we filtered the resulting VCF file with GATK VariantFiltration using the following \nfiltering parameters: “QD < 5.0, MQ < 40.0, FS > 60.0, ReadPosRankSum < -8.0, MQRankSum \n< -12.5”.  \nWe converted this merged VCF file into a SNP table containing AD (allele depth) and DP \n13 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n(unfiltered depth) as the alternate allele count and coverage for all SNPs passing the quality \nfilters. To ensure high-quality data for downstream analysis, we filtered out SNPs with coverage \nbelow 50 or above 500 in at least one sample (Schlötterer et al., 2014). Additionally, we imposed \na minor allele frequency filter to exclude SNPs with a combined frequency of 1% across all \nsampled populations, eliminating both variants fixed in all samples (uninformative) and spurious \nvariants likely caused by sequencing or variant calling errors.  \nAllele frequency analysis \nAfter quality filtering, we calculated allele frequencies by dividing the alternative allele count \n(AD) by the coverage (DP). We then performed a Principal Component Analysis (PCA) \n(function stats::prcomp) in R to visualize the extent to which ancestry versus selection treatment \nshaped genome-wide allele frequencies. We applied a K-means clustering algorithm (k = 3) on \nallele frequency data to verify grouping patterns. We chose k=3 because we expected three \ngroups, one with experimental samples at generation 20, and one per ancestry containing \nexperimental samples at generation 01 and control samples. To identify alleles associated with \nthe observed phenotypes, we applied an adapted CMH test (Spitzer et al., 2020) on a scaled SNP \ntable (we set DP to 100 for all SNPs, and adjusted AD proportionally). Scaling is applied to \ncontrol for coverage heterogeneity. We corrected for multiple testing bias (FDR-corrected), and \nconsidered the SNPs with adjusted p-values below a threshold of 1-100 as statistically significant \n(roughly top 0.1% quantile). \nSNP annotation and Gene Ontology (GO) term analysis \nWe used TxDb.Dmelanogaster.UCSC.dm6.ensGene v3.12.0 (Bioconductor Core Team, 2017), \nAnnotationDbi v1.70.0 (Hervé Pagès, 2017), and GenomicRanges v1.60.0 (Lawrence et al., \n14 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n2013) to to obtain a list of genomic features containing significant SNPs. To explore the GO \nterms associated with the genes, we used the R packages org.Dm.eg.db v3.21.0 (Carlson, 2017) \nand biomaRt v2.64.0 (Durinck et al., 2009). We excluded GO terms annotated with the TAS, \nNAS, IC, and ND codes from the analysis, as these are considered low confidence annotation \ncodes. To identify enriched Gene Ontology (GO) terms at the Biological Process level, we used \nthe package clusterProfiler v4.16.0 (G. Yu, 2024), more specifically the function \nclusterProfiler::enrichGO set to a q-value cutoff of 0.05 and to only capture “Biological Process” \nterms. To reduce redundancy, we used the clusterProfiler::simplify function with default \nparameters. We used the package enrichplot v1.28.4 (Guangchuang Yu, 2018) to plot the \nenriched GO terms. Our background gene set was all Drosophila melanogaster genes, which was \nprovided by the package.  \n15 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nRESULTS \nPhenotypes \nAcross all phenotypic assays, O-type flies consistently differed in trait values relative to the \nB-type controls after 20 generations (and after 10 generations, as shown in Supplementary Figure \n1). They displayed longer lifespans, delayed development, higher fecundity, and increased \nresistance to starvation, desiccation, and fungal infection. These patterns indicate a broad \nphenotypic shift associated with selection to extended generation times. All analyses were \nconducted at the population level (N = 20 populations per model). \nMortality rate & longevity \nO-type flies had significantly lower age-specific mortality rates (Figure 2A, Supplementary \nFigure 2A) than B-type flies in both sexes (Cox Proportional-Hazards test, P<0.001). Male flies \nhad a higher hazard rate of death (Cox Proportional-Hazards test, P<0.001). Consistent with this \npattern, mean population longevity (Figure 2B, Supplementary Figure 2B) was estimated to be \nhigher in O-type flies when compared to B-type flies (linear model, P<0.001), and lower in male \nflies when compared to female flies (linear model, P<0.001). Data for an intermediate time point \nare available as Supplementary Figure 1. \nDevelopment time \nPopulation development time (Figure 2C, Supplementary Figure 2C) was significantly longer in \nO-type flies compared to B-type flies (linear model, P<0.001). However, this difference was \nsmaller in O-type flies with BO ancestry (linear model interaction term, P=0.005). Intermediate \ngeneration data are available as Supplementary Figure 1. \n16 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nFecundity \nO-type populations exhibited higher mean population fecundity (Figure 2D, Supplementary \nFigure 2D) at all ages compared to B-type populations (Poisson generalized linear model, \nP<0.001). Also, we observed a weak significant effect of ancestry for the fecundity phenotype, \nsuch that mean population fecundity in the BO-derived populations was estimated to be higher \nthan in the B-derived populations, regardless of selection regime (Poisson generalized linear \nmodel, P<0.058). Age-specific fecundity trajectories are displayed in Supplementary Figure 3.  \nBody weight \nO-type flies had higher estimated mean population weight than B-type flies (linear model, \nP<0.001, Figure 2E, Supplementary Figure 2E), and female flies were estimated to have higher \nmean population weight than male flies (linear model, P<0.001). Dry weight and water loss data \nare reported in Walsh, 2022. \nStarvation Resistance and Desiccation Resistance \nO-type flies showed increased resistance to starvation and desiccation, with a higher proportion \nof flies surviving under both stresses compared to B-type flies. O-type flies had higher mean \npopulation starvation resistance (Figure 2F, Supplementary Figure 2F) than B-type flies (linear \nmodel, P<0.001). Also, females were estimated to have higher mean population starvation \nresistance than males under starvation conditions (linear model, P<0.001). For desiccation \nresistance (Figure 2G, Supplementary Figure 2G), O-type flies were estimated to have higher \nmean population desiccation resistance than B-type flies (linear model, P<0.001) and females \nwere estimated to have higher mean population desiccation resistance than males (linear model, \nP<0.001). \n17 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nImmune defense \nO-type flies showed a higher probability of surviving Beauvaria bassania infection compared \nwith B-type flies (Figure 2H, Supplementary Figure 2H). O-type flies were estimated to have a \nreduced chance of dying if compared to B-type populations, regardless of infection status or sex \n(Binomial generalized linear model, P<0.001). Infection predictably reduced survival, with \nInfected flies having higher odds of dying when compared to non-infected Control flies \n(Binomial generalized linear model, P<0.001), and O-type selection regime being estimated to \nhave increased survival in the infected flies (Binomial generalized linear model interaction term, \nP=0.038), indicating that O-type flies experienced less infection-induced mortality than B-type \nflies. \nGenomics \nAfter filtering steps, the resulting dataset consisted of 1,086,149 biallelic SNPs across the five \nmajor chromosome arms. The mean sequencing genome-wide coverage ranged from 56X to \n183X, with a mean of 106X across all samples. Coverage was reasonably consistent across the \ngenome, with a coefficient of variation ranging from 0.20 to 0.30.  \nPCA analysis showed the landscape of allele frequencies across all populations (Figure 3). We \nobserved patterns of convergence based on selection treatment: O-type populations clustered \ntogether regardless of ancestry, based on similar patterns of genetic variation after 20 \ngenerations. Populations at generation 1 clustered according to ancestry. \nAs the PCA analysis indicated convergence between OBO1-5 and OB1-5, we performed CMH tests \ncomparing the allele frequencies between generations 01 and 20 of all the O-type populations \ncombined (referring to them as O1-10). Benjamini-Hochberg FDR corrected, log-transformed \n18 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \np-values were plotted in Figure 4A. We identified a few dozen defined peaks evenly distributed \nacross all 4 main chromosome arms and X, with 1,218 statistically significant SNPs, located \nwithin 290 unique genes (240 of which have documented annotations). The annotated list of \ncandidate SNPs is available in Supplementary Table 2, and the list of candidate genes is available \nin Supplementary Table 3. When analyzed individually, CMH test results considering only the \nOBO and OB populations largely implicated the same regions, but with highly reduced statistical \nsignificance (Figure 4, panels B and C). When comparing B1-10 between generations 01 and 55, \nwe observe three peaks that are located in unannotated genomic regions (Supplementary Figure \n4). These peaks are not shared with the other comparisons. When comparing O1-10 at generation \n20 with B1-10 at generation 56, we observe differentiation in similar regions as when comparing \nO1-10 across generations, albeit with less statistical support (Supplementary Figure 5). \nOut of the 290 candidate genes, 240 genes have annotated GO terms, 214 of which are \nBiological Process terms. A GO term enrichment analysis suggested an enrichment for 42 terms \n(Supplementary Table 4) over the 290 unique genes implicated by our candidate SNP list \n(Supplementary Table 2). Removing redundant terms (see Methods for more details) returns a \nconsolidated list of 21 terms (Figure 5). The terms GO:0048667 (cell morphogenesis involved in \nneuron differentiation), GO:0048812 (neuron projection morphogenesis), GO:0120039 (plasma \nmembrane bounded cell projection morphogenesis), GO:0048858 (cell projection \nmorphogenesis), and GO:0031175 (neuron projection development) had the lowest q-values (< \n0.0001) and were enriched in 26 of the 214 considered genes. Figure 5 shows the simplified \nenriched GO terms with adjusted p-values and ratios. Though we observed no GO term \nenrichment for any longevity-related term, six genes on the list have been previously associated \nwith the “determination of adult lifespan” (GO:0008340) GO term (or any child term): Adcy1 or \n19 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nrut (Tong et al., 2007), Egfr (Kamakura, 2011), mthl5 (Araújo et al., 2013), Gnmt (Obata & \nMiura, 2015), Lkb1 (Funakoshi et al., 2011), and puc (Libert et al., 2008; M. C. Wang et al., \n2003). Similarly, we observe 9 genes previously associated with the “Immune response” \n(GO:0006955) (or any child) GO term: snk (Irving et al., 2001), dia (Howell et al., 2012), Atf3 \n(Rynes et al., 2012), Spn88Ea (Ahmad et al., 2009), ths (Dragojlovic-Munther & \nMartinez-Agosto, 2013), Gbp1 (Tsuzuki et al., 2012), tefu (Petersen et al., 2012), Elp6 (Howell \net al., 2012), puc (Libert et al., 2008), Antp (Mandal et al., 2007), and sick (Foley & O’Farrell, \n2004).  \n20 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nDISCUSSION \nPhenotypic consequences of selection for postponed reproduction \nUnder O-type selection, all populations consistently evolved extended longevity along with \nseveral correlated traits: increased development time, fecundity, immune defense, body weight, \nand resistance to starvation and desiccation. These phenotypic changes align with expectations \nfrom prior experimental evolution work (cf. Rose et al., 2004) and confirm that postponed \nreproduction imposes strong selection on life-history traits in the DEEP system.  \nNotably, O-type flies exhibited higher fecundity at all ages compared to B-type flies \n(Supplementary Figure 3). This contrasts with earlier findings of a trade-off between longevity \nand early-life reproduction (Rose, 1984). This discrepancy may reflect experimental differences, \nsuch as dietary yeast supplementation, which has been shown to mitigate early fecundity costs \n(Chippindale et al., 1993). O-type flies also survived longer than B-type flies under starvation \nand desiccation stress. As body weight influences the response to these stressors (Djawdan et al., \n1998), the observed increased weight in O-type flies may underlie their enhanced resistance. \nThis increase in weight may have resulted from prolonged larval feeding (associated with the \nincrease in development time in O-type populations). However, since we only measured time to \neclosion, we cannot determine at which developmental stage the B- and O-type flies diverge. \nFollow-up experiments dissecting the developmental timeline in finer detail could clarify \nwhether differences stem from the larval or pupal period. \nOverall, phenotypic evolution under O-type selection led to expected increases in organismal \nrobustness, with no detectable trade-offs aside from increased development time, which could \nincur ecological costs under natural conditions. \n21 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nAbsence of trade-offs between immune defense and longevity \nTrade-offs between fecundity and immune defense are well-documented in insects (Schwenke et \nal., 2016), suggesting that a physiological investment in reproduction can weaken immune \ndefense, while immune activation reduces fecundity. Relating immune defense to longevity, \nShahrestani et al. (2021) reported that Drosophila populations selected for improved defense \nagainst Beauveria bassiana exhibited higher post-infection survival but reduced lifespan under \nuninfected conditions. Thus, reduced immune defense in O-type populations would be consistent \nwith a trade-off hypothesis; yet, we observed consistently improved measures of immune defense \nacross all O-type populations. Our observations instead align with studies showing positive \nassociations between longevity and immune performance. Fabian et al. (2018) reported that \npopulations with experimentally-evolved extended lifespan also exhibited improved immune \ndefense against multiple pathogens, including B. bassiana. Bagheri et al. (2025) similarly found \nno evidence of trade-offs between longevity and immune defense in other DEEP populations. \nTogether, these observations contribute to the ongoing narrative that correlated traits don’t \nalways respond similarly when one or the other is selected upon directly (reviewed by Burke and \nRose 2009), and they also suggest that correlations between immune defense and longevity \ndepend on the nature of the immune strategies favored by selection.  \nRecent conceptual work emphasizes that immune defense is not a single trait, but rather a suite \nof strategies that differ in their physiological costs and evolutionary consequences (e.g. Troha \nand Ayres 2022). Antagonistic immune strategies, such as chronic or constitutive immune \nactivation, can impose substantial costs through immunopathology or energetic expenditure, \nwhereas cooperative or condition-dependent immune strategies scale with organismal condition \nand may enhance resistance or tolerance with fewer trade-offs. Selection for postponed \n22 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nreproduction may favor the latter, particularly if extended lifespan increases cumulative exposure \nto pathogens in crowded cage environments.  \nIn addition, O-type populations show increased adult body weight, presumably due to increased \nlarval feeding over their extended development time, which may further contribute to enhanced \nimmune defense. Mounting an immune response is an energetically costly process that remodels \nmetabolism to inhibit nutrient storage and catalyze fat stores (DiAngelo et al., 2009). In \nDrosophila, the fat body functions as a central immune and metabolic organ, integrating nutrient \nstorage with systemic immune signaling (reviewed by Yu et al. 2022). Because fat body mass \nand lipid reserves scale with overall body condition, variation in body size or energetic state \nshould influence immune performance. Consistent with this framework, larger and more robust \nO-type flies may show enhanced immune performance because immune defense scales with \noverall condition rather than being constrained by zero-sum allocation trade-offs. However, \nbecause immune defense was measured at a single age (14 days from egg), the relationship \nbetween body condition, immune defense, and longevity across the lifespan remains unresolved. \nFuture experiments assaying immune defense across multiple ages, multiple pathogens or \ninfection intensities will be necessary to more comprehensively examine the impact of evolved \nlongevity on age-specific immune competence. \nEffects of ancestry on phenotypic evolution \nAlthough selection consistently drove robust phenotypic divergence between O-type and B-type \npopulations, we detected a few ancestry-associated differences when comparing BO-derived \n(OBO1-5) and B-derived (OB1-5) lines within treatments (Supplementary Figure 2). This finding \ngenerally aligns with previous observations that populations derived from this experimental \n23 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nsystem rapidly converged at both phenotypic (Burke et al., 2016) and genomic levels (Graves et \nal., 2017), suggesting that strong directional selection can quickly overwhelm consequences of \nevolutionary history. Our statistical models only weakly support ancestry as being a significant \nfactor in a limited set of comparisons. The presence of ancestry effects across multiple \nlife-history traits suggests that convergence under experimental evolution regimes is not \nuniformly rapid, but instead may proceed at different rates depending on the trait and the \nevolutionary context. Among these traits, fecundity stands out as the most consistently affected \nand therefore provides a useful point of comparison to prior work. \nBurke et al. (2016) found that early-life fecundity did not fully converge between selection \ntreatments that differed in long-term, but not recent evolutionary history. Specifically, early \nfecundity of “ACO” populations did not fully converge with “AO” populations after the latter \nhad transitioned from a 70-day to a 9-day generation cycle over ~160 generations. Given that our \nO-type populations have experienced substantially fewer generations of selection, the persistence \nof ancestry-associated fecundity differences, and by extension, differences in correlated traits, is \nconsistent with a lag in convergence rather than a failure of parallel evolution. In contrast, Burke \net al. (2016) reported full convergence of early fecundity between B-type treatments that differed \nin their long-term, but not recent evolutionary history (the same B and BO lines used to initiate \nthe nB and nBO lines here). The observation of fecundity differences between nB and nBO \npopulations in the present study is therefore less intuitive. One potential explanation is that \nBurke et al. (2016) quantified early fecundity (ages 11-14 days from egg), whereas we measured \nlifetime fecundity. It is possible that lifetime reproductive output, or age-specific fecundity later \nin life, did not fully converge in the earlier experiment, and that effect has continued to persist. \nAdditionally, in the present study, nB populations were transitioned to cage environments prior \n24 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nto the onset of the experiment, whereas nBO populations had long been maintained in cages. \nSuch environmental differences could plausibly influence reproductive schedules, potentially \nreintroducing detectable phenotypic differences even among populations that had previously \nconverged under earlier assay conditions. \nTaken together, these results do not contradict the broader conclusion of Burke et al. (2016) that \nstrong selection rapidly drives phenotypic convergence. Rather, they indicate that convergence \nmay be trait-specific and/or age-specific, and likely shaped by the strength of selection imposed \nby the reproductive regime. Under the more extreme O-type conditions, which impose strong \nselection by restricting reproduction to late life, populations have rapidly diverged from the \nB-type phenotype but may require additional generations to fully converge with one another. In \ncontrast, under the more permissive B-type conditions, weaker selection may allow residual \nancestry effects or context-dependent differences to persist. These patterns are also reflected in \nthe genomic comparisons among populations as described below. \nGenomic consequences of selection \nExperimental evolution for postponed reproduction produced widespread and consistent genomic \nchanges in outbred Drosophila populations, consistent with a polygenic response. Despite clear \nphenotypic divergence in longevity and immune defense, our genomic analysis did not reveal \noverrepresentation of genes traditionally associated with those life-history traits. Instead, the \nstrongest signals emerging from allele frequency change over 20 generations of O-type selection \nwere found in genes associated with GO-terms related to development, and particularly neural \ndevelopment (Figure 5).  \nOur list of candidate genes does not include many canonical “aging loci” identified through \n25 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nmutant screens, such as mth (Petrosyan et al., 2014), Indy (P.-Y . Wang et al., 2009), chico (Slack \net al., 2015), mTor (Kapahi et al., 2004), etc. However, we found seven genes previously \nassociated with “determination of adult lifespan”, and nine genes previously associated with \n“immune response”. Among these, only one (growth-blocking peptide Gbp1) is involved with \nantimicrobial peptides (AMPs), a group of molecules critical in the innate immune response. \nUnlike studies based on inbred lines and de novo mutations, our experimental system applies \nlong-term selection to genetically diverse populations, allowing polygenic adaptation via subtle \nshifts in allele frequency across many loci. In this context, our observed enrichment in Biological \nProcess terms suggests that selection may have favored genes involved in maintaining \nphysiological homeostasis and somatic maintenance, consistent with generalized robustness \nphenotypes rather than targeted aging pathways.  \nComparing our gene list to those of other investigators working with long-lived lines \nexperimentally evolved for increased longevity reveals relatively little overlap among our \nstudies. Carnes et al. (2015) identified 99 candidate genes in females, eight 8.08%) of which \nmatch our candidate genes (Amy-d, Gem3, ko, Ccdc85, CG11400, Gbp2, CG45263, and \nasRNA:cr45272). Given that this study used populations derived from the same founding Ives \npopulation, we might have expected even more overlap. Gbp2 is a notable gene on this list; the \nGBP pathway is suggested as an important pathway in mediating acute innate immune reactions \n(Tsuzuki et al., 2012), and Gbp2 regulates Gbp1 expression (Ono et al., 2024). In a study with \npopulations that come from a completely independent study system, Fabian et al. (2018) \nidentified 868 candidate genes, with twenty (2.18%) also being present in our list (Ace, Atf3, \nCG13280, CG1677, CG33203, CG33459, CG34354, CG42673, comm3, DIP-gamma, kirre, Lar, \nMICU3, myd, Nepl19, path, sqz, ths, Ugt37D1, and wat). No genes are simultaneously reported \n26 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nas candidates by the three studies, highlighting the challenge of drawing connections across \nsimilar experiments at the gene level. Differences in the methods used to assess significance and \ndefine candidate gene lists likely contributed to this limited overlap. \nRole of replication and ancestry in genomic interpretation \nOne strength of our design is 10-fold replication among O-type populations, which enables us to \ndetect consistent selection responses despite the modest effect sizes typical of polygenic traits. \nPrevious studies with fewer replicates may have lacked the power to identify these subtle, \nrepeatable patterns, which may help explain why candidate genes often differ across \nexperiments. \nCompared to what we observed among phenotypes, ancestry played an even subtler role in \nshaping the genomic response. Comparisons between OBO1–5 and OB1–5 revealed minimal \ndivergence, reinforcing the idea that strong selection can drive parallel genomic changes across \ndistinct ancestral backgrounds. This convergence at the genomic level suggests that key targets \nof selection are shared across genetic backgrounds, despite variable phenotypic expression. \nWe also compared genomic signals across the B-type populations through generations \n(Supplementary Figure 4). While we observed negligible differentiation between the nB1-5 \npopulations over time, three weak peaks emerged from comparing the nBO1-5 populations \ninitially and after 55 generations. These three peaks persist when grouping all 10 B-type \npopulations together and comparing them over time, though they become non-significant with \nthe increased replication. While these three noncoding regions could implicate effects of \ndomestication and ongoing selection in the nBO lines, they may also reflect differences in \nmapping quality. This further emphasizes the importance of biological replication in the genomic \n27 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \ninterpretation of selection experiments. \nWe also compared signals of genomic differentiation between O1-10 at generation 20 with B1-10 at \ngeneration 56 (Supplementary Figure 5). This analysis implicated broadly similar genomic \nregions to those identified in comparisons among O1-10 across generations, though with reduced \nstatistical power. The weaker signal is consistent with the expectation that control populations \nare themselves evolving under laboratory conditions, including responses to domestication and \nhusbandry that are independent of the O-type life-history regime. As a result, contrasts that rely \nexclusively on contemporary control populations may incorporate genomic change unrelated to \nthe focal selective pressure, potentially obscuring loci most relevant to adaptation to O-type \nselection. This interpretation is consistent with prior results demonstrating substantial \nevolutionary change in laboratory control populations (e.g. Phillips et al. 2016) and highlights \nthe value of longitudinal sampling in Evolve & Resequence experimental design.  \nConclusions and future directions \nOur study demonstrates that selection for postponed reproduction consistently drives both \nphenotypic and genomic changes in D. melanogaster, with broad convergence across replicate \nlines. Diverged phenotypes largely recapitulate those observed in prior evolution experiments, \nthough our method of assaying immune defense is novel in this study system. Our observation \nthat immune defense is enhanced in O-type populations is therefore noteworthy and invites \ndeeper lines of inquiry. Links between longevity and innate immunity are well-established, and \nthe DEEP population resource provides a tantalizing opportunity to explore the connections \nbetween these two complex traits. While our genomic results do not directly implicate genes in \nknown Drosophila immune pathways, future work, particularly the integration of gene \n28 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nexpression and other -omics data types, should enrich our understanding of these traits in outbred \npopulations.  \n29 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nData availability \nThe data underlying this article are available in the Dryad Digital Repository, at \nhttps://doi.org/10.5061/dryad.mcvdnckf5. Data analysis R scripts are available at \nhttps://github.com/gacrestani/Gamboa-Santarosa_et_al. Raw sequencing data is available at SRA \nPRJNA1392952. \nAuthor contributions \nKaren Alma Gamboa-Santarosa: Formal analysis, Investigation, Validation, Writing – original \ndraft. \nGiovanni A. Crestani: Formal analysis, Investigation, Validation, Visualization, Writing – \noriginal draft.\n \nAlejandro Moran: Investigation. \nDolly Modha: Investigation. \nHannah S. Dugo: Investigation. \nMansour Abdoli: Formal analysis. \nMolly K. Burke: Conceptualization, Data curation, Funding acquisition, Methodology, Project \nadministration, Resources, Supervision, Validation, Writing – review & editing.\n \nParvin Shahrestani: Conceptualization, Data curation, Funding acquisition, Methodology, \nProject administration, Resources, Supevision, Validation, Writing – review & editing. \nFunding \nThis work was supported by National Institute of Health grant R35GM147402 to Molly K. \nBurke and National Institute of Health grant R15GM147869 to Parvin Shahrestani. \nConflict of Interest \n30 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nThe authors declare no conflict of interest. \nAcknowledgements \nWe thank Dr. Michael Rose (UC Irvine) for generously providing the ancestral populations that \nwere used to found this experiment. We also thank Avani Modha from Dr. Parvin Shahrestani’s \n(CSUF) laboratory for her notable contributions as well as the entire Longevity Selection Team. \nWe declare the usage of ChatGPT (models GPT-4 and GPT-3.5) and Google Gemini (models 2.0 \nFlash and 3.0 Pro) for initial stages of data analysis and figure generation to streamline R coding \n(subsequently reviewed and verified before usage) and sentence-level editing of the manuscript’s \ntext.  \n31 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nFigure 1. Schematic of the evolutionary history of experimental populations under O-type \nand B-type selection. Populations represented by blue lines experienced O-type selection \nwhereas populations represented by red lines experienced B-type selection. Dashed gray lines \nindicate periods of population continuity not relevant to the current study. The splitting of \npopulations (e.g. O1-5 into BO1-5, and B1-5 & BO1-5 into the final 20 populations) happened \ninstantaneously and therefore violates the y-axis scale. \nALT TEXT: Phylogeny of the twenty populations used in the study, from 1984 to 2024, with \n32 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \ncolor showing selection regimen.  \n33 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nFigure 2. B-type and O-type selection regimes lead to differentiation in life-history and \nstress-resistance phenotypes. Across all panels, phenotypes measured in the B-type populations \n34 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n(nBO1-5 combined with nB1-5) are shown in red, and phenotypes measured in the O-type \npopulations (OBO1-5 combined with OB1-5) are shown in blue. Each boxplot in panels B-H is \nderived from 10 data points (n=10 for all groups), with each data point representing a single \nvalue per replicate population per sex. Female and male data are shown separately where \napplicable. Statistical significance of the effect of selection regime was assessed using t-tests on \npopulation mean values, comparing B-type and O-type regimes. A Age-specific mortality rate \nover time; each line represents one replicate population. B Mean longevity (days). C Mean \ndevelopment time (hours). D Mean lifetime fecundity (average number of eggs per female). E \nMean bodyweight (mg). F Mean starvation resistance (survival time in hours). G Mean \ndesiccation resistance (survival time in hours). H Immune defense, measured as percent \nmortality following infection (death percentage); plots above the dashed line show infected files, \nwhile plots below the dashed line show uninfected control flies. \nALT TEXT: Graphs comparing phenotype of O-type flies and B-type flies, with statistical \nsignificance markers.  \n35 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nFigure 3. Genome-wide allele frequencies converge by selection regime rather than \nancestry after 20 generations of selection. Principal component analysis (PCA) of SNP allele \nfrequencies (1,086,149 filtered SNPs) across all populations shows that evolved populations \ncluster by selection regime rather than ancestry. A K-means clustering (k=3) of allele \nfrequencies. The top left cluster contains OBO1-5 populations at generation 1 and nBO1-5 \npopulations at generations 1 and 56, grouping all non-evolved populations derived from BO1-5. \nThe bottom left cluster contains OB1-5 at generation 1 and nB1-5 at generations 1 and 56, grouping \nall non-evolved populations derived from B1-5. The right cluster contains OBO1-5 at generation 20 \nand OB1-5 at generation 20, grouping all evolved experimental populations. B Principal \ncomponent 1 plotted by generation. By generation 20, experimental populations diverge from \ntheir ancestral and control populations, which remain clustered together, consistent with \nconvergence under shared selection regimes. \nALT TEXT: Scatterplot of first two principal components, with samples grouping by treatment \n36 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nand generation, and line plot showing principal component one by generations.  \n37 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nFigure 4. Combining replicate O-type populations increases power to detect evolved \ngenome-wide allele frequency change Manhattan plots show Cochran-Mantel-Haenszel (CMH) \ntest results based on scaled SNP frequencies; x-axis values represent cumulative physical \ndistance along the genome and y-axis values represent FDR-corrected p-values. The horizontal \nred line indicates the significance threshold (adjusted p-value = 10-100). A Combined analysis of \nthe ten O-type populations (OBO1-5 & OB1-5, relabeled here as O1-10) at generation 20 compared \nto generation 1. B OBO1-5 populations at generation 20 compared to generation 1. C OB1-5 \npopulations at generation 20 compared to generation 1. \n38 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nALT TEXT: Manhattan plots of O-type populations compared between generations 1 and 20, \nwith significance threshold indicating statistically significant single nucleotide polymorphisms.\n39 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nFigure 5. Enriched biological process GO terms among genes containing significant SNPs. \nThe y-axis lists enriched Biological Process GO terms. Point size corresponds to the number of \ngenes associated with each term, and point color indicates the Benjamini-Hochberg FDR \nadjusted p-value correction). The x-axis shows the gene ratio, defined as the number of genes \nassociated with a given GO term divided by the total number of genes analyzed. \n40 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nALT TEXT: Graph showing gene enriched ontology terms, with datapoint color depicting \nP-value and datapoint size depicting gene count.  \n41 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nSUPPLEMENTARY INFORMATION\n \nSupplementary Figure 1. Phenotypic divergence is evident after ~12 generations of O-type \nselection. Phenotypic data were collected at an intermediate timepoint, corresponding to \ngenerations 12 and 14 of O-type selection and approximately 19 generations of B-type selection. \nAcross all panels, data from the B-type regime (nBO1-5 combined with nB1-5) are shown in red, \nand data from the O-type regime (OBO1-5 combined with OB1-5) are shown in blue. Panels B and \nC, show boxplots derived from 10 data points (n=10 for all groups), with each data point \nrepresenting a sex-specific population mean from each replicate population. Data are shown \nseparately for males and females. A Instantaneous mortality rate over time; each line represents \none replicate population, with plots split by sex. B Mean population longevity (days). C Mean \ndevelopment time (hours). \n42 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nALT TEXT: Graphs of phenotypes measured at an intermediate generation time point, with \nstatistical comparisons.  \n43 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nSupplementary Figure 2. Phenotypic convergence across ancestral backgrounds is observed \nafter 20 generations of O-type selection. Across all panels, data from nBO1-5 and nB1-5 \ntreatments are shown in red, and data from OBO1-5 and OB1-5 treatments are shown in blue. \n44 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nPanels B-H, show boxplots derived from 5 data points per group (n=5 for all groups), with each \ndata point representing a sex-specific population mean per replicate population. Male and female \ndata are shown separately where applicable. Statistical differences between ancestries were \nassessed using t-tests on population means, with p-values reported above comparison brackets. A \nAge-specific mortality rate over time; each line represents one replicate population, with plots \nsplit by ancestry (B- or BO-derived) and sex. B Mean population longevity (days). C Mean \ndevelopment time (hours). D Mean lifetime fecundity (average number of eggs laid per female). \nE Mean body weight (mg). F Starvation resistance (survival in hours). G Desiccation resistance \n(survival in hours). H Immune defense, measured as percent mortality following infection (death \npercentage); plots above the dashed line show infected files, and plots below the dashed line \nshow uninfected control flies. With the exception of female development time in the OBO vs. \nOB populations (p=0.049) trait values do not significantly differ between ancestral backgrounds. \nALT TEXT: Graphs comparing phenotype of O-type flies and B-type separated by ancestry, \nwith statistical significance markers.  \n45 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nSupplementary Figure 3. Age specific fecundity trajectories show consistently higher \nreproductive output in O-type flies across the adult lifespan. Age-specific fecundity was \nmeasured at fine temporal resolution after approximately 20 generations of O-type selection. A \n46 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nThe average number of eggs laid per female is shown per replicate, separated by selection regime \n(B-type and O-type) and ancestry (B and BO). B Age-specific fecundity trajectories averaged \nacross replicates and ancestral backgrounds within each selection regime. \nALT TEXT: Scatterplots of age-specific fecundity split by ancestry and selection regimen, and \nsplit only by selection regimen.  \n47 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nSupplementary Figure 4. B-type populations show little genome-wide allele frequency \nchange over time, consistent with convergence. Manhattan plots show CMH test results based \non scaled SNP frequencies; x-axis values represent cumulative physical distance along the \ngenome and y-axis values represent FDR-corrected p-values. The horizontal red line indicates a \nsignificance threshold of adjusted p-value=10-100. A The combined ten B-type populations \n(nBO1-5 & nB1-5) B1-10 at generation 56 vs B1-10 at generation 1. B nBO1-5 at generation 56 vs \nnBO1-5 at generation 1. C nB1-5 at generation 56 vs nB1-5 at generation 1. \nALT TEXT: Manhattan plots of B-type populations compared between generations 1 and 56, \n48 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nwith significance threshold indicating statistically significant single nucleotide polymorphisms.\n49 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \n \nSupplementary Figure 5. Manhattan plot of adapted CMH tests on scaled allele depth (AD) \nand unfiltered depth (DP) of all SNPs. SNPs are represented as a function of genomic position \nand FDR-corrected p-value. The horizontal red line represents a significance threshold of \nadjusted p-value 10-100. A The combined ten O-type populations (OBO1-10 & OB1-10) at \ngeneration 20 vs the combined ten B-type populations (nBO1-5 & nB1-5) B1-10 at generation 56. B \nOBO1-5 at generation 20 vs nBO1-5 at generation 56. C OB1-5 at generation 20 vs nB1-5 at \ngeneration 56. \nALT TEXT: Manhattan plots of O-type populations compared to B-type populations, with \n50 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nsignificance threshold indicating statistically significant single nucleotide polymorphisms.\n51 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nSupplementary Table 1 - SNPs showing significant allele frequency change under O-type \nselection and their associated genes \nSupplementary Table 2 - Candidate genes underlying the response to O-type selection and their \nassociated GO terms \nSupplementary Table 3 - GO terms significantly enriched among candidate genes harboring \nsignificant SNPs \nSupplementary Table 4 - Statistical model outputs for phenotypic traits analyzed across selection \nregimes  \n52 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nReferences \nAhmad, S. T., Sweeney, S. T., Lee, J.-A., Sweeney, N. T., & Gao, F.-B. (2009). Genetic screen \nidentifies serpin5 as a regulator of the toll pathway and CHMP2B toxicity associated with \nfrontotemporal dementia. Proceedings of the National Academy of Sciences, 106(29), \n12168–12173. https://doi.org/10.1073/pnas.0903134106  \nAraújo, A. R., Reis, M., Rocha, H., Aguiar, B., Morales-Hojas, R., Macedo-Ribeiro, S., Fonseca, \nN. A., Reboiro-Jato, D., Reboiro-Jato, M., Fdez-Riverola, F., Vieira, C. P., & Vieira, J. \n(2013). The Drosophila melanogaster methuselah Gene: A Novel Gene with Ancient \nFunctions. PLOS ONE, 8(5), e63747. https://doi.org/10.1371/journal.pone.0063747  \nAuwera, G. van der, & O’Connor, B. D. (2020). Genomics in the cloud: Using Docker, GATK, \nand WDL in Terra (First edition). O’Reilly Media.  \nBagheri, E., Yin, H., Bengo, A. L., Rai, K. E., Conyers, T., Courville, R., Abdoli, M., Burke, M., \n& Shahrestani, P. (2025). Age-Dependent Immune Defense Against <em>Beauveria \nbassiana</em> in Long- and Short-Lived <em>Drosophila Populations</em>. MDPI \nAG. https://doi.org/10.20944/preprints202506.0507.v1  \nBaldwin-Brown, J. G., Long, A. D., & Thornton, K. R. (2014). The power to detect quantitative \ntrait loci using resequenced, experimentally evolved populations of diploid, sexual \norganisms. Molecular Biology and Evolution, 31(4), 1040–1055. \nhttps://doi.org/10.1093/molbev/msu048  \nBioconductor Core Team, B. P. M. O. [Cre. (2017). TxDb.Dmelanogaster.UCSC.dm6.ensGene \n[Computer software]. Bioconductor. \n53 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nhttps://doi.org/10.18129/B9.BIOC.TXDB.DMELANOGASTER.UCSC.DM6.ENSGENE  \nBurke, M. K., Barter, T. T., Cabral, L. G., Kezos, J. N., Phillips, M. A., Rutledge, G. A., Phung, \nK. H., Chen, R. H., Nguyen, H. D., Mueller, L. D., & Rose, M. R. (2016). Rapid \ndivergence and convergence of life-history in experimentally evolved Drosophila \nmelanogaster. Evolution, 70(9), 2085–2098. https://doi.org/10.1111/evo.13006  \nBurke, M. K., Dunham, J. P., Shahrestani, P., Thornton, K. R., Rose, M. R., & Long, A. D. \n(2010). Genome-wide analysis of a long-term evolution experiment with Drosophila. \nNature, 467(7315), 587–590. https://doi.org/10.1038/nature09352  \nBurke, M. K., King, E. G., Shahrestani, P., Rose, M. R., & Long, A. D. (2014). Genome-Wide \nAssociation Study of Extreme Longevity in Drosophila melanogaster. Genome Biology \nand Evolution, 6(1), 1–11. https://doi.org/10.1093/gbe/evt180  \nBurke, M. K., Liti, G., & Long, A. D. (2014). Standing Genetic Variation Drives Repeatable \nExperimental Evolution in Outcrossing Populations of Saccharomyces cerevisiae. \nMolecular Biology and Evolution, 31(12), 3228–3239. \nhttps://doi.org/10.1093/molbev/msu256  \nBurke, M. K., & Rose, M. R. (2009). Experimental evolution with Drosophila. American Journal \nof Physiology-Regulatory, Integrative and Comparative Physiology, 296(6), \nR1847–R1854. https://doi.org/10.1152/ajpregu.90551.2008  \nCarlson, M. (2017). Org.Dm.eg.db [Computer software]. Bioconductor. \nhttps://doi.org/10.18129/B9.BIOC.ORG.DM.EG.DB  \n54 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nCarnes, M. U., Campbell, T., Huang, W., Butler, D. G., Carbone, M. A., Duncan, L. H., \nHarbajan, S. V ., King, E. M., Peterson, K. R., Weitzel, A., Zhou, S., & Mackay, T. F. C. \n(2015). The Genomic Basis of Postponed Senescence in Drosophila melanogaster. PLOS \nONE, 10(9), e0138569. https://doi.org/10.1371/journal.pone.0138569  \nChippindale, A. K., Leroi, A. M., Kim, S. B., & Rose, M. R. (1993). Phenotypic plasticity and \nselection in Drosophila life-history evolution. I. Nutrition and the cost of reproduction. \nJournal of Evolutionary Biology, 6(2), 171–193. \nhttps://doi.org/10.1046/j.1420-9101.1993.6020171.x  \nCox, D. R. (1972). Regression Models and Life-Tables. Journal of the Royal Statistical Society: \nSeries B (Methodological), 34(2), 187–202. \nhttps://doi.org/10.1111/j.2517-6161.1972.tb00899.x  \nDi Tommaso, P., Chatzou, M., Floden, E. W., Barja, P. P., Palumbo, E., & Notredame, C. (2017). \nNextflow enables reproducible computational workflows. Nature Biotechnology, 35(4), \n316–319. https://doi.org/10.1038/nbt.3820  \nDiAngelo, J. R., Bland, M. L., Bambina, S., Cherry, S., & Birnbaum, M. J. (2009). The immune \nresponse attenuates growth and nutrient storage in Drosophila by reducing insulin \nsignaling. Proceedings of the National Academy of Sciences, 106(49), 20853–20858. \nhttps://doi.org/10.1073/pnas.0906749106  \nDjawdan, M., Chippindale, A. K., Rose, M. R., & Bradley, T. J. (1998). Metabolic Reserves and \nEvolved Stress Resistance in Drosophila melanogaster. Physiological Zoology. (world). \nhttps://doi.org/10.1086/515963  \n55 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nDragojlovic-Munther, M., & Martinez-Agosto, J. A. (2013). Extracellular matrix-modulated \nHeartless signaling in Drosophila blood progenitors regulates their differentiation via a \nRas/ETS/FOG pathway and target of rapamycin function. Developmental Biology, \n384(2), 313–330. https://doi.org/10.1016/j.ydbio.2013.04.004  \nDurinck, S., Spellman, P. T., Birney, E., & Huber, W. (2009). Mapping identifiers for the \nintegration of genomic datasets with the R/Bioconductor package biomaRt. Nature \nProtocols, 4(8), 1184–1191. https://doi.org/10.1038/nprot.2009.97  \nFabian, D. K., Garschall, K., Klepsatel, P., Santos-Matos, G., Sucena, É., Kapun, M., Lemaitre, \nB., Schlötterer, C., Arking, R., & Flatt, T. (2018). Evolution of longevity improves \nimmunity in Drosophila. Evolution Letters, 2(6), 567–579. \nhttps://doi.org/10.1002/evl3.89  \nFoley, E., & O’Farrell, P. H. (2004). Functional Dissection of an Innate Immune Response by a \nGenome-Wide RNAi Screen. PLOS Biology, 2(8), e203. \nhttps://doi.org/10.1371/journal.pbio.0020203  \nFranceschi, C., Bonafè, M., Valensin, S., Olivieri, F., De Luca, M., Ottaviani, E., & De \nBenedictis, G. (2000). Inflamm-aging. An evolutionary perspective on \nimmunosenescence. Annals of the New York Academy of Sciences, 908, 244–254. \nhttps://doi.org/10.1111/j.1749-6632.2000.tb06651.x  \nFunakoshi, M., Tsuda, M., Muramatsu, K., Hatsuda, H., Morishita, S., & Aigaki, T. (2011). A \ngain-of-function screen identifies wdb and lkb1 as lifespan-extending genes in \nDrosophila. Biochemical and Biophysical Research Communications, 405(4), 667–672. \n56 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nhttps://doi.org/10.1016/j.bbrc.2011.01.090  \nGraves, J. L., Jr, Hertweck, K. L., Phillips, M. A., Han, M. V ., Cabral, L. G., Barter, T. T., Greer, \nL. F., Burke, M. K., Mueller, L. D., & Rose, M. R. (2017). Genomics of Parallel \nExperimental Evolution in Drosophila. Molecular Biology and Evolution, 34(4), \n831–842. https://doi.org/10.1093/molbev/msw282  \nGuangchuang Yu. (2018). Enrichplot [Computer software]. Bioconductor. \nhttps://doi.org/10.18129/B9.BIOC.ENRICHPLOT  \nHamilton, W. D. (1966). The moulding of senescence by natural selection. Journal of Theoretical \nBiology, 12(1), 12–45. https://doi.org/10.1016/0022-5193(66)90184-6  \nHervé Pagès, M. C. (2017). AnnotationDbi [Computer software]. Bioconductor. \nhttps://doi.org/10.18129/B9.BIOC.ANNOTATIONDBI  \nHoedjes, K. M., Van Den Heuvel, J., Kapun, M., Keller, L., Flatt, T., & Zwaan, B. J. (2019). \nDistinct genomic signals of lifespan and life history evolution in response to postponed \nreproduction and larval diet in Drosophila. Evolution Letters, 3(6), 598–609.  \nHowell, L., Sampson, C. J., Xavier, M. J., Bolukbasi, E., Heck, M. M. S., & Williams, M. J. \n(2012). A directed miniscreen for genes involved in the Drosophila anti-parasitoid \nimmune response. Immunogenetics, 64(2), 155–161. \nhttps://doi.org/10.1007/s00251-011-0571-3  \nIrving, P., Troxler, L., Heuer, T. S., Belvin, M., Kopczynski, C., Reichhart, J.-M., Hoffmann, J. \nA., & Hetru, C. (2001). A genome-wide analysis of immune responses in Drosophila. \n57 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nProceedings of the National Academy of Sciences, 98(26), 15119–15124. (world). \nhttps://doi.org/10.1073/pnas.261573998  \nIves, P. T. (1970). Further genetic studies of the south amherst population of Drosophila \nmelanogaster. Evolution, 24(3), 507–518. \nhttps://doi.org/10.1111/j.1558-5646.1970.tb01785.x  \nKamakura, M. (2011). Royalactin induces queen differentiation in honeybees. Nature, \n473(7348), 478–483. https://doi.org/10.1038/nature10093  \nKapahi, P., Zid, B. M., Harper, T., Koslover, D., Sapin, V ., & Benzer, S. (2004). Regulation of \nLifespan in Drosophila by Modulation of Genes in the TOR Signaling Pathway. Current \nBiology, 14(10), 885–890. https://doi.org/10.1016/j.cub.2004.03.059  \nKassambara, A. (2016). ggpubr: “ggplot2” Based Publication Ready Plots (p. 0.6.2) [Dataset]. \nhttps://doi.org/10.32614/CRAN.package.ggpubr  \nKassambara, A., Kosinski, M., Biecek, P., & Fabian, S. (2025). survminer: Drawing Survival \nCurves using “ggplot2” (Version 0.5.1) [Computer software]. \nhttps://cran.r-project.org/web/packages/survminer/index.html  \nKofler, R., & Schlötterer, C. (2014). A Guide for the Design of Evolve and Resequencing \nStudies. Molecular Biology and Evolution, 31(2), 474–483. \nhttps://doi.org/10.1093/molbev/mst221  \nLawrence, M., Huber, W., Pagès, H., Aboyoun, P., Carlson, M., Gentleman, R., Morgan, M. T., \n& Carey, V . J. (2013). Software for Computing and Annotating Genomic Ranges. PLoS \n58 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nComputational Biology, 9(8), e1003118. https://doi.org/10.1371/journal.pcbi.1003118  \nLibert, S., Chao, Y ., Zwiener, J., & Pletcher, S. D. (2008). Realized immune response is \nenhanced in long-lived puc and chico mutants but is unaffected by dietary restriction. \nMolecular Immunology, Special Section: Theories and Modelling of T Cell Behaviour, \n45(3), 810–817. https://doi.org/10.1016/j.molimm.2007.06.353  \nLong, A., Liti, G., Luptak, A., & Tenaillon, O. (2015). Elucidating the molecular architecture of \nadaptation via evolve and resequence experiments. Nature Reviews Genetics, 16(10), \nArticle 10. https://doi.org/10.1038/nrg3937  \nLuckinbill, L. S., Arking, R., Clare, M. J., Cirocco, W. C., & Buck, S. A. (1984). Selection for \ndelayed senescence in Drosophila melanogaster. Evolution; International Journal of \nOrganic Evolution, 38(5), 996–1003. https://doi.org/10.1111/j.1558-5646.1984.tb00369.x  \nMandal, L., Martinez-Agosto, J. A., Evans, C. J., Hartenstein, V ., & Banerjee, U. (2007). A \nHedgehog- and Antennapedia-dependent niche maintains Drosophila haematopoietic \nprecursors. Nature, 446(7133), 320–324. https://doi.org/10.1038/nature05585  \nMcHugh, K. M., & Burke, M. K. (2022). From microbes to mammals: The experimental \nevolution of aging and longevity across species. Evolution, 76(4), 692–707. \nhttps://doi.org/10.1111/evo.14442  \nMedawar, P. B. (1952). An unsolved problem of Biology.  \nObata, F., & Miura, M. (2015). Enhancing S-adenosyl-methionine catabolism extends \nDrosophila lifespan. Nature Communications, 6(1), 8332. \n59 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nhttps://doi.org/10.1038/ncomms9332  \nPartridge, L., & Fowler, K. (1992). DIRECT AND CORRELATED RESPONSES TO \nSELECTION ON AGE AT REPRODUCTION IN DROSOPHILA MELANOGASTER. \nEvolution; International Journal of Organic Evolution, 46(1), 76–91. \nhttps://doi.org/10.1111/j.1558-5646.1992.tb01986.x  \nPetersen, A. J., Rimkus, S. A., & Wassarman, D. A. (2012). ATM kinase inhibition in glial cells \nactivates the innate immune response and causes neurodegeneration in Drosophila. \nProceedings of the National Academy of Sciences, 109(11), E656–E664. \nhttps://doi.org/10.1073/pnas.1110470109  \nPetrosyan, A., Gonçalves, Ó. F., Hsieh, I.-H., & Saberi, K. (2014). Improved functional abilities \nof the life-extended Drosophila mutant Methuselah are reversed at old age to below \ncontrol levels. AGE, 36(1), 213–221. https://doi.org/10.1007/s11357-013-9568-1  \nPhillips, M. A., Kutch, I. C., Long, A. D., & Burke, M. K. (2020). Increased time sampling in an \nevolve-and-resequence experiment with outcrossing Saccharomyces cerevisiae reveals \nmultiple paths of adaptive change. Molecular Ecology, 29(24), 4898–4912. \nhttps://doi.org/10.1111/mec.15687  \nPoplin, R., Ruano-Rubio, V ., DePristo, M. A., Fennell, T. J., Carneiro, M. O., Auwera, G. A. V . \nder, Kling, D. E., Gauthier, L. D., Levy-Moonshine, A., Roazen, D., Shakir, K., Thibault, \nJ., Chandran, S., Whelan, C., Lek, M., Gabriel, S., Daly, M. J., Neale, B., MacArthur, D. \nG., & Banks, E. (2018). Scaling accurate genetic variant discovery to tens of thousands \nof samples (p. 201178). bioRxiv. https://doi.org/10.1101/201178  \n60 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nR Core Team. (2025). R: A Language and Environment for Statistical Computing. R Foundation \nfor Statistical Computing. https://www.R-project.org/  \nRemolina, S. C., Chang, P. L., Leips, J., Nuzhdin, S. V ., & Hughes, K. A. (2012). GENOMIC \nBASIS OF AGING AND LIFE-HISTORY EVOLUTION IN DROSOPHILA \nMELANOGASTER. Evolution, 66(11), 3390–3403. \nhttps://doi.org/10.1111/j.1558-5646.2012.01710.x  \nRose, M. R. (1984). Laboratory Evolution of Postponed Senescence in Drosophila melanogaster. \nEvolution, 38(5), 1004–1010. https://doi.org/10.2307/2408434  \nRose, M. R., Passananti, H. B., & Matos, M. (2004). Methuselah Flies: A Case Study in the \nEvolution of Aging. WORLD SCIENTIFIC. https://doi.org/10.1142/5457  \nRynes, J., Donohoe, C. D., Frommolt, P., Brodesser, S., Jindra, M., & Uhlirova, M. (2012). \nActivating Transcription Factor 3 Regulates Immune and Metabolic Homeostasis. \nMolecular and Cellular Biology, 32(19), 3949–3962. \nhttps://doi.org/10.1128/MCB.00429-12  \nSchlötterer, C., Tobler, R., Kofler, R., & Nolte, V . (2014). Sequencing pools of \nindividuals—Mining genome-wide polymorphism data without big funding. Nature \nReviews Genetics, 15(11), 749–763. https://doi.org/10.1038/nrg3803  \nSchwenke, R. A., Lazzaro, B. P., & Wolfner, M. F. (2016). Reproduction–Immunity Trade-Offs \nin Insects. Annual Review of Entomology, 61(1), 239–256. \nhttps://doi.org/10.1146/annurev-ento-010715-023924  \n61 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nShahrestani, P., King, E., Ramezan, R., Phillips, M., Riddle, M., Thornburg, M., Greenspan, Z., \nEstrella, Y ., Garcia, K., Chowdhury, P., Malarat, G., Zhu, M., Rottshaefer, S. M., Wraight, \nS., Griggs, M., Vandenberg, J., Long, A. D., Clark, A. G., & Lazzaro, B. P. (2021). The \nmolecular architecture of Drosophila melanogaster defense against Beauveria bassiana \nexplored through evolve and resequence and quantitative trait locus mapping. G3 \nGenes|Genomes|Genetics, 11(12), jkab324. https://doi.org/10.1093/g3journal/jkab324  \nSlack, C., Alic, N., Foley, A., Cabecinha, M., Hoddinott, M. P., & Partridge, L. (2015). The \nRas-Erk-ETS-Signaling Pathway Is a Drug Target for Longevity. Cell, 162(1), 72–83. \nhttps://doi.org/10.1016/j.cell.2015.06.023  \nSpitzer, K., Pelizzola, M., & Futschik, A. (2020). Modifying the Chi-square and the CMH test \nfor population genetic inference: Adapting to overdispersion. The Annals of Applied \nStatistics, 14(1), 202–220. https://doi.org/10.1214/19-AOAS1301  \nTherneau, T. M., until 2009), T. L. (original S.->R port and R. maintainer, Elizabeth, A., & \nCynthia, C. (2024). survival: Survival Analysis (Version 3.8-3) [Computer software]. \nhttps://cran.r-project.org/web/packages/survival/index.html  \nTong, J. J., Schriner, S. E., McCleary, D., Day, B. J., & Wallace, D. C. (2007). Life extension \nthrough neurofibromin mitochondrial regulation and antioxidant therapy for \nneurofibromatosis-1 in Drosophila melanogaster. Nature Genetics, 39(4), 476–485. \nhttps://doi.org/10.1038/ng2004  \nTroha, K., & Ayres, J. S. (2022). Cooperative defenses during enteropathogenic infection. \nCurrent Opinion in Microbiology, 65, 123–130. \n62 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nhttps://doi.org/10.1016/j.mib.2021.11.003  \nTsuzuki, S., Ochiai, M., Matsumoto, H., Kurata, S., Ohnishi, A., & Hayakawa, Y . (2012). \nDrosophila growth-blocking peptide-like factor mediates acute immune reactions during \ninfectious and non-infectious stress. Scientific Reports, 2(1), 210. \nhttps://doi.org/10.1038/srep00210  \nTurner, T. L., Stewart, A. D., Fields, A. T., Rice, W. R., & Tarone, A. M. (2011). \nPopulation-Based Resequencing of Experimentally Evolved Populations Reveals the \nGenetic Basis of Body Size Variation in Drosophila melanogaster. PLOS Genetics, 7(3), \ne1001336. https://doi.org/10.1371/journal.pgen.1001336  \nVasimuddin, Md., Misra, S., Li, H., & Aluru, S. (2019). Efficient Architecture-Aware \nAcceleration of BWA-MEM for Multicore Systems. 2019 IEEE International Parallel \nand Distributed Processing Symposium (IPDPS), 314–324. \nhttps://doi.org/10.1109/IPDPS.2019.00041  \nWalsh, A. K.-G. (2022). Experimental evolution for longevity differentiation in Drosophila \nmelanogaster. California State University, Fullerton.  \nWang, M. C., Bohmann, D., & Jasper, H. (2003). JNK Signaling Confers Tolerance to Oxidative \nStress and Extends Lifespan in Drosophila. Developmental Cell, 5(5), 811–816. \nhttps://doi.org/10.1016/S1534-5807(03)00323-X  \nWang, P.-Y ., Neretti, N., Whitaker, R., Hosier, S., Chang, C., Lu, D., Rogina, B., & Helfand, S. \nL. (2009). Long-lived Indy and calorie restriction interact to extend life span. \nProceedings of the National Academy of Sciences, 106(23), 9262–9267. \n63 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nhttps://doi.org/10.1073/pnas.0904115106  \nWickham, H. (with Sievert, C.). (2016). ggplot2: Elegant graphics for data analysis (Second \nedition). Springer international publishing.  \nWickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L., François, R., Grolemund, G., \nHayes, A., Henry, L., Hester, J., Kuhn, M., Pedersen, T., Miller, E., Bache, S., Müller, K., \nOoms, J., Robinson, D., Seidel, D., Spinu, V ., … Yutani, H. (2019). Welcome to the \nTidyverse. Journal of Open Source Software, 4(43), 1686. \nhttps://doi.org/10.21105/joss.01686  \nWickham, H., & Bryan, J. (2015). readxl: Read Excel Files (p. 1.4.5) [Dataset]. \nhttps://doi.org/10.32614/CRAN.package.readxl  \nWilliams, G. C. (1957). Pleiotropy, Natural Selection, and the Evolution of Senescence. \nEvolution, 11(4), 398–411. https://doi.org/10.1111/j.1558-5646.1957.tb02911.x  \nYu, G. (2024). Thirteen years of clusterProfiler. The Innovation, 5(6), 100722. \nhttps://doi.org/10.1016/j.xinn.2024.100722  \nYu, S., Luo, F., Xu, Y ., Zhang, Y ., & Jin, L. H. (2022). Drosophila Innate Immunity Involves \nMultiple Signaling Pathways and Coordinated Communication Between Different \nTissues. Frontiers in Immunology, 13. https://doi.org/10.3389/fimmu.2022.905370  \nZwaan, B., Bijlsma, R., & Hoekstra, R. F. (1995). ARTIFICIAL SELECTION FOR \nDEVELOPMENTAL TIME IN DROSOPHILA MELANOGASTER IN RELATION TO \nTHE EVOLUTION OF AGING: DIRECT AND CORRELATED RESPONSES. \n64 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint \n\n \nEvolution, 49(4), 635–648. https://doi.org/10.1111/j.1558-5646.1995.tb02300.x   \n65 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted February 27, 2026. ; https://doi.org/10.64898/2026.02.25.708047doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}