Gene expression changes in response to aging compared to heat stress, oxidative stress and ionizing radiation in Drosophila melanogaster

1Heat shock protein (Hsp) genes are induced in response to stresses that cause protein denaturation, through activation of the heat shock factor (HSF). Up-regulation of Hsp genes is also observed during normal aging. For example, both Hsp70 and Hsp22 are up-regulated during normal Drosophila aging, and this up-regulation requires functional HSF binding sites (Heat Shock Elements, or HSEs) in the promoters of these genes. Genome-wide studies of gene expression changes during Drosophila aging have revealed additional features of a stress response, including the up-regulation of additional oxidative stress-response genes, and the dramatic up-regulation of innate immune response genes. In addition, Drosophila aging is characterized by a small but across-the-board down-regulation of mitochondrial metabolism and electron transport chain (ETC) genes, and this pattern is also observed in aging mammalian tissues, and at early adult ages in both Drosophila and C. elegans, indicating a conservation of aging mechanisms across species. Both innate immune response genes and Hsp genes have been shown to be predictive biomarkers of individual animal life span when the gene promoters are fused to GFP to create transgenic reporters, thereby supporting the significance of the identified gene expression changes. Here normal aging was compared with multiple stressors to provide further insight into common and unique features.

2Micro-array analysis was used to identify genes whose expression was altered in response to normal aging, hyperoxia, hydrogen peroxide, ionizing radiation and heat stress. A core set of 18 stress-response genes were up-regulated ≥1.5-fold in response to each of the tested stresses as well as during normal aging (Table 1).

3These up-regulated genes included the heat shock protein genes Hsp70, Hsp83 (which is the single Drosophila Hsp90-class member), and the small Hsp gene l(2)efl. The up-regulation of Hsp70 and l(2)efl in response to selected stressors was confirmed using quantitative real-time PCR analysis (Figure 1), and in addition Hsp70 was analyzed by Northern blot analysis (Supplemental Figure S1; results summarized in Table 2). Also up-regulated by aging and each stressor were the glutathione S-transferase gene GstD2, the central metabolic regulatory enzyme gene Pepck, and the mitochondrial unfolded protein response (mUPR) gene ref(2)P. Down-regulated genes included several associated with sugar metabolism and proteolysis (Table 1).

4Each stress had gene expression changes that were unique to that stress (listed in Supplemental Table S1) and the enriched GO terms that uniquely characterize each stress are summarized (Table 3). Hyperoxia stress had no enriched GO terms in the uniquely up-regulated genes, and a single enriched GO term, Signal peptide processing (3 genes) among the down-regulated genes. In contrast, there were numerous up-regulated genes unique to hydrogen peroxide stress, and these up-regulated genes were enriched for many GO terms involved in developmental pathways, signaling pathways, and nucleobase metabolism (Table 3). Genes uniquely up-regulated upon heat stress included many of the Hsp60-class, and this list was consequently enriched for the GO term Protein folding (16 genes), whereas down-regulated genes unique to heat stress were enriched for the GO terms Defense response and Melanization defense response (Table 3). Finally, genes uniquely up-regulated in response to ionizing radiation included several proteasome subunit genes (Supplemental Table S1), and this gene list was enriched for the GO terms Protein catabolic process and Macromolecular catabolic process (Table 3), whereas there were no GO terms enriched among down-regulated genes.

5As described above, a core set of stress response genes was induced during aging and by each of the stressors tested. Aging shared additional changes in gene expression with each individual stressor (Supplemental Table S2), and was found to be more similar to the stresses most associated with oxidative stress (hyperoxia, hydrogen peroxide, ionizing radiation) than it was to heat stress, based on cluster analysis (Supplemental Figure S2) and by comparison of the GO categories that were enriched in the groups of up-regulated and down-regulated genes (Supplemental Table S3). While aging shared a significant overlap in up-regulated and down-regulated genes with each of the stresses, aging shared the greatest number of gene expression changes with hyperoxia (Table 4).

6A number of gene expression changes were found to be unique to aging. These included up-regulation of numerous innate immune response genes, and down-regulation of numerous mitochondrial metabolism genes, including ones encoding components of the ETC (Supplemental Table S1; enriched GO terms listed in Table 4). While up-regulation of innate immune response genes is a feature of aging that is shared with hyperoxia (Supplemental Table S3), the number of up-regulated innate immune response genes was significantly greater for aging, resulting in many changes in this category that were unique to aging. Also uniquely up-regulated during aging were the odorant receptor genes Obp56a and Obp57d.

7Down-regulation of mitochondrial genes is a feature of aging that is shared with hydrogen peroxide and ionizing radiation (Supplemental Table S3), but the number of down-regulated mitochondrial genes was greater for aging, resulting in many changes in this category that were unique to aging. Among these many down-regulated mitochondrial genes were ones encoding mitochondrial ribosomal proteins and components of the mitochondrial membrane protein translocases (TIM and TOM), as well as the mitochondrial form of superoxide dismutase (MnSOD or Sod2).

8Changes in gene expression caused by one or more stressors were confirmed by quantitative real-time PCR (Figure 1 and Supplemental Figure S3) and by Northern blot analysis (Supplemental Figure S1), and in general an excellent concordance was observed with the micro-array data and with the published literature (Summarized in Table 2). One exception was for expression of the innate immune response gene Drosomycin upon heat stress, which was observed to increase in the micro-array analysis, but to decrease in the qPCR analysis (Table 2). Because bacterial load and Drosomycin gene expression can vary significantly between different flies and vials of flies, we conclude that this discrepancy was most likely due to a small difference in bacterial load and Drosomycin gene expression in the control flies used for the qPCR analysis relative to the control flies used for micro-array analysis.

9Comparison of the responses to the different stresses reveals preferential induction of certain genes. For example, Hsp70 (Figure 1) and Hsp22 (Supplemental Figure S3) were induced to the greatest extent by heat stress, whereas l(2)efl (Figure 1) and ade3 (Supplemental Figure S3) were induced to a greater extent by ionizing radiation and hyperoxia. In addition significant sexual dimorphism in the magnitude of responses was observed. For example, the induction of Hsp70 (Figure 1) and Hsp22 (Supplemental Figure S3) in response to heat stress was greater in males than in females, and the induction of l(2)efl (Figure 1) and ade3 (Supplemental Figure S3) in response to ionizing radiation was greater in males than in females.

10TheGstD1 gene encodes a glutathione-S-transferase, and is induced in adult flies during normal aging and when flies are challenged with oxidative stress produced by hyperoxia and paraquat, and was also found to be up-regulated in response to hydrogen peroxide and ionizing radiation stress (Summarized in Table 2). The GstD1 promoter region contains consensus binding motifs for the stress-responsive transcription factors Nrf2 and Foxo (diagrammed in Supplemental Figure S4). A transgenic reporter has been characterized where the regulatory sequences of the GstD1 gene are fused to GFP, and the resulting GstD-GFP reporter is induced in the adult fly by feeding flies with the oxidative stressors paraquat, arsenic or hydrogen peroxide. A clustered point mutation was created to disrupt the antioxidant response element (ARE) in the GstD-GFP reporter to yield a mutant reporter called GstD-deltaARE-GFP (diagrammed in Supplemental Figure S4). These reporters have been used to demonstrate that the GstD-GFP transgene is positively regulated in the adult fly in response to genetically-altered Nrf2 expression, and in response to the cancer chemotherapeutic compound Oltipraz which is known to activate Nrf2, and these regulations required the intact ARE. Quantitative PCR analysis of adult flies indicated that induction of the GstD1 gene by paraquat is reduced in flies hemizygous for the JNKK gene hemipterous, suggesting additional positive regulation of GstD genes by the JNK pathway in response to oxidative stress. The JNK pathway activates the transcription factor Foxo suggesting that the JNK pathway may activate GstD gene expression through the Foxo binding motif located in the GstD gene promoter region(diagrammed in Supplemental Figure S4). A GstD1-LacZ reporter has been reported to be up-regulated during normal aging in the enteroendocrine cells (ECs) of the fly intestine, but to be reduced during aging in the intestinal stem cells (ISCs).

11Here the expression of the GstD-GFP and GstD-deltaARE-GFP reporters were examined in whole adult flies during normal aging. The GstD-GFP reporter was expressed at low levels in young flies, and exhibited robust induction throughout the body of the fly during aging, including the head, thorax, abdomen and legs (Figure 2A), consistent with the whole-body micro-array analyses presented above. Induction was apparent even at moderate ages (30 days; Figure 2A) and continued at high levels for the remainder of the life span (data not shown). In contrast, the GstD-deltaARE-GFP reporter was robustly induced in thorax and legs, particularly in flight muscle and leg muscle, whereas induction in the head and abdomen was either greatly reduced or absent. The GstD-deltaARE-GFP reporter was also observed to produce slightly more expression in young flies in the upper abdomen. The mean GFP intensity throughout the body was quantified from captured images of multiple flies using Image J software, and this analysis confirmed the up-regulation of both reporters during aging, in both males and females (Figure 2B). Despite the absence of GFP induction in head and abdomen tissue, the mean intensity of fluorescence produced by the GstD-deltaARE-GFP reporter throughout the fly was comparable to the un-mutated reporter (Figure 2B), as the expression in thorax and legs was relatively greater (see Figure 2A); and this difference may be due to some effect of the different chromosomal insertion sites on the overall expression levels for the reporters. Taken together, these data confirm the up-regulation of the GstD1 gene during aging in the majority of adult tissues, and suggest that efficient expression in the head and abdomen may require the consensus ARE motif (Diagrammed in Supplemental Figure S4).

12Changes in gene expression that were common to aging and all tested stresses identified a core set of stress response genes. Pepck encodes an enzyme critical in gluconeogenesis and glyceroneogenesis, and its up-regulation may be part of a basic metabolic adaptation to stress. Interestingly, in mice, over-expression of PEPCK specifically in muscle tissue increases movement, life span and muscle mitochondrial proliferation. Also among the core set of induced genes were Starvin, ref(2)P, and the Hsp genes Hsp70, Hsp83 and lefl. Starvin encodes a co-chaperone involved in autophagy and muscle maintenance and its up-regulation is consistent with its role in protein turnover and the cellular response to proteotoxicity. Similarly, ref(2)P encodes a component of the mitochondrial unfolded protein response (mUPR) pathway, consistent with proteotoxicity in the mito-chondrial compartment. Hsp genes are induced in response to protein denaturation and misfolding through activation of the HSF transcription factor, which in turn binds to HSEs in the Hsp gene promoters and activates transcription. In stressed cells certain Hsps have been shown to function to reduce proteotoxicity by favoring protein re-folding as well as the turnover of damaged proteins through the ubiquitin/proteasome and autophagy pathways. Induction of Drosophila Hsp genes during normal aging and upon oxidative stress has been shown to be dependent upon functional HSEs in the gene promoters, consistent with an increased abundance of misfolded proteins and conse-quent HSF activation under these conditions. The presence of Hsp genes in the set up-regulated by aging and each stressor indicates that protein denaturation/misfolding and HSF activation are common features of aging and each of the tested stresses. Induction of Hsps during aging may be part of a stress response that favors fly function by helping the fly to cope with aging-associated proteotoxicity. Consistent with this idea, increased expression of certain Hsps is associated with increased fly life span. In addition, it is also possible that chronic Hsp induction, particularly at late ages, may sometimes be mal-adaptive. Interestingly, 8/18 of the common up-regulated genes and 21/32 of the common down-regulated genes have uncharacterized functions, indicating that there is much yet to be learned about the core stress response.

13While the pattern of gene expression changes during aging shared features with each of the tested stresses, each stress also had unique features (Summarized in Table 3). For example, hydrogen peroxide stress caused up-regulation of numerous genes involved in developmental pathways and signaling pathways, consistent with the fact that hydrogen peroxide also normally functions as a signaling molecule, during development and in adults, in Drosophila and other metazoans. Genes uniquely up-regulated upon heat stress included ones of the Hsp60-class, which are important for protein trafficking and protein import into organelles. Down-regulated genes unique to heat stress were enriched for the GO terms Defense response and Melanization defense response, suggesting that responses to wounding and bacterial challenge may be impaired. Finally, genes uniquely up-regulated in response to ionizing radiation included several proteasome subunit genes, which may indicate a particular requirement for protein turnover, perhaps in response to protein backbone cleavage, or alternatively this might reflect the critical role of the proteasome in DNA repair.

14While aging shared features with each stress, aging was found to be more similar to the stresses most associated with oxidative stress (hyperoxia, hydrogen peroxide, ionizing radiation) than it was to heat stress. These observations are consistent with the conclusion that aging eukaryotic cells are in a pro-oxidant state associated with up-regulation of oxidative stress-response genes including ones encoding Gsts. In addition to the shared features, a number of gene expression changes were found to be unique to aging (Supplemental Table S1). For example, the gene encoding the mitochondrial form of superoxide dismutase (MnSOD or SOD2) was uniquely down-regulated during aging, and this is of potential interest given the fact that augmenting the expression ofMnSOD can favor life span in adult flies and in C. elegans. Also uniquely up-regulated during aging were the odorant receptor genes Obp56a and Obp57d, which is interesting in light of reports of negative effects of other odorant receptor genes on fly life span.

15The changes in gene expression that were found to be unique to aging included up-regulation of numerous innate immune response genes, and down-regulation of numerous mitochondrial metabolism genes, including ones encoding components of the ETC (Table 4). While up-regulation of innate immune response genes is a feature of aging that is shared with hyperoxia (Supplemental Table S3), the number of up-regulated innate immune response genes was significantly greater for aging, resulting in many changes in this category that were unique to aging. Similarly, down-regulation of mitochondrial and ETC genes is a feature of aging that is shared with ionizing radiation and hydrogen peroxide stress (Supplemental Table S3), but the number of down-regulated mitochondrial and ETC genes was greater for aging, resulting in many changes in this category that were unique to aging. Girardot et al examined gene expression changes during Drosophila aging separately for the head, thorax and abdomen, and found that down-regulation of mitochondrial genes is observed preferentially in the thorax; because the thorax is composed primarily of flight muscle this observation suggests that mitochondrial gene down-regulation may occur preferentially in muscle tissue.

16Up-regulation of innate immune response genes during Drosophila aging is in part due to a dramatic increase in microbial load during aging, as eliminating bacteria reduces the response. However, innate immune response genes are still up-regulated during aging in the absence of detectable microbes, suggesting additional mechanisms for activation of these genes during aging. Consistent with this conclusion, innate immune response genes are also up-regulated in response to oxidative stress caused by hyperoxia (; this study), and therefore one possibility is that an aging-related failure in mitochondrial maintenance leads to oxidative stress that can induce innate immune response gene expression. Similarly, studies in mammals reveal that damaged mitochondria also release DNA fragments and formyl-peptides that can induce innate immune response genes, and therefore this may be an additional mechanism for innate immune response gene induction during aging that is a consequence of a failure in mitochondrial maintenance. The across-the-board down-regulation of Drosophila mitochondrial genes, ETC genes and mitochondrial metabolism genes observed during aging suggests a possible mechanism for a failure in mitochondrial maintenance during aging (Diagrammed in Supplemental Figure S5). The ETC and mitochondria turn over at a basal rate, and more rapidly in response to signals such as starvation, and a reduced rate of replacement is expected to result in longer-lived structures that will be more susceptible to time-dependent damage and malfunction. This idea is consistent with the observed accumulation of structural-ly abnormal mitochondria during Drosophila aging, reduced mitochondrial transcription, decreased ATP and increased production of ROS. Decreased ATP flux is expected to reduce rates of bulk protein synthesis and turnover, and increased ROS will increase protein damage, consistent with the accumulation of damaged and misfolded proteins (proteotoxicity) and the induction of Hsp genes.

17Taken together, the data support a model wherein the down-regulation of mitochondrial and ETC genes during aging leads to a failure in mitochondrial maintenance and the accumulation of abnormal mitochondria, which in turns leads to oxidative stress and proteotoxicity; these stresses in turn cause the oxidative-stress-like and proteotoxic-stress-like patterns of gene expression observed during aging (Supplemental Figure S5). Placing oxidative stress down-stream of an aging-associated failure in mitochondrial maintenance is consistent with the observation that oxidative stress correlates with, but does not directly regulate life span in Drosophila, and with the implication of mitochondrial malfunction in mammalian aging-related metabolic disorders. Consistent with the importance of mitochondrial maintenance in aging, certain interventions that increase mitochondrial proliferation, such as over-expression of PGC1alpha in gut tissue, have recently been reported to increase life span and tissue function in aging Drosophila, and PGC1alpha activity is also implicated in maintaining tissue function during aging in mammals. In contrast, other manipulations that increase Drosophila mitochondrial proliferation, such as increased tissue-general expression of PGC1alpha or cyclin D/Cdk4 had negative consequences for life span and oxidative stress levels, and taken together these studies indicate that effective interventions in mitochondrial maintenance during aging will require tissue-specific targeting. Notably, certain carefully-timed interventions that reduce activity of ETC components have been shown to increase life span in both invertebrates and mammals, and this might function through a hormetic response to increase production of new mitochondria, or conceivably by inhibiting the activity of abnormal mitochondria. Critical questions for the future include determining the causes and mechanisms for the observed down-regulation of mitochondrial and ETC genes during aging - a pattern shared by Drosophila and mammalian tissues. Possible explanations include the inherently shorter-lived nature of mitochondrial genome sequences relative to nuclear genome sequences, genetic conflicts resulting from the uni-parental inheritance of mitochondrial genomes, and trade-offs between the costly production of new mitochondria and investments in growth, sexual differentiation and reproduction(Supplemental Figure S5), and these will be interesting areas for future research.

18Drosophila melanogaster flies were cultured on a standard agar/dextrose/corn meal/yeast media at 25°C. The transgenic strains GstD1-GFP and GstD1-deltaARE-GFP were generously provided by Dirk Bohmann. Age-synchronized cohorts of flies were generated by collecting newly-eclosed flies over a period of 48 hours, followed by maintenance in vials at approximately 20 flies per vial, with every-other day transfer to fresh media, until the indicated age time points. Visible images, GFP fluorescence images, and image overlays for flies were generated using the Leica MZFLIII fluorescence stereomicroscope. GFP fluore-scence was quantified using captured GFP images and Image J software, with mean and standard deviation calculated using 6 flies per sample. Flies used for stress treatments, RNA analyses and microarray analyses were generated as follows: males of wild-type strain Oregon-R were crossed to virgins of transgenic laboratory stock w[1118];rtTA(3)E2/TM3 Sb to generate hybrid progeny of genotype w[1118];rtTA(3)E2/+, as was used for previous microarray analyses, and 9-10 day-old male flies were used for each stress treatment. Old flies were 61 days of age, which corresponds to approximately the 50% survival point for the cohort. Vials containing 1% sucrose were prepared by adding 1.5 ml of 1% sucrose in deionized water to a Drosophila vial containing a single folded Kimwipe (Kimberly-Clark). For each stress treatment and the sugar-treated controls, replicate vials of 25 flies each were subjected to the treatment, and then the flies from each vial were separately processed for RNA, and each sample was used to generate probe for one micro-array hybridization, such that each treatment is represented by at least three biological replicates. For hyperoxia treatment flies in standard food vials were subjected to 100% oxygen atmosphere for 5 days as previously described. For ionizing radiation treatment flies in standard food vials were irradiated with 5666Rads/hour for 16 hours using a Cesium source (Grammacell 40-Cesium 137, Atomic Energy, Canada) at the USC Norris Cancer Center facility, and then transferred to 1% sucrose vials for two days followed by processing for RNA. For qPCR analysis 9 hour irradiation samples were also generated. Because ionizing radiation is inhibitory to transcription, the two-day recovery period was included to allow the gene expression response to develop; recovery in sucrose vials was employed because the newly-irradiated flies have greatly reduced mobility and will adhere to the surface of a regular food vial. For hydrogen peroxide treatment flies were placed in sucrose vials adjusted to 3% hydrogen peroxide for two days, and then processed for RNA. For heat stress treatment flies were placed in sucrose vials at 37oC for 5.5 hours and then processed for RNA. Controls for the effects of sucrose vials (“sugar-treated controls”) were generated by placing flies in sucrose vials for two days prior to processing for RNA.

19An average of 35 μg RNA was isolated from groups of 25 adult male flies using Trizol reagent (Life Technologies, Grand Island) according to the manufacturer's instructions. The RNA was further purified using the RNAqueous kit, and concentration was determined using NanoDrop spectrophotometer. A portion of the RNA (3 μg) was fractionated on 1.0% agarose gels to determine purity. 10 μg of total RNA was then used as substrate to generate biotinylated cRNA according to standard Affymetrix protocol (Childrens Hospital, Los Angeles, CA). A total of 35 Affymetrix gene chips were analyzed including at least four biological replicates for each experimental condition and control, with the exception of heat stress in which one array was omitted due to poor quality. The old, hyperoxia, and young samples were derived from our previous study in which six arrays were used for the hyperoxia and young conditions and four arrays were used for old flies. Quantitative real-time RT-PCR analyses and Northern blot analyses were performed as previously described, using RNA samples derived independently from those used for the microarrays.

20Gene expres-sion measures were computed based on a non-linear multi-chip model of the perfect match signal. This approach enables the separation of specific and non-specific components of the microarray signal and circumvents the issue of saturation bias in the high-intensity range. The background and concentration parameters were both fit within a single global routine (rather than estimating the background parameter before computing gene expression measures), and the model that best described the observed data selected. Linear modeling and empirical Bayes analysis was performed in the R statistical programming language (http://www.r-project.org/) using the Limma: Linear Models for Microarray Data package to identify genes significantly differentially expressed in response to multiple stressors or during aging; Limma computes an empirical Bayes adjustment for the t-test. Because the identification of genes altered in multiple condiions was a major objective of this study, a nested F-test approach was employed as this can be more powerful at detecting genes altered in multiple contrasts. Multiple testing was corrected for using the Benjamini and Hochberg method, which controls the false discovery rate (FDR) in this framework on a per-gene basis (but not across contrasts). Using this robust method, genes were found to be significantly differentially expressed both by biological and statistical criteria (±1.2 fold change, FDR 1% (p < 0.01); (Supplemental Table S4a, Supplemental Table S4b, Supplemental Table S4c, Supplemental Table S4d, Supplemental Table S4e, Supplemental Table S4f, Supplemental Table S4g, Supplemental Table S4h, Supplemental Table S4i, Supplemental Table S4j, Supplemental Table S4k, Supplemental Table S4l). Gene expression changes of ±1.5 fold were used for subsequent comparisons, as indicated. Hierarchical cluster analysis of the top 1000 differentially expressed genes for each condition based on the F-test p-value from the linear model fit was performed to visualize the gene expression patterns across different stressors, using the R package mclust. The microarray data discussed in this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) and are accessible through GEO Series numberGSEXXX.

21Statistically over-represented GO categories were identified using Flymine, by the calculation of a p-value denoting the probability that the observed numbers of counts could have resulted from randomly distributing a particular GO term between the test and the reference group. Multiple testing was controlled for using the Holm-Bonferoni method.

22The statistical significance of the overlap between various gene sets was evaluated by computing the p-value representing the probability of obtaining the observed number of overlaps by chance under a hypergeometric distribution, using the R function phyper.

23Gene annotations for the AffyDrosGenome1 arrays were updated to the latest information from Flybase using the online tool Flymine for all genes with expression altered ≥1.5 fold. Gene annotations identified by Flymine as matching more than one entry in the current database were resolved where possible, as follows: The Affymetrix probe ID was obtained from the limma files, and the corresponding probe sequence was obtained from the Affymetrix website. The probe sequence was then used to query the current Drosophila genome sequence annotation using the Flybase website and BLAST function to identify the correct gene. Ten probe sequences had ambiguous match that could not be resolved and were not included in the GO term analyses (Supplemental Table S5), and an additional 27 identifiers did not match genes in the current database. To control for any possible effects on gene expression patterns caused by two days maintenance of flies in sucrose vials, GeneChip analysis was performed on flies transferred to sucrose vials for two days in the absence of added stressors as a control. 258 genes were found to be up-regulated and 362 genes were found to be down-regulated relative to controls maintained on normal media (Supplemental Table S6), and these gene sets had no GO terms enriched among the up-regulated genes, and 4 GO terms enriched among the down-regulated genes: Proteolysis, Post-mating behavior, Insemination, and Lipid metabolic process (Supplemental Table S7). The genes that were up-regulated and down-regulated in response to sucrose were subtracted from the lists of genes up-regulated and down-regulated in response to hydrogen peroxide and ionizing radiation treatment to generate the final lists presented in Tables 1, 3, 4, and Supplemental Table S1, Supplemental Tables S2, Supplemental Table S3. While this simple subtraction procedure does not account for possible gene expression changes caused by interactions of sucrose with the stressors, we observe that the major GO term categories enriched in the gene sets up-regulated and down-regulated by hydrogen peroxide and ionizing radiation do not differ significantlywhen the gene expression changes caused by sucrose alone are included or excluded from the analysis (compare Supplemental Table S8 and Supplemental Table S9 where the effects of sucrose are included, to Supplemental Table S10, Supplemental Tables S11 where the effects of sucrose are excluded). In addition, cluster analysis demonstrated that the gene expression changes caused by the hydrogen peroxide stress treatment and ionizing radiation stress treatment were more similar to each other and to aging and hyperoxia than they were to the sucrose-treated control flies (Supplemental Figure S3), providing further evidence that the gene expression changes due to sucrose transfer do not make a significant contribution to the gene expression changes observed in the hydrogen peroxide and ionizing radiation samples.