| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1 Institut für Genetik
2 Institut für Angewandte Informatik, Forschungszentrum Karlsruhe, Karlsruhe, Germany
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
longevity; microarray analysis; starvation; larval development; metabolic adaptation
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
We have used Drosophila and mice as model organisms to identify genes that are regulated by nutrient signals (2, 39). In the Drosophila study, we carried out a microarray analysis of larvae under complete (saline only) and protein starvation (sugar fed only) conditions. Various components of signaling transduction cascades were identified as being affected by these conditions, including genes known to be involved in nutrient signaling and growth control such as the translation regulator Thor/Drosophila initiation factor 4E-binding protein (d4E-BP) (34, 35). In addition, patterns of gene expression were revealed for which it would have been difficult to predict. For example, entire classes of genes were identified in which no regulation was seen upon complete starvation but were strongly affected upon protein starvation (39). The changes in gene expression most likely reflect changes in the metabolic program designed to increase survival and sustain growth during nutrient stress. The specific nature of the metabolic program and its underlying gene expression pattern is, in turn, dependent on the different physiological requirements of the animal. For Drosophila larvae, the main impetus is organismal growth, where the animals greatly increase their mass and overall size within a very short period of time. Adults, in contrast, are faced with a different task. They do not grow in size, and they spend a much longer time as adults than as larvae. These differences are also reflected in the distinct feeding behaviors of larvae and adults (22, 38). In addition, the sex difference is expected to play a much greater role in adults than larvae. There have been numerous studies, for example, on the sex differences in response to starvation resistance and lifespan as well as the reallocation of available metabolic resources under nutrient stress (26, 33).
To identify some of the metabolic and genetic pathways that serve as possible points of control for differential nutrient reallocation, we analyzed gene expression profiles in Drosophila males and females through the use of microarrays. Microarrays have been used previously in Drosophila to analyze gene expression changes under various conditions, including aging, stress, nutrient deprivation, and caloric restriction (10, 20, 27, 39, 40). However, in all cases, they have been done at the larval stage or on only one sex in the adult. The present study provides a direct comparison of the gene expression profiles of males and females under different nutrient conditions. Although microarrays cannot provide information on metabolic flux through biochemical pathways, expression changes in genes that encode enzymes of given pathways may provide hints as to the mechanisms involved. Through these analyses, we were able to identify potential metabolic control points for resource reallocation under nutrient stress and discuss its relevance for understanding the observed lifespan-reproduction tradeoff.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
93% of the genes predicted in version 1.0 of the Drosophila genome from sequences released by the Berkeley Drosophila Genome Project and Celera Genomics in March 2000. The amplicons represented 75% of the genes predicted in the January 2003 release 3.1 Drosophila genome annotation update. The regions amplified to build the array ranged in size from 150 to 600 bp with an average size of 410 bp. PCR fragments were amplified, and each amplicon was quality control tested for DNA concentration, purity, and fragment size. A total of 11,796 reactions led to a single product of correct size and sufficient amount. The 675 reactions that produced multiple products were run on a gel again, and the band of the correct size was excised cloned into pCR TOPO 2.1 (Invitrogen; Karlsruhe, Germany). The 2,034 reactions where no band was visible were also subjected to a "blind" clone attempt. Those 2,709 cloning reactions led to 2,033 positive clones. From them, the inserts were amplified with vector-specific primers by colony PCR. Those PCR products were handled as described above for the directly amplified open reading frame-specific amplicons and were included on the microarrays. Probes from nonsequence-verified clones were excluded from data analysis in the present study. A mixture of all these 2,033 clones was used to make a titration dilution as a hybridization control [the "incyte amplification titration standard" (IATS)], which was included in each of the 48 blocks of our array.
Arrays were spotted with an Omnigrid 100 (GeneMachines; San Carlos, CA; now Genomic Solutions; Ann Arbor, MI, USA) with 24 SMP3 pins (Telechem; Sunnyvale, CA) on CMT-GAPSII slides (Corning; Corning, NY). By having two identical subarrays next to each other on each slide, we obtained 48 blocks with 26 rows and 26 columns each. Spot to spot spacing was 150 µm; average spot diameter was 100 µm.
Nutrient Conditions and Fly Handling
Larvae were kept on apple juice-agar plates with yeast and raised at 25°C. For the developmental control, larvae were collected 40 (±1) h after egg laying (AEL) for the developmental time points at 52 (±1), 64 (±1), 76 (±1), 88 (±1) and 100 (±1) h AEL.
Adult flies were kept at 25°C in 50 ml flat-bottom PS tubes (Greiner bio-one; Frickenhausen, Germany) on standard fly food (16.5 g yeast, 81.5 g cornmeal, 8 g agar, 100 ml sugar beet syrup, and 200 ml of 10% nipagin in ethanol in 2 liters water). The flies collected had eclosed over a 24-h period and were kept in groups of 200 flies/tube. After 5 days, we separated males and females and put them in groups of 80 flies/28-ml PS tube. We allowed for 1-day CO2 recovery on standard fly food before flies were moved to 1) tubes with yeast paste on PBS-soaked filter paper (normal-fed control); 2) tubes with only PBS-soaked filter paper (total starvation); or 3) tubes with filter paper soaked in PBS with 20% sucrose (sugar condition, also referred to as protein starvation). After 24 or 48 h, flies were shock frozen in liquid nitrogen and stored at –80°C for further handling.
Males and females were assayed separately. For females, we performed 24- and 48-h total starvation and 24- and 48-h sugar feeding (protein starvation). For the males, only 24-h total starvation was done because most died upon 48-h total starvation; 24- and 48-h sugar feeding was also performed. These experiments essentially parallel the experimental setup of previous larval experiments done in our laboratory (39).
RNA Preparation, Labeling, and Hybridization
Larval total RNA was prepared using the Macherey and Nagel (Düren, Germany) RNA L kit as previously described (39). For RNA preparation from adult animals, 80 collected shock-frozen flies were transferred to 1.5-ml reaction tubes (Eppendorf; Hamburg, Germany) containing 400 µl TRIzol (Gibco-Invitrogen; Carlsbad, CA). Samples were homogenized for 5 min using a disposable polypropylene pellet pestle. After another 600 µl TRIzol was added, the manufacturer's protocol was followed. From the derived total RNA, polyA RNA was extracted with the Ambion Purist kit (Austin, TX) according to the manufacturer's protocol.
Microarray Hybridization
A total of 63 microarrays were used to monitor gene expression changes in this study.
Flies and larvae for each time point or condition were obtained. For the larval development, the 40-h AEL sample was used as a reference. The larval collections were obtained independently; the adult collections were always done in parallel for the different feeding conditions. From each biological sample, total RNA and mRNA were made (see Supplemental Table 9; available at the Physiological Genomics web site).1 Equimolar pools of the mRNA level were made to balance the variance of the biological repeats.
Labeled cDNA was synthesized from 1.5 µg polyA RNA using the Amersham Direct cDNA Labeling Kit (Amersham Europe; Freiburg, Germany). Removal of incorporated dyes and concentration of the target were done with Microcon 30 spin columns (Millipore; Bedford, MA) according to the manufacturer's instructions. The concentrated probes were hybridized to the microarray in 1x dig easy hyb buffer (Hoffman-la Roche; Basel, Switzerland) overnight (minimum 16 h) at 42°C.
Each labeled cDNA sample derived from a pooled mRNA probe was used to hybridize at least two times with each dye (Cy3 and Cy5, duplicate dye swap) (see Supplemental Table 9).
Microarray Scanning
Arrays were scanned using the dual-laser scanner Axon 4000B and the corresponding software Genepix 4 (Axon; Union City, CA). Both channels (532 nm for Cy3 and 635 nm for Cy5) were scanned in parallel and stored as 16-bit TIFF files. The absolute intensity values span the range from 0 to 65,535. The scans were performed with a resolution of 10 µm. From each spot with a mean diameter of 100 µm, 100 data pixels were recorded. The individual local background areas around the spots were defined, which included 400 pixels and excluded neighboring spots. For each channel, raw data were calculated as median intensities of all foreground pixels from which the median intensity of the background pixels was subtracted.
Each array was scanned three times (low, medium, and high scan) with different signal amplification factors (voltage settings of the photo multiplier tubes) but with the same laser power. The channels for Cy3 and Cy5 were balanced in each scan for approximately the same intensity profile. In the low scan, no spot was saturated (higher or equal to the absolute maximum intensity value); in the high scan, the signal amplification for Cy5 was set to 80% of maximum and the Cy3 amplification was adjusted to this. The settings used in the medium scan lay between the low and high scans. This method for scanning has several advantages, as described previously (2). In the low scan, where no spot is in saturation, it is possible to calculate the real ratio for genes with high expression levels; however, those with a low expression level are most likely not recognized. In order to not lose these, a high scan is made; in this case, the information on the saturated spots is lost, so the two scans complement each other. The medium scan produces additional values for subsequent calculations. By scanning the arrays three times, errors that occurred during recording and that might have increased the error factor in the normalization were averaged.
Data Processing and Normalization
Data processing was carried out with the Karlsruhe Microarray Analysis Tool (KAMAAN), a program written in the R programming language. It consists of four sequential steps, as follows.
Step 1.
Background-corrected intensity values were calculated for each array and both color channels for each spot from the values of the three different scans. This calculation is based on the assumption that background-corrected intensities of different scans can be brought to the same scale by a affine-linear transformation, i.e., yij = si x yij + oi [where the number of scans i equals 1, 2, or 3 and the number of spots j equals 1, 2, 3,..., with estimation of the si, oi by location (median) and scale (median of absolute deviation) parameters of the distributions of yij (for a description, see Ref. 14)].
Step 2.
Intensity values from step 1 of both color channels and all redundantly hybridized arrays (including dye swaps of compared cDNA probes) were brought on the same scale and transformed with the arsinh function to stabilize their variance over the intensity range, based on the assumptions (12, 29) that 1) intensity values from different arrays and color channels have an affine-linear relation, i.e., yij
''} = 
i x y{ij + ßi [where the number of arrays i equals 1, 2, 3..., 2 x number of arrays (multiplication by 2, because of two colors)]; 2) i can be decomposed in an offset (ai) and a zero mean Gaussian variable; and 3) ßi is a product of a gain factor (bi) and a log-normal distributed variable, with estimation of the ai, bi with a robustified maximum likelihood estimator, assuming that the majority of genes are not differentially expressed, using the application of the R package VSN (13).
Step 3.
A further normalization step reduced spatial systematic effects (specific to spot position). A two-dimensional median filter was applied on the differences of calibrated and transformed intensity values from step 2 for each array separately; this is comparable to the spatial normalization method in Ref. 36.
Step 4.
A scale normalization was performed to bring the M values of different arrays onto the same scale using the method described in Ref. 36. This effectively results in an array-specific weighting of M values when calculating the mean of M values for finding differentially regulated genes where weights are inversely proportional to variance (or a robust estimate) of the M values of a chip. In the case of our experiments, differences in the overall variances of the arrays should be due to technical reasons.
Differentially expressed genes were identified using a regularized t-test (1), and adjustment of P values for control of the false discovery rate (FDR) used the method of Benjamini and Hochberg (4). In subsequent examinations, only a gene whose ratio placed it in the 99.5% confidence interval, based on at least six data points, was included, which resulted in a FDR cutoff of 1%.
For quality control, the results of all steps were assessed using diagnostic plots. To judge the raw data, foreground and background intensities for each channel, linear (not normalized) background-corrected ratios, and logarithmic (not normalized) background-corrected ratios were plotted by color coding the rank of the values. Additionally, the spots that showed no detectable signal were plotted. The raw data were also visualized by diagrams that show the background-corrected values against the uncorrected values of each scan and channel. Between all scans, the background-corrected intensities were averaged, logarithmically transformed, and plotted one channel against the other.
The normalization process can be assessed by the plots of the non-normalized M values, normalization constants, and normalized M values as well by the plots of non-normalized M values against the rank of A and the plots of normalized M values against the rank of A [MA plots (36)].
Ontological Data Analysis
Gene categories were obtained by taking the strongest regulated genes of the different experiments as landmarks for sorting other genes into functional groups. This manual method takes the extent of regulations into account. By doing a literature inquiry on a single gene basis, information that was missed in the gene ontology (GO), e.g., in Flybase, were also included.
As complementary approach, we used the web-based tool Onto-Express (17). Onto-Express is available online (http://vortex.cs.wayne.edu/ontoexpress/) and can be used, after registration, cost free. As input files, lists of Flybase identifiers of all significantly regulated genes of an experiment were submitted together with a reference file that contained the identifiers of all genes on the used microarrays. The data files and the reference list, as well as a description of the files, can be found in the Supplementary Material. Onto-Express was run with the following settings: organism = "drosophila melanogaster," input type = "flybase id," reference array = "my own array," distribution = "hypergeomatric distribution," correction = "fdr," and search for = (all tags set) "biological process," "cellular component," "molecular function," and "chromosome information." In the results, the complete GO tree could be explored and the genes of each GO category listed. Subsequently, the results were saved locally, and the identified biological processes were further analyzed using MS Excel. Only those GO groups were considered as significantly affected that contained a minimum of three genes (column "unique input total") and showed a corrected P value of <0.05.
In this way, a table of GO categories was obtained for each experiment. The lists of the 24-h total starvation experiments were combined into one table as well as the lists of the 24- and 48-h sugar feeding experiments. A GO group was considered as equally affected between the sexes according to the following criteria: 1) if the GO categories were found to be significantly affected in all experiments and 2) if the difference in the number of affected genes between females and males was <26% of the sum of all affected genes in the category.
For reasons of clarity, four tables were created, which can be found as Tables 7 and 8 in the text [GO categories that show a sex-dependent bias under 24-h total starvation (Table 7) and under protein starvation (Table 8)] and Supplemental Tables 7 and 8 [GO categories that were found to be equally affected in female and male flies under 24-h total starvation (Supplemental Table 7) and under protein starvation (Supplemental Table 8)].
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
The data on which this paper are based were deposited with the National Center of Biotechnology Information (NCBI) Gene Expression Omnibus (http://www.ncbi.nml.nih.gov/geo/) in the sample series GSE1672.
Reproductive Physiology
We first analyzed genes that are known to be expressed in a sex-specific manner. This served not only as an internal control for the experiments but also provided information on the genes that underlie known physiological responses, such as the decrease in egg laying seen in protein-deprived conditions. As expected, there was a strong downregulation of the genes encoding chorion and vitelline proteins in females (Table 1). These gene products are required in high amounts for egg production; the chorion genes, in fact, are amplified to meet the high demands of chorion biosynthesis during oogenesis (31). They were downregulated upon all starvation and sugar conditions. Analogously, the accessory protein encoding genes of the male were downregulated slightly but consistently under the starvation condition only in males (Table 1). This was also consistent with the decrease in reproduction under nutrient stress conditions. A major conclusion from this analysis is that the egg production genes are not dependent on the energy source per se because the addition of sugar had little effect compared with total starvation. In contrast, the accessory gland genes in males were completely dependent on sugar.
|
|
|
|
Purine and Folate Metabolism
In our view, the most intriguing difference between males and females upon nutrient stress was found in purine metabolism (Table 5). The largest difference was seen after 48 h under protein starvation (sugar condition), in which genes were downregulated to a much greater degree in males. For example, the ade3 gene was unchanged in 48-h sugar-fed females but was 10-fold downregulated in sugar-fed males. This difference was not observed for total starvation, where both males and females showed a similar, low degree of downregulation. These differences in the expression profile raise two questions. First, why would genes be more strongly regulated in protein-starved flies than completely starved flies? Second, why is this observed most prominently in males but not in females?
|
The second point can be explained in terms of differential nutrient resource allocation between the sexes. We have shown above that many genes required for egg production, such as chorion genes, are highly downregulated in females under nutrient stress (Table 1). No comparable regulation of male-specific reproductive genes is seen. Perhaps the females do not decrease their purine biosynthesis because they can reallocate more amino acids by shutting down reproduction. In other words, they have other metabolic resources to utilize, including those normally available for egg production. This would allow females to maintain their purine biosynthesis. In contrast, males do not have this extra metabolic resource pool, so they must react by decreasing their purine biosynthesis. As we will outline later, these considerations point to purines as playing an important role in the allocation of resources between reproduction and longevity.
Immune Response and Growth Control
There was also a sex-specific regulation of genes involved in the immune response (Table 6). Dpt and mtk were, in fact, the two highest upregulated genes in our array in females starved for 48 h. The large difference between 24- and 48-h data in females indicates that these immune response genes were abruptly upregulated upon prolonged starvation. Interestingly, aging flies also showed a large increase in the immune response genes shortly before death (20, 27). Therefore, the upregulation of immune response genes upon prolonged starvation may reflect impending death and adds to the speculation that one of the major causes of death in laboratory conditions is infection.
|
Larval Nutrition and Adult Physiology
There is a major influence of larval nutrition on adult physiology, as illustrated by the observations of Beadle and colleagues (3). Starving larvae before 70 h AEL results in larval death, whereas starving them after this time point (the "70-h change") results in viable but small adults (3). As part of our interest in the genetic mechanisms underlying the 70-h change (39), we analyzed global gene expression across this time period and examined how the genes that were regulated in a sex-dependent manner behaved (Supplementary Tables 1–6). Several interesting patterns were revealed through this comparison. First, there was a huge increase in expression of the Lsp genes after the 70-h time point (Supplementary Table 2). The differences among the specific Lsp genes were also striking: they all increased dramatically during late larval growth, but their regulation in adults was highly specific (e.g., Lsp2 was the highest regulated gene upon nutrient stress in adults, whereas Lsp1gamma was not regulated at all).
We also observed a significant increase in Thor (d4E-BP) expression near the end of larval growth (Supplementary Table 6); we have previously shown that Thor expression is upregulated during complete and protein starvation in larvae (39). These data are consistent with the view that increased Thor expression correlates with stoppage of cellular growth, because larvae must stop their growth in preparation for metamorphosis at the late stage. Recent studies have also pointed to Thor as a negative regulator of growth under stress conditions (34, 35).
The most dramatic gene expression pattern connecting larval and adult physiology was found for two Obp genes: Obp99a and Obp99b (Supplementary Table 4). Obp99a was downregulated only in females during nutrient stress and showed essentially no change during larval growth; in stark contrast, Obp99b was regulated only in males during nutrient stress and showed a huge increase (over 100-fold) in expression during late larval growth. It would be interesting to see whether Obp99b is expressed in a sex-biased manner (i.e., only in males) already at the larval stage.
Additional Categories of Sex-Biased Changes in Gene Expression
The above categorization of genes was based on the initial semimanual selection of the highest regulated genes in each of the nutrient conditions and then using these as guideposts to identify other genes that could be involved in a particular process. To complement this approach, we also analyzed our data by using Onto-Express (see MATERIAL AND METHODS), which categorizes regulated genes by using GO (18). Briefly, we took all significantly regulated genes and identified GO categories that displayed differential expression patterns between males and females (Tables 7 and 8). This approach revealed several informative insights. First, there was a large overlap compared with the functionally based semimanual approach we used above. Expected gene categories included insect chorion production, vitellogenesis and vitelline membrane formation, and postmating behavior (the Acp genes), which showed a clear sex bias in the Onto-Express list as well. Unexpected sex differences, such as in phototransduction and in specific metabolic pathways (e.g., amino acid catabolism), were also revealed.
|
|
Concluding Remarks
Purines as metabolic currency in the tradeoff between reproduction and lifespan.
Microarray analysis clearly cannot give information on the direction of metabolic flux, but it can generate hypotheses that can guide one in where to look. The majority of the large regulatory changes under nutrient stress was in genes involved in amino acid and protein metabolism. One of the most revealing differences in the transcription profile between males and females occurred under protein starvation in genes that encode enzymes of purine biosynthesis. These genes become greatly downregulated in males compared with females (Table 5), and we suggested above that this could be due to the different abilities of the sexes to reallocate available metabolic resources. Strikingly, a recent microarray study (20) showed that all the genes in the purine biosynthetic pathway become upregulated in oxidatively stressed and in aging male flies (Fig. 1). As suggested by Landis et al. (20), this upregulation of purine biosynthesis might be required for repair processes upon increased oxidative damage and aging. By the same token, the downregulation of purine biosynthesis under nutrient stress would lead to a decreased ability for repair processes.
|
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
Present address of J. Jaekel: Hochschule für Technik, Wirtschaft und Kultur Leipzig, Institut für Mess-, Steuerungs-, und Regelungstechnik, Postfach 301166, Leipzig 04251, Germany.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: M. J. Pankratz, Institut für Genetik, Forschungszentrum Karlsruhe, Postfach 3640, Karlsruhe 76021, Germany (e-mail: michael.pankratz{at}itg.fzk.de).
* M. Bauer and J. D. Katzenberger contributed equally to this work. 
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONGRANTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Missirlis, S. Kosmidis, T. Brody, M. Mavrakis, S. Holmberg, W. F. Odenwald, E. M. C. Skoulakis, and T. A. Rouault Homeostatic Mechanisms for Iron Storage Revealed by Genetic Manipulations and Live Imaging of Drosophila Ferritin Genetics, September 1, 2007; 177(1): 89 - 100. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
