Quantitative trait transcripts for nicotine resistance in Drosophila melanogaster Gisele Passador-Gurgel1, Wen-Ping Hsieh2, Priscilla Hunt1, Nigel Deighton3 & Greg Gibson1 Although most genetic association studies are performed with the intention of detecting nucleotide polymorphisms that are correlated with a complex trait, transcript abundance should also be expected to associate with diseases or phenotypes. We performed a scan for such quantitative trait transcripts in adult female heads of the fruit fly (Drosophila melanogaster) that might explain variation for nicotine resistance. The strongest association was seen for abundance of ornithine aminotransferase transcripts, implicating detoxification and neurotransmitter biosynthesis as mediators of the quantitative response to the drug. Subsequently, genetic analysis and metabolite profiling confirmed a complex role for ornithine and GABA levels in modification of survival time upon chronic nicotine exposure. Differences between populations from North Carolina and California suggest that the resistance mechanism may be an evolved response to environmental exposure. Gene expression profiling has been used in a variety of ways to gain insight into the architecture of complex traits. Two common applications are multivariate clustering to identify molecular signatures of disease status1 and whole-transcriptome scans for expression quantitative trait loci (eQTL), also known as genetical genomics2. Neither of these approaches uses the full range of variation in nature: the former typically reduces continuous variation to dichotomous traits, such as affected or unaffected, and the latter concentrates on the genetic variation segregating in a single cross between two divergent genotypes. The next logical extension is to search for correlation between transcript abundance across a diverse population of organisms, with the values of a continuous trait. By analogy with nucleotide polymorphisms that associate with phenotypic variation, namely quantitative trait nucleotides3, we propose to refer to such transcript associations as quantitative trait transcripts, or QTT. Like many vertebrates4, when adult D. melanogaster are exposed to nicotine, they are initially stimulated, but chronic ingestion reduces activity within 24 h and is eventually lethal to flies5. There is ample genetic variation for survival time on nicotine6. A survey of over 200 nearly isogenic lines from two populations in North Carolina and

California demonstrated higher variability in North Carolina lines, with approximately 30% of the North Carolina lines more resistant than all but one California line (Fig. 1). It is not known whether this reflects resistance to residual environmental nicotine in proximity to tobacco fields, an evolved response to nicotine-based insecticides, or some other cause, such as genetic drift. In order to identify genes that might mediate resistance to nicotine, we performed cDNA microarray analysis on adult female heads of flies in the absence or presence of nicotine, from a sample of 108 inbred wild-type lines. Differences in overall gene expression profiles in response to the drug within and between the populations defined several categories of metabolic response (Supplementary Methods online). Identification of genes that respond globally to the drug says little about which genes mediate differential resistance and susceptibility. To address this question, we reasoned that only a fraction of transcripts would show a high correlation between transcript abundance and lifespan on nicotine and that such transcripts might define a mechanism for differential survival. For each of the 4,385 transcripts

70 60

Survival time (h)

© 2007 Nature Publishing Group http://www.nature.com/naturegenetics

LETTERS

North Carolina California

50 40 30 20 10 0 71 Rank order by population

115

Figure 1 Distribution of survival times on nicotine. Survival time in hours is plotted by rank order for 115 North Carolina inbred lines and 71 California inbred lines. Note that 36 of the North Carolina lines are more resistant to nicotine than all of the California lines.

1Department of Genetics, North Carolina State University, Raleigh, North Carolina 27695, USA. 2Department of Statistics, National Tsing Hua University, Hsinchu 30013, Taiwan. 3Metabolomics and Proteomics Laboratory, North Carolina State University, Raleigh, North Carolina, 27695 USA. Correspondence should be addressed to G.G. ([email protected]).

Received 16 October 2006; accepted 17 November 2006; published online 21 January 2007; doi:10.1038/ng1944

264

VOLUME 39

[

NUMBER 2

[

FEBRUARY 2007 NATURE GENETICS

LETTERS but as can be seen for one example in Figure 2, the regression for CG8745 was largely due to Number meeting significance level elevated transcript abundance in the highly resistant lines. As we did not observe such Source of RNAa Treatmentb P o 0.00001 P o 0.0001 P o 0.001 resistance in the California sample, the elevated expression in North Carolina flies may Total RNA, North Carolina Control 1 9 54 be an adaptation to environmental nicotine, Nicotine 0 0 9 although it is also possible that the California Amplified RNA, North Carolina Control 0 0 5 flies have lost the ability to resist nicotine. To Nicotine 0 2 15 confirm the association, we replicated the Amplified RNA, California Control 0 0 0 microarray analysis using total RNA labeling Nicotine 0 0 11 in place of linear amplification, and we found Amplified RNA, Combined Control 0 0 24 Nicotine 0 14 94 54 genes significant at P o 0.001, including three of the previous five genes, CG8745, a a‘Total RNA’ refers to total RNA labeling for each of the 58 North Carolina lines, and ‘amplified RNA’ to linear amplification of each of the 58 North Carolina or 50 California lines. ‘Combined analysis’ refers to reevaluation fitting all 108 lines in a single specific alkaline phosphatase (CG8147) and a b model. ’Control’ refers to RNA extracted from adult female heads of flies kept on normal cornmeal; ‘nicotine’ refers to flies genetically uncharacterized elastin-like potenexposed to nicotine for 8 h. tial phosphate transporter (Ela, CG7021). These genes thus represent candidate mediaon the array, relative abundance both in control and nicotine-treated tors of nicotine resistance in D. melanogaster. Transcript abundance flies was regressed on survival time. The number of significant explains close to 20% of the variance for survival time in both regressions for each comparison in each population and a combined replicates, and together CG8745 and CG8147 explain 30% of the analysis is shown in Table 1. As there is a large difference in phenotype variance. Although the magnitude of the significance of the associadistributions between the two populations, the combined analysis is tion is not as high as is typically observed between nucleotide variants confounded by population structure, so we focus on significant and quantitative traits8–11, the amount of variance explained is greater, regressions within each population. and given the error associated with measurement of a continuous At a significance level of 0.001, only four associations were expected variable (gene expression) as opposed to dichotomous genotypes, the by chance in either population. In both the North Carolina and quantitative trait transcript effect is very strong. California samples, 15 and 11 genes, respectively, showed this level of CG8745 is predicted to encode one of two enzymes in D. melanoregression of transcript abundance on survival time after exposure to gaster with ornithine-oxo-acid transaminase activity (Oat, EC nicotine, but none of these genes was shared between the analyses 2.1.6.13) that catalyze the reduction of ornithine to L-glutamate and (Supplementary Table 1 online). When we pooled all of the data from L-glutamate semialdehyde and sit at a key branch point that is likely to both populations, we observed 94 significant genes (at a false discovery be important for detoxification12 (Supplementary Fig. 1 online). rate of just 2%), mostly in the categories of nucleic acid and lipid Upon exposure to nicotine, CG8745 is downregulated by 33% (t test, metabolism, as well as proteolysis, suggesting that survival is asso- P o 10
23), presumably allowing ornithine to be diverted into the ciated with redirection of metabolism in specific directions. Nicotine is urea cycle and probably inhibiting the accumulation of glutamate. an appetite suppressant7, and survival is correlated with starvation Glutamate has several roles, functioning as a major stimulatory resistance, although it is clear that there is also a component of the neurotransmitter13 but also serving as a substrate for synthesis of response that is specific to nicotine6. the inhibitory neurotransmitter GABA by the enzyme Gad1 (ref. 14). Considering mRNA levels in control flies not exposed to nicotine, Glutamate activity is downregulated by conversion to glutamine, but, the survival time of flies in response to nicotine in the North Carolina notably, transcription of glutaminase is elevated by nicotine exposure, sample was associated with the abundance of five transcripts at P o suggesting that the neurotransmitter is recycled to offset the reduction 0.001. None of these effects was replicated in the California sample, in biosynthesis from ornithine. In mammals, the distribution of Gad1 (EC 4.1.1.15) is directly correlated with the concentration of GABA15. This gene was not present on our microarrays, but quantitative realtime PCR16 showed (as expected) that it was expressed at a signifi1.5 2 cantly lower level in nicotine-sensitive lines than in nicotine-resistant R = 0.2307 Transcript abundance in adult female heads

© 2007 Nature Publishing Group http://www.nature.com/naturegenetics

Table 1 Number of transcripts correlated with survival time on nicotine

P < 0.00005

1

Table 2 Downregulation of Gad1 by nicotine exposure and in sensitive flies

0.5 0

Category

Diet

Mean Cta

Comparisonb

Effectc

Resistant

Control Nicotine

18.34 ± 0.28 18.49 ± 0.12

Resistant lines Sensitive lines

0.91 0.50

Sensitive

Control Nicotine

18.79 ± 0.10 19.79 ± 0.21

Control diet Nicotine diet

0.74 0.41

–0.5

–1 0

10

20 30 40 50 60 Female survival time on nicotine (h)

70

Figure 2 Regression of transcript abundance on survival time for CG8745. The highly significant regression shows that survival time explains 23% of the variation in ornithine aminotransferase transcript abundance in adult female heads, for total RNA–labeled samples. Linear amplification of the same samples uncovered a similar regression that explained slightly less (22%) of the variation.

NATURE GENETICS VOLUME 39

[

NUMBER 2

[

FEBRUARY 2007

aMean Ct (threshold cycle at inflexion point of product accumulation curve) across replicates, ± 1 s.d. bComparison of expression in flies on nicotine diet versus control diet in the resistant or sensitive lines, or of expression in nicotine-sensitive versus nicotine-resistant flies on control or nicotine diets. cEstimated reduction in expression computed by DDCt method16, showing expression in flies on a nicotine diet as a multiple of that in flies on a control diet (or expression in nicotine-sensitive flies as a multiple of that in resistant flies) using normalization against average of all 24 measurements of 5HT2.

265

LETTERS

a

b

70

DoF

MS

F ratio

1 729 7 1,216 1 10,432 7 63 1 21 7 280 7 64 64 26

27.7 46.1 395.8 2.4 0.8 10.6 2.4

P value

60 Genotype Line Sex L×G S×G L×S L×S×G Vial

50 40 30 20

Mutant

<0.0001*** <0.0001*** <0.0001*** 0.032* 0.379 <0.0001*** 0.028*

Balancer

Figure 3 Haploinsufficiency of CG8745e00991 for survival time on nicotine. (a) Lines link average survival time from three replicate vials for Mutant/+ and Balancer/+ siblings, averaged across the sexes in each of eight crosses. (b) ANOVA table showing the degrees of freedom, mean square, F ratio and P value associated with each effect, namely Genotype (Mutant versus Balancer), Line (the eight crosses) and Sex.

only one line showed a slight, nonsignificant increase (Fig. 3a). Threeway analysis of variance (Fig. 3b) indicated that despite large differences in survival time among the eight lines and between the sexes, the overall genotype term was highly significant (P o 0.0001), with an average decrease in survival time owing to haploinsufficiency of 12% for the enzyme. The lack of significance of the genotype-by-sex interaction term confirms that reduction of Oat activity decreases the resistance of both male and female flies to nicotine to a similar degree. We obtained further support for a role for ornithine transaminase in mediation of nicotine response by mass spectrometric profiling of metabolites in adult female heads. Both GABA (P ¼ 0.01; Fig. 4a) and ornithine (P ¼ 0.002; Fig. 4b) concentrations showed significant regression on survival time in whole adult female heads after exposure to nicotine. Notably, both the mean and variance of ornithine and GABA levels were higher in the North Carolina sample than the California sample, but this was not the case for an alternate source of GABA biosynthesis, putrescine (data not shown) or for nicotine measured in the same samples (Supplementary Fig. 3 online). The absence of any relationship between nicotine levels and survival time rules out metabolism of the drug as a source of variation for survival time. Population structure explains much of the positive correlation between GABA or ornithine and survival time, but, notably, the slope of the regression for ornithine was the same in the North Carolina sample as it was in the combined sample. As CG8745 was downregulated upon ingestion of nicotine, possibly as part of a negative feedback mediated by GABA21, overall ornithine levels rose after exposure. We suggest that more efficient downregulation of the enzyme in resistant flies quickly turns off glutamate production and produces a greater pool of ornithine that is available to fuel the urea detoxification cycle. However, it should also be noted that fivefold replication of metabolite measurements in two of the highest–nicotine resistance and two of the lowest–nicotine resistance lines did not uncover any large sample variance associated with measurement of ornithine, and we did not detect any heritable variation for concentration of ornithine, whereas 35% of the variance in GABA levels was attributable to line differences in the same samples (data not shown). Finally, we also observed an excess of transcripts whose abundance is associated with ornithine levels in flies from North Carolina. Whereas, as expected by chance, only one or two transcripts regress on concentrations of nicotine, cotinine or GABA at P o 0.0001, 13 transcripts regress on ornithine levels in control flies, and five on ornithine levels after nicotine exposure (Supplementary Table 2 online). None of the genes is shared in common between the treatments; most represent unannotated genes, and those that are

266

Survival time (h)

lines (F1,21 ¼ 107, P o 0.0001, analysis of variance (ANOVA)) and that transcriptional downregulation in the presence of nicotine was much stronger in nicotine-sensitive lines than in nicotine-resistant lines (F1,21 ¼ 26, P o 0.0001, ANOVA) (Table 2 and Supplementary Figure 2 online). This suggests that nicotine resistance might be due in part to relatively high ratios of GABA to glutamate before exposure to nicotine, which is also neurostimulatory. According to the Kyoto Encyclopedia of Genes and Genomes database, the particular alkaline phosphatase (EC 3.1.3.1, CG8147) that is associated with survival time is implicated in degradation of g-hexachlorocyclohexane, also known as lindane, which is a pesticidal neurotoxin that antagonizes function of GABAergic chloride channels17. A similar function is shown by the cytochrome P450 Cyp6a20 (EC 1.14.14.1; ref. 18) that was associated with nicotine resistance in the combined analysis of both California and North Carolina samples. The two enzymes converge on benzoate degradation, which is also in the pathway of metabolism of another potent insecticide, Parathion. Notably, singular value decomposition has shown19 that lindane degradation is the most strongly differentially expressed pathway in airway epithelium of human smokers compared with non-smokers. Furthermore, in our experiment, transcription of the fast-acting GABA-A receptor gene was upregulated by nicotine exposure, whereas levels of the slow-acting GABA-B receptor transcript are lower in sensitive flies than resistant flies. All of these data would be consistent with the model that survival is mediated by reduction of the potency of nicotine’s activation of nicotinic acetylcholine receptors by inhibitory GABAergic signaling on dopamia 60 b 60 nergic neurons. 50 50 To test for a genetic effect of reduction in 40 40 CG8745 activity, we compared survival time 30 30 of hemizygous and heterozygous siblings derived from the cross of CG8745e00991 / 20 20 TM6B, Hu Tb to a panel of eight of the inbred 10 10 wild-type lines from North Carolina that 0 0 0.8 0.2 0.4 –0.6 –0.4 –0.2 0 0.6 –1 –0.8 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 were chosen to cover the range of nicotine Normalized log [ornithine] Normalized log [GABA] responses. The CG8745e00991 allele is a lethal insertion of a mini-w+ piggyBac{RB} element Figure 4 Regression of GABA and ornithine metabolites on survival time. (a) GABA levels after in the 3¢ UTR of the gene20. Consistent with exposure to nicotine regress weakly on survival time across the entire sample, although the withinthe microarray results, seven of the eight population regression is not significant for either North Carolina or California. (b) Ornithine levels after crosses showed a reduction in lifespan in exposure to nicotine regress strongly on survival time, with very similar regression lines for the entire flies carrying the CG8745 mutation, whereas sample (solid line) and just the North Carolina sample (dashed line). Survival time (h)

© 2007 Nature Publishing Group http://www.nature.com/naturegenetics

Mean survival time (h)

Effect

VOLUME 39

[

NUMBER 2

[

FEBRUARY 2007 NATURE GENETICS

LETTERS known do not obviously share functions. Nevertheless, these data are consistent with the notion that regulation of ornithine level is an important biochemical control point that deserves investigation for its role in detoxification and mediation of drug response in D. melanogaster. Our experiments also suggest that population sampling should be considered in parallel with genetical genomic strategies for dissecting quantitative traits via gene expression profiling.

© 2007 Nature Publishing Group http://www.nature.com/naturegenetics

METHODS Drug treatment. Female flies were collected from 108 isofemale lines sampled from two populations of Drosophila melanogaster, previously evaluated for survival time in nicotine food. Fifty-eight lines were sampled from a population originally collected from peach orchards in West End, North Carolina22, and 50 lines were from a vineyard population in Wolfskill, California23. Flies were collected between 1 and 3 d after eclosion and kept at a density of 50–60 flies per vial, separated by sex, and were placed separately in vials containing regular cornmeal media (control food) or nicotine food. Nicotine (Sigma) was directly dissolved in molten cornmeal media food just before pouring into empty plastic vials at concentration of 3 ml/ml. After 8 h of chronic exposure to the nicotine food, whole flies were collected and immediately placed in liquid nitrogen.

the basis of a dominant Humeral marker. Survival time on 3 ml/ml nicotine was recorded at 6 h intervals over 5 d for three replicate vials of ten 3- to 6-d-old male or female flies of each genotype class. Crossing of line means was assessed by three-way ANOVA on vial means with fixed effects of line, genotype and sex. Metabolite profiling. Concentrations of ornithine and g-aminobutyric acid (GABA) were determined by LC-MS/MS on a Thermo LTQ mass spectrometer using extracts from five adult female heads (Supplementary Methods). Nicotine and cotinine determinations were also performed by LC-MS on parallel samples of five heads processed independently. PCR primers. Primers for quantitative real-time PCR are listed in Supplementary Table 3 online. Accession codes. All expression data are available from the MIAME-compliant public repository at ArrayExpress (expression profile E-TABM-109) and on the website listed below. URLs. ArrayExpress (http://www.ebi.ac.uk/arrayexpress). Expression data are also available from http://statgen.ncsu.edu/ggibson/SupplInfo/ SupplInfo13.htm. Note: Supplementary information is available on the Nature Genetics website.

RNA preparation, labeling and hybridization. Total RNA from each line and drug treatment was extracted with Trizol reagent (Invitrogen) from about 400 North Carolina and 30 California fly heads. For experiments using total RNA from North Carolina lines, 10 mg of total RNA was used as a template for each of two replicate cDNA synthesis reactions using Im-Prom II (Promega). Indirect labeling was carried out with aminoallyl-linked Cy3 or Cy5 esters after cDNA synthesis. Labeled cDNAs were hybridized onto 2 cm  2 cm arrays for 15 h at 42 1C. After high-stringency washes, the slides were scanned with a ScanArray 4000 Microarray Analysis System scanner (Packard Bioscience), and images were processed using UCSF Spot software24. For experiments using amplified RNA from North Carolina and California lines, total RNA was used for linear amplification of RNA using the Agilent Low-Input Fluorescent Linear Amplification Kit (Agilent Technologies).

ACKNOWLEDGMENTS We thank S. Nuzhdin for providing the California inbred lines and R. Carrillo for initial sampling of nicotine resistance. This paper is dedicated to the memory of R. Rose, who first pointed out the lindane connection in our data. This work was supported by US National Institutes of Health grant P01-GM45344 to G.G.

Microarray preparation and experimental design. The cDNA microarrays comprised 4,385 ESTs arrayed on aminosilane-coated glass slides (Corning GAPS II). The experiments consisted of three independent loop designs25, one for each population (North Carolina and California) and RNA source (for North Carolina, total and amplified RNA; for California, amplified RNA only). In these experiments, each sample was replicated twice (once labeled with Cy3, once with Cy5). RNA from line 1 in control food (labeled with Cy3) was hybridized with RNA from line 2 in nicotine food (labeled with Cy5) on the first slide; the same line 2 (now labeled with Cy3) was compared with line 3 in control food (Cy5) on the second slide, and so on. A total of 332 hybridizations involving 664 samples were analyzed.

Published online at http://www.nature.com/naturegenetics Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions

Statistical analyses. Analyses for individual populations were carried out using two sequential ANOVA models26,27 using PROC MIXED in SAS as described in detail in the Supplementary Methods. For the combined analysis of amplified RNA from North Carolina and California lines, Population and its interactions terms Population-by-Drug and Population-by-Dye were introduced into both mixed-models ANOVA to account for the genotypic differences between the two populations. Normalized quantitative least-square mean measures of transcript levels were estimated for each individual line and drug treatment for each gene, and drug response was measured as survival time (in hours) on nicotine food for each individual line. Estimates of expression were then applied to regression of control- or nicotine-induced expression on nicotine survival time using the PROC REG statement in SAS version 8. Complementation testing. The CG8745e00991 allele in Bloomington Stock 17912 carries a lethal insertion of a piggyBac{RB} element into the 3¢ UTR of the gene20. Eight inbred wild-type North Carolina lines were selected to cover the range of nicotine sensitivities. The cross of Mutant / TM6B Balancer by wild-type North Carolina line yields heterozygous Mutant / North Carolina and wild-type Balancer / North Carolina progeny that were distinguished on

NATURE GENETICS VOLUME 39

[

NUMBER 2

[

FEBRUARY 2007

AUTHOR CONTRIBUTIONS G.P.-G. performed all of the data analysis and experimental components, with the exception of the genetic complementation test (P.H.) and metabolite profiling (N.D.). She was assisted by W.-P.H. in the statistical analysis. G.G. conceived the experiment and wrote the manuscript. COMPETING INTERESTS STATEMENT The authors declare that they have no competing financial interests.

1. Liotta, L. & Petricoin, E. Molecular profiling of human cancer. Nat. Rev. Genet. 1, 48–56 (2000). 2. de Koning, D.J. & Haley, C.S. Genetical genomics in humans and model organisms. Trends Genet. 21, 377–381 (2005). 3. Mackay, T.F. The genetic architecture of quantitative traits. Annu. Rev. Genet. 35, 303–339 (2001). 4. Cohen, C., Welzl, H. & Battig, K. Effects of nicotine, caffeine and their combination on locomotor activity in rats. Pharmacol. Biochem. Behav. 40, 121–123 (1991). 5. Wolf, F.W. & Heberlein, U. Invertebrate models of drug abuse. J. Neurobiol. 54, 161–178 (2003). 6. Carrillo, R. & Gibson, G. Unusual genetic architecture of natural variation affecting drug resistance in Drosophila melanogaster. Genet. Res. 80, 205–213 (2002). 7. Jessen, A., Buemann, B., Toubro, S., Skovgaard, I.M. & Astrup, A. The appetitesuppressant effect of nicotine is enhanced by caffeine. Diabetes Obes. Metab. 7, 327–333 (2005). 8. Palsson, A. & Gibson, G. Association between nucleotide variation in Egfr and wing shape in Drosophila melanogaster. Genetics 167, 1187–1198 (2004). 9. Dworkin, I., Palsson, A., Birdsall, K. & Gibson, G. Evidence that Egfr contributes to cryptic genetic variation for photoreceptor determination in natural populations of Drosophila melanogaster. Curr. Biol. 13, 1888–1893 (2003). 10. Macdonald, S.J., Pastinen, T. & Long, A.D. The effect of polymorphisms in the Enhancer of split gene complex on bristle number variation in a large wild-caught cohort of Drosophila melanogaster. Genetics 171, 1741–1756 (2005). 11. De Luca, M. et al. Dopa decarboxylase (Ddc) affects variation in Drosophila longevity. Nat. Genet. 34, 429–433 (2003). 12. Seiler, N. Ornithine aminotransferase, a potential target for the treatment of hyperammonemias. Curr. Drug Targets 1, 119–153 (2000). 13. Kelly, A. & Stanley, C.A. Disorders of glutamate metabolism. Ment. Retard. Dev. Disabil. Res. Rev. 7, 287–295 (2001). 14. Sivilotti, L. & Nistri, A. GABA receptor mechanisms in the central nervous system. Prog. Neurobiol. 36, 35–92 (1991).

267

© 2007 Nature Publishing Group http://www.nature.com/naturegenetics

LETTERS 15. Beleboni, R.O. et al. Pharmacological and biochemical aspects of GABAergic neurotransmission: pathological and neuropsychobiological relationships. Cell. Mol. Neurobiol. 24, 707–728 (2004). 16. Livak, K.J. & Schmittgen, T.D. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)). Methods 25, 402–408 (2001). 17. Vale, C., Damgaard, I., Sunol, C., Rodriguez-Farre, E. & Schousboe, A. Cytotoxic action of lindane in neocortical GABAergic neurons is primarily mediated by interaction with flunitrazepam-sensitive GABA(A) receptors. J. Neurosci. Res. 52, 276–285 (1998). 18. Parmar, D. et al. Effect of lindane on hepatic and brain cytochrome P450s and influence of P450 modulation in lindane induced neurotoxicity. Food Chem. Toxicol. 41, 1077–1087 (2003). 19. Tomfohr, J., Lu, J. & Kepler, T.B. Pathway level analysis of gene expression using singular value decomposition. BMC Bioinformatics 6, 225 (2005). 20. Thibault, S.T. et al. A complementary transposon toolkit for Drosophila melanogaster using P and piggyBac. Nat. Genet. 36, 283–287 (2004).

268

21. Yoneda, Y., Roberts, E. & Dietz, G.W., Jr. A new synaptosomal biosynthetic pathway of glutamate and GABA from ornithine and its negative feedback inhibition by GABA. J. Neurochem. 38, 1686–1694 (1982). 22. Palsson, A., Rouse, A., Riley-Berger, R., Dworkin, I. & Gibson, G. Nucleotide variation in the Egfr locus of Drosophila melanogaster. Genetics 167, 1199–1212 (2004). 23. Yang, H.P. & Nuzhdin, S.V. Fitness costs of Doc expression are insufficient to stabilize its copy number in Drosophila melanogaster. Mol. Biol. Evol. 20, 800–804 (2003). 24. Jain, A.N. et al. Fully automatic quantification of microarray image data. Genome Res. 12, 325–332 (2002). 25. Churchill, G.A. Fundamentals of experimental design for cDNA microarrays. Nat. Genet. 32(Suppl.), 490–495 (2002). 26. Jin, W. et al. The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster. Nat. Genet. 29, 389–395 (2001). 27. Wolfinger, R.D. et al. Assessing gene significance from cDNA microarray expression data via mixed models. J. Comput. Biol. 8, 625–637 (2001).

VOLUME 39

[

NUMBER 2

[

FEBRUARY 2007 NATURE GENETICS