“Plasticity and Gene by Environment interactions in complex phenotypes of adult Drosophila melanogaster”

By Clement F. Kent III

A thesis submitted in conformity with the requirements for the degree of Ph.D., Graduate Department of Ecology and Evolutionary Biology, in the University of Toronto

© Copyright by Clement F. Kent III , 2009

“Plasticity and Gene by Environment interactions in complex phenotypes of adult Drosophila melanogaster” A thesis submitted by Clement F. Kent III in conformity with the requirements for the degree of Ph.D., Graduate Department of Ecology and Evolutionary Biology, in the University of Toronto, 2009.

Abstract Behaviour genetics deals with complex phenotypes which respond flexibly to environments animals experience. Change of phenotype in response to environment is phenotypic plasticity. A central question is how genes influence plasticity. I study plasticity and gene by environment interactions (GEI) relating to behaviours, metabolic, and genomic phenotypes of adults of the fruit fly Drosophila melanogaster. Chapters 1-3 study cuticular hydrocarbon (CH) levels of male flies. Chapter 1 shows male CH levels respond to time of day and light. Methods are developed to reduce high variability of CH. I show variation in CH parallels activity of two classes of CH synthesis hormones. Analysis of rate of variation gives estimates of turnover rates of CH and the metabolic cost of signaling. Chapter 2 studies mixed groups of genetically different flies, “hosts” and “visitors”. GEI of CH are found with both abiotic factors and with social mix. Social mix results in GEI as strong as abiotic factors. Indirect Genetic Effects (IGE) theory is used to show frequencydependent IGE interactions. Chapter 3 shows that males in mixed social environments reduce expression of clock and CH synthesis genes, resulting in different signals. Females mate more often with males in a mixed group than with single-genotype males. Plasticity in male gene expression in response to social environment leads to different signals, mating levels, and potentially different fitness. Chapter 4 deals with behaviour, metabolite, and genomic phenotypes in flies differing in foraging gene alleles, as the food environment is changed. Strong GEI is found, structured by food type, chemical class of metabolite, and gene metabolic roles. A concept called “relative nutrient sensitivity” suggests an interaction between foraging and the insulin signaling pathway. I demonstrate epistasis between for and insulin with quantitative genetic methods and bioinformatics. These results lead to the conclusion that GEI are common in many fly phenotypes in response to well studied environments such as food and less studied ones such as social group. Some implications of this for maintenance of genetic variance are discussed.

ii

Acknowledgments I thank Prof. Marla B. Sokolowski for support and encouragement throughout, and Prof. Joel Levine for stimulating discussions and steadfast belief in the quality of our work. I thank committee members Prof. Gary Sprules and Prof. Timothy Westwood for comments and advice, all external reviewers of the papers making up Chapters 1-4 for their comments, and of course my many collaborators in this work. Prof. Allen Moore asked stimulating questions about plasticity and was a constructive external examiner. I thank the many members of the Sokolowski and Levine labs whose advice and help made this work far more pleasant. Thanks for all the flies! Thanks to Karen Williams for many thoughtful discussions on insulin and flies; to Amsale Belay and Karla Kaun for lots of help with learning fly wrangling and understanding fly metabolism; to Ken Dawson-Scully for many scientific discussions ranging from ion channels to stress-related behaviours to computer geekdom; to Craig Riedl for computer wizardry and fly breeding schemes; to Hiwote Belay for indispensable help in the molecular lab, to Hiwote and Mark Fitzpatrick for rubber bands and other practical jokes, and to Mark for great references on evolutionary topics; to Scott Douglas for fly stocks and listening to my complaints on same; and especially to Brandon Sheffield and Bianco Marco for indispensable technical help with flies, behaviours, and spectrophotometer assays. Special thanks to Reza Azanchi for a ton of work and unfailing good humour, to Ben Smith, Joshua Krupp for the steadiest hands on the planet, Julia Schonfeld, Jayed Atallah, Amanda Formosa, Adrienne Chu, and Olga Sizova for GC advice and data, and putting up with weird R printouts. Thanks to Jean-Christophe Billeter for many discussions and for help with Adobe Illustrator.

Special thanks to Prof. Ralph Greenspan, whose ideas and data jumpstarted this research and whose patience in awaiting results was exemplary. Thanks for scientific advice and feedback on talks go to Prof. Mark Blows and Prof. Stephen Goodwin, and for a few words on fly sex appeal to Dr. Jean-Marc Jallon.

Patience was also noteworthy from several people: my partner Leena Raudvee whose unfailing support and listening post was essential, my parents Elizabeth Casey Kent and Clement F. Kent Jr. whose early teachings in biology and mathematics helped set me on this path, and to Prof. Roger Hansell, Prof. Nick Collins, and especially the late Prof. Jyri Paloheimo who helped guide an earlier foray into ecology.

Special thanks are due to long time friend and collaborator Jonathan Wong for listening, encouragement, and ideas: who else understood vapor trails in hyperspace?

I acknowledge support from Grant DK70141-2 from NIDDK to MBS, support from the Department of Ecology and Evolutionary Biology at University of Toronto, and to all faculty, staff, and students of the Department of Biology at University of Toronto Mississauga.

iii

Last but not least, I acknowledge the many volunteers who donate their time and expertise to several outcroppings of the “open information” movement: the editors at PLoS journals, the maintainers of GEO Express, Homologene, UCSC Genome Browser, Flybase, KeGG, Wormbase, and SGD, and most of all the creators of R and the many packages available in it.

iv

Table of Contents Abstract

ii

Acknowledgements

iii

Table of Contents

v

List of Figures

vi

List of Tables

viii

List of Abbreviations

x

Introduction

1

Chapter 1

84

Chapter 2

126

Chapter 3

167

Chapter 4

208

Discussion

255

v

List of Figures Introduction I.1 Major insulin signaling pathway components I.2 Insulin signaling and glycogen I.3 Regulation of Insulin and Tor signaling Chapter 1 1.1 Distribution of individual fly total abundance (TA) values 1.2 Test of RA model assumptions 1.3 Abundance Variability AVj for 21 compounds in wild-type flies in DD 1.4 Cluster analyses of 22 CHs in LD and DD 1.5 Abundance Variability and effect of light 1.6 Compound TA-mean abundance ratio, LD/DD 1.7 Difference between subjective day and night FA-normalized mean compound abundance 1.8 Diurnal changes in compounds 1.9 Difference between LD and DD at each hour 1.10 Abundance Variability versus time of day for monoenes 1.11 Effect of LD AV variation at CT13 1.12 Desaturation Index DI versus total hydrocarbon abundance TA over 24 hours in LD and DD 1.13 Temporal variation for representative compounds in LD and DD 1.14 Loss rate of applied 9-C25:1 in LD Chapter 2 2.1 Overview of Cuticular Hydrocarbon Types and Synthesis Pathways, and Design of Group Effects Experiment 2.2 Social and Genotype by Environment (GxE) effects 2.3 Group composition changes Principal Coordinate levels and time-course 2.4 Indirect genetic effects of wild-type flies depends on social context S2.1 Hourly levels of four cuticular hydrocarbons show that passive transfer cannot explain observed social effects S2.2 Levels of methyl-branched alkane 2-MeC28 in DD S2.3 Clustering analyses S2.4. Social Effects versus PC2 S2.5. Social effects on Percent Monoenes S2.6 Distribution of individual fly total abundance S2.7 Light effects versus PC1 S2.8 Observed versus predicted direction of social effect S2.9 Hourly Host-Control differences in DD S2.10 Isolates-Communals PC2 and PC1 show social effects S2.11 Indirect genetic effects in social contexts S2.12. Fitted versus true Ψ S2.13. Match of corrected and true Ψ Chapter 3 3.1 Cytology of Adult Drosophila Oenocytes 3.2 Oenocytes Contain a per-Dependent Clock 3.3 Cuticular Hydrocarbon Accumulation is Regulated by a per-Dependent Clock and is Influenced by Exposure to Light 3.4 Cuticular Hydrocarbon Accumulation is Regulated by the Expression Level of desat1 3.5 Social Interactions Affect the Temporal Profile of Gene Expression and Cuticular Hydrocarbon Accumulation 3.6 Social Context Changes the Amount and Temporal Distribution of Mating S3.1 Expression Pattern of the Gal4 Driver, 1407-Gal4 S3.2 Clock Gene Expression is Cyclic in Heads with an Advanced Phase Relationship Relative to that of the Oenocytes

vi

Page 48 50 52 86 87 90 93 96 97 99 100 102 103 104 106 108 110 128 129 131 135 142 146 148 150 151 152 153 155 156 157 160 163 165 170 171 172 175 177 180 191 192

S3.3. A Peripheral Clock Mechanism is Required for Gene Expression and Cuticular Hydrocarbon Accumulation S3.4 Temporal pattern of DESAT1 expression in the oenocytes S3.5 7-Tricosene Accumulation is Regulated by the Expression Level of desat1 S3.6 The Effects of Social Interactions on Cuticular Hydrocarbon Profiles are Reduced in LD S3.7 Visitors are Affected by Social Context Chapter 4 4.1 Behavioral foraging gene by food interaction 4.2 Rovers and sitters use energy stores differently 4.3 Fed rovers and sitters store energy differently 4.4 Transcriptional interactions between foraging alleles and food 4.5 foraging GEI is due to plasticity differences: rovers respond more to food than sitters 4.6 Insulin pathway genes interact with foraging alleles in expression and in food-leaving assay 4.7. Meta-analysis of 3 manipulations of the insulin/Tor signaling identifies rover-biased genes S4.1 Behaviour testing apparatus S4.2 Positive regulators of insulin signaling – rovers change expression more than sitters S4.3 qRTPCR results Discussion D-1 Barton-Turelli K for rover-sitter genes D-2 Haldane-Jayakar model of larval fitness

vii

193 194 195 196 197 211 212 213 215 218 220 223 234 239 247 272 274

List of Tables Chapter 1 1.1 Percent of within-hours variance removed by normalization methods 1.2 Cluster Membership Chapter 2 T2.1 η2 effect sizes in factorial ANOVA T2.2 Hosts – Control differences during virtual day and night in DD T2.3 Cluster Membership T2.4 Social effects on Host-wild type Principal Coordinates T2.5 Host-Control differences during CT 12-17 in DD T2.6 Social effects in Isolates-Communals experiment T2.7 Levels of IGE coefficient Ψ in Host-Visitors and Controls Chapter 3 S3.1. Fitted Cosine Curves for Normalized Relative RNA Expression Levels from Oenocytes S3.2. Phase Differences Between Clock Genes in Oenocytes RNA Expression Level Differences Between LD and DD for Wild Type Oenocytes, and Between Wild Type and per0 Oenocytes in DD S3.4. Fitted Cosine Curves for Wild Type Controls and Hosts in Oenocytes and Heads S3.5. Phase Differences Between Wild Type and Hosts in Oenocytes and Heads S3.6. Expression Level Differences Between Wild Type and Hosts in Oenocytes and Heads S3.7. Phase Differences in Clock Gene Expression Between Oenocytes and Heads Chapter 4 T4.1. Analysis of variance of behaviour and four food media T4.2. FTICR MS metabolite data T4.3 Gene groups with significant GxE interactions T4.4. Relative Nutrient Sensitivity (RNS) for metabolites T4.5. Complementation cross analysis of variance for insulin mutants T4.6. Meta-analysis: Genes that respond to insulin differ between rovers and sitters Discussion D-1. Genetic covariance between lipid and carbohydrate gene expression

viii

Page 88 92 140 144 147 149 154 158 159 198 199 200 201 202 202 203 235 238 240 243 244 245 280

List of Abbreviations Abbreviation Meaning D. melanogaster adipokinetic hormone gene, no human homolog but analgous to Akh AKT, Akt1 alkane BKCa cAMP CBP cGKI, cGKII cGMP CH CHPA CR CREB DAG desat1 dilp Dp110, Dp110B egl-4 ER for forR fors fors2 foxo GC-FID GC-FTMS GEI GlyP GO GxE

glucagon; product is a hormone acting to stimulate fat and glycogen breakdown in fat body v-akt murine thymoma viral oncogene homolog, a kinase involved in insulin signaling cascade with growth and proliferation phenotypes; human oncogene AKT2 in the context of CH, refers to linear (non-branched) alkane compound big potassium conductance calcium-activated ion channel; KCNMA1 is official human gene name for unit whose fly homolog is slo, slowpoke cyclic AMP CREB-binding protein. Transcriptional co-activator of CREB and histone acetyltransferase, fly nej, nejire, human EP300 or p300. cyclic GMP-dependent kinase I; human and mouse homolog of for; cGKII is homolog of fly PKG21D cyclic GMP Cuticular Hydrocarbon Constrained Hierarchical Parsimony ANOVA – systems of linear equations which most parsimoniously describe data patterns, constrained in their hierarchy to a priori known groups or pathways of genes Caloric restriction cAMP responsive element binding protein; transcription factor. Fly CrebB-17A, human CREB1I. Transduces cAMP/PKA signaling to gene expression. diacylglycerol; intermediate in synthesis/breakdown of TAG and phospholipids; DAG has signaling properties (see PIP2, IP3) D. melanogaster desaturase 1 gene, homologous to mouse Scd2, stearoyl-Coenzyme A desaturase 2; product is fatty acid desaturase enzyme drosophila insulin-like peptide; dilp1-dilp7 encode these. synonym for Pi3K92E gene. Dp110B is a mutant hypomorph allele. egg-laying defective 4; C. elegans homolog of for endoplasmic reticulum D. melanogaster foraging gene, homologous to human PRKGI; product is cGMPdependent protein kinase, type I Allele of foraging gene found in natural rovers; high PKG activity Allele of foraging gene found in natural sitters; lower PKG activity Allele of foraging gene found in mutant sitters; low PKG activity D. melanogaster foxo gene, homologous to human FOXO3. product, transcriptional inhibitory member of insulin signaling cascade Gas Chromatograph Flame IDentification Gas Chromatograph Fourier Transform Mass Spectroscopy Gene by Environment Interaction(s) D. melanogaster glycogen phsophorylase gene, homologous to human PYGM, stearoylproduct is a rate limiting enzyme in glycogen breakdown Gene Ontology; a GO group defines a set of genes annotated as being related to a single common biological pathway, molecular, or cellular component. Gene by Environment

ix

HBB I IGE InR, InR IP3 KATP KEGG LTD, LTP MAG MB methyl MOT1 monoene MW NC2 nej NFAT NO PDE per, per0 PGC-1α Pi3K92E PIP3, PIP2, etc PKA PKG PLC PP2A PRKG1 Pten QTL RI RNS ROS SERCA sGC slo

human hemoglobin beta gene A measure of magnitude and direction of gene by environment interactions Indirect Genetic Effect; effect of genes in another individual on phenotype/fitness of a focal individual D. melanogaster InR gene, homologous to human IGF1R; product is insulin-like receptor inositol 1,4,5-triphosphate; product of hydrolysis of PIP2 by PLC signals to various pathways ATP activated potassium channel Kyoto Encyclopedia of Genes and Genomes; an online database structured around Enzyme Commission (EC) chemical reaction definitions. long term depression and long term potentiation of synapses monoacylglycerol; intermediate in synthesis/breakdown of TAG and phospholipids mushroom body in the context of CH, refers to methyl-branched alkane compound yeast gene name for transcription factor which antagonizes TBP activity; homologs are human BTAF1 and fly Hel89B in the context of CH, refers to an alkane which has been modified by the insertion of one double bond in the carbon chain, typically in positions 5,7, or 9. Molecular Weight (in Daltons) A heterodimeric transcription factor associated with Mot1 and DPE, q.v. nejire, fly homolog of CBP, q.v. nuclear factor of activated T-cells, transcription factor. synonyms NF/AT, NF-AT nitric oxide phosphodiesterase; PDE1-12 are cAMP- or cGMP-PDE’s, PDE5 is primary cGMP PDE D. melanogaster period gene, homologous to human Per1, Per2, Per3. Core member of circadian clock. per0 is a mutant allele which disrupts clock function. peroxisome proliferator-activated receptor gamma, coactivator 1 alpha; official human gene name is PPARGC1A D. melanogaster Pi3K92E gene, homologous to human PI3KCD. product phosphoinositide 3-kinase, member of insulin signaling cascade phosphatidylinositol 3,4,5-triphosphate; phosphatidylinositol 4,5-bisphosphate; phospholipids associated with insulin and many other signaling cascades Protein Kinase A; cAMP-dependent protein kinase Protein Kinase G; cGMP-dependent protein kinase phospholipase C breaks PIP2 into DAG and IP3; major signaling pathway protein phosphatase 2A protein kinase G1; official human gene name for PKG form homologous to for D. melanogaster Pten gene, homologous to human PTEN. product phosphatase and tensin homolog, inhibitory member of insulin signaling cascade quantitative trait locus or loci Recombinant inbred [lines] Relative Nutrient Sensitivity Reactive oxygen species Sarco-Endoplasmic Reticulum Calcium ATPase soluble guanylyl cyclase enzyme produces cGMP when stimulated by NO slowpoke, fly ion channel gene; see BKCa

x

S/N SCD TAG tax-2, tax-4 Tbp TCA TGF-β Trp VASP Ψ, ψa,b

Signal to Noise ratio Sickle-cell disease triacylglycerol; primary constituent of lipid droplets in fat cells abnormal chemotaxis 2,4, C. elegans genes for cGMP gated channel TATA binding protein transcription factor tricarboxylic acid cycle, syn. Krebs cycle transforming growth factor beta transient receptor potential calcium channel gene, also a class of such channels vasodilator-stimulated phosphoprotein; fly homolog is ena, enabled measure of interaction strength between two individuals in IGE theory; ψa,b is effect of genotype b on genotype a; Ψ is a matrix of effects of multiple genotypes on each other. Dimensionless value between -1 and 1, with 0 indicating no effect.

xi

Introduction This thesis examines complex phenotypes of adult fruit flies (Drosophila melanogaster), in the context of gene by environment (GxE) interactions (GEI) and phenotypic plasticity. The phenotypes examined include gene expression data, metabolite and other physiological data, and behavioural data. The environmental influences investigated include nutrition, light cycle, and the composition of the group in which flies live. I apply mathematical models, multivariate statistical techniques, and both traditional and new experimental designs (a) to identify where important new biology can be found within the information overload of very large numerical datasets, (b) to test new hypotheses using large datasets, and (c) to develop new, more powerful techniques experimenters can apply to the large, rich datasets now available. I am a theoretical biologist who works closely with experimenters. This thesis therefore uses fairly abstract or mathematical methods at times to look at very concrete biological problems. A goal of the thesis therefore is to present to the reader not just a number of interesting results in biology but also some useful approaches to finding such results in large datasets.

It’s traditional in a thesis introduction to restrict oneself to prior results and theory in the fields studied, then to discuss results and models in the Discussion. However, I have been fortunate to work with two supervisors with divergent interests, and to be able to work on topics that intersect many specialized areas of modern biology. There are multiple threads in this thesis. To provide context about these threads and how they are woven together in this thesis, I first present a brief overview of the main areas of investigation. The purpose of this overview is to motivate the review later in the Introduction of seemingly disparate results from realms as diverse as economics, probability theory, cell and systems biology, and genetics, whose relevance might not otherwise be obvious. A unifying theme here is that this a work of theoretical biology applied to large amounts of experimental data, so I indicate what kinds of theoretical approaches are available and how they help generate and test hypotheses about the experimental data. This provides a context which I hope helps readers recognize common theoretical models applied to quite different kinds of data. Finally, I introduce the biological literature for GEI, IGE, fly cuticular hydrocarbons, circadian rhythms, mating, foraging, nutrition and insulin-like signaling. As is clear from this long list of topics, a full literature survey for all these topics would require hundreds of pages of Introduction. I ask the reader to therefore bear in mind that references are those most relevant to the work of the thesis, with an effort being made to cite recent results, reviews, and especially influential older papers.

Overview of work – social interactions In my work with Prof. Levine, I examine chemical signaling of adult male flies via cuticular hydrocarbon (CH) profiles. CH chemical signals are involved in signaling within and between the sexes

1

(Jallon 1984; Wyatt 2003; Ferveur 2005). CH signals help determine whether mating will happen among conspecifics, act as barriers between species (Cobb and Jallon 1990; Marcillac, Houot et al. 2005), and are used as aggregation pheromones among females. Drosophila mating has been viewed as a prototypical example of stereotyped behaviors which are especially amenable genetic dissection (Greenspan and Ferveur 2000). The converse question is, how responsive to environmental influences are fly mating behaviours? (Svetec and Ferveur 2005; Rundle, Chenoweth et al. 2006; Ruedi and Hughes 2008) Focusing on the subset of mating behaviours represented by the complex suite of CH signals, I ask how these vary in response to abiotic and biotic environmental factors (light, and social interactions of flies with other flies in their local group). We use the complex phenotype of the expression levels of 24 CH compounds over 24 hours of the day, and show that although it has strong gene by abiotic environment interactions, gene by social environment interactions are even stronger.

The demonstration that that the social environment (the number and frequency of flies of different genotypes in the group) has strong main and interaction effects on male chemical signals raises the possibility that the social environment may therefore effect both male-male and male-female interactions. With collaborator Joshua Krupp, we demonstrate that male-male interactions change depending on the group composition, not only in the final expressed signal (CH levels) but also in the expression of a key gene (desat1) which regulates the production of CH signals (Marcillac, Bousquet et al. 2005). I use multivariate methods to find that the majority of variation in CH signals could be attributed to two principle coordinate (PC) axes of variation. One axis maps corresponds to levels of compounds affected by desat1, while the other axis corresponds to the actions of elongase enzymes, which lengthen the carbon backbones of CHs. Excitingly, the two PC axes respond quite differently to social environment, with the desat1influenced axis showing main effects of social group, while the elongase-influenced axis shows strong genotype by social environment interactions.

This GEI analysis also shows that the branched methylalkane (“methyls”) group of CH compounds responds to social environment. The laboratory of Mark Blows implicated these compounds as pheromones in the Australian Drosophila serrata (Blows, Chenoweth et al. 2004; Petfield, Chenoweth et al. 2005; Rundle, Chenoweth et al. 2005; Skroblin and Blows 2006) and finds that there is heritable variation for D. melanogaster methyl production (Foley, Chenoweth et al. 2007). Our results represent the first demonstration that methyls may play a role in D. melanogaster chemical communication.

Given that we have shown that the social group influences expression of genes important in CH production, and that CH signals from the group are perceived by others in the group and may affect their behaviour, we have a recursive cause and effect loop between genes in group members, their signals to others, and changes in gene expression and behaviour of other group members.

2

Resolving the net or final effects of such loops or cycles of cause and effect was a concern of Sewall Wright, who developed the method of path coefficients to solve complex equations of interdependent causes and effects (Wright 1921; Wright 1934). Wright was primarily concerned to solve for the net effect of multi-gene systems within a single individual. Extended later by Ronald Fisher into a more comprehensive mathematical theory (Fisher 1958), Wright’s insights led Hamilton to new applications to the evolution of social behaviour under the name of kin selection (Hamilton 1964). In kin selection, the inclusive fitness impacts of behaviours on the animal behaving and on its kin of various degrees of relatedness determine the net evolutionary effect of a behaviour. Kin selection and inclusive fitness theory was able to solve problems in the evolution of social behaviours first identified as major challenges for evolutionary theory by Darwin.

In the 1980’s and 1990’s, work by Cheverud, Moore, and collaborators extended kin selection arguments to maternal care (Cheverud 1984), sibling competition and cooperation (Cheverud and Moore 1994), and finally to interactions of groups of unrelated individuals (Moore, Brodie et al. 1997). These insights were crystallized in an important series of papers under the names “Interacting Phenotype Effects” and “Indirect Genetic Effects” (IGE) (Moore, Brodie et al. 1997; Wolf, Brodie et al. 1998; Agrawal, Brodie Iii et al. 2001; Moore, Haynes et al. 2002). The key result from IGE theory is that the equations defining the effects of individuals of different genotypes on each others’ phenotypes can be resolved in a conceptually simple form using the method of path coefficients. In IGE theory, these path coefficients are represented by the Greek letter ψ, which captures the strength and direction of the interaction between two individuals (Moore, Brodie et al. 1997). When genotypes a and b interact, IGE theory allows for the possibility that the effect of b on a (ψa,b) may be different from the effect of a on b (ψb,a). For n interacting genotypes there is an n,n matrix Ψ of all pairwise interaction strengths. Just as Lande’s G-matrix crucially determines evolution of genetically correlated traits (Lande 1979; Lande and Arnold 1983), the Ψ matrix determines evolution of behaviourally interacting genotypes. This crucial discovery makes clear how social interaction can coevolve with the underlying genotypes which influence it.

Male flies interact with each other in groups (McRobert and Tompkins 1983; Levine, Funes et al. 2002; Kim, Phillips et al. 2004; Svetec, Cobb et al. 2005; Svetec and Ferveur 2005) and change chemical signals (Krupp, Kent et al. 2008). Thus, IGE theory should apply to this behavioural interaction. However, methodological difficulties exist. The assays which determine CH levels kill the fly. The original IGE papers were based on a repeated-measures model (observe the focal animal before and after interaction with another and measure how its phenotype has changed) (Moore, Brodie et al. 1997; Wolf, Brodie et al. 1998). To apply IGE theory to one-time only fly assays, I develop an extension of IGE theory to describe how to estimate ψ coefficients from measurements of multiple individuals within a group, repeated over multiple groups. In addition, our research demonstrates that there are temporal patterns CH expression, so I further extend IGE equations to cases in which expressed phenotype varies over time. Solution of these

3

extended IGE equations involves terms nonlinear in ψ so I develop the sampling theory and mathematical techniques to estimate true ψ from data.

Appying these extended IGE equations to CH data for interacting groups of male flies, I find that the ψ values found in single-genotype groups are different from the same ψ coefficients for males in a mixed-genotype group. Thus, the genotype-genotype interaction strength ψ depends on the genotypic composition of the group. In one case a compound which is known to influence male- male interactions had same genotype ψ switch from positive in a single-genotype group (males emulate signals of other males) to negative in mixed groups (males avoid the signal levels of same-genotype males). And, the betweengenotype ψ for the same compound was significantly positive. This is consistent with males changing their CH response to others from emulation of own genotype to emulation of “stranger” genotypes. This pattern holds over many CH compounds, especially including the methyls the Blows group finds important to D. serrata behaviour (Petfield, Chenoweth et al. 2005). Hence, IGE theory needs to encompass frequency dependent IGE phenotypic effects.

Application of IGE theory to evolution requires measures of the fitness of traits affected by indirect phenotypic effects. We examine matings of females with males from single- and mixed-genotype groups and find that matings increase in the mixed compared to the single genotype groups (Krupp, Kent et al. 2008). This suggests that interactions in mixed groups affect male fitness, although more work is required to confirm this.

This overview of material presented in the first three chapters intersects fly biology in four areas: cuticular hydrocarbons, circadian rhythms, mating, and social behaviours. So, later in the Introduction previous publications from these four areas will be discussed.

Overview of work – rovers and sitters In my work with Prof. Sokolowski, I examine genomic, physiological, and behavioural responses of adult flies with differing foraging (for) gene alleles to their nutritional environment. A much more in depth review of foraging is presented later in this introduction. Here I note briefly that Sokolowski isolated a naturally occurring gene affecting larval locomotion in the presence of food, one of the earlier demonstrations of natural genetic variation for behaviour caused by a single gene (Sokolowski 1980; Sokolowski and Hansell 1983; Sokolowski, Kent et al. 1983). Sokolowski and colleagues determined that for encodes protein kinase G (PKG) and that the natural variant with higher PKG activity (rovers) has a suite of behaviours differing from those with the lower PKG variant; a mutant of for in the rover background causing it to become a sitter confirmed the identification of gene and behavioural effects (Osborne, Robichon et al. 1997). Since then for has been shown to have other behavioural effects in flies (Engel, Xie et al. 2000; Scheiner, Sokolowski et al. 2004; Mery, Belay et al. 2006; Kaun, Hendel et al.

4

2007), and PKG-encoding genes are associated with behaviour differences in worms (Raizen, Cullison et al. 2006; Hallem and Sternberg 2008; Raizen, Zimmerman et al. 2008; van der Linden, Wiener et al. 2008; You, Kim et al. 2008), bees (Ben-Shahar, Robichon et al. 2002), and ants (Ingram, Oefner et al. 2005).

I show in Chapter 4 that trends in behavioral plasticity in response to food availability and type are explained by both genetic (foraging allele type) and environmental (food quality) factors, and that there are strong GEI in these. From these results, I hypothesized that foraging may interact with the insulin pathway. Applying both bioinformatic and direct experimental methods, I demonstrate that there is a strong interaction between foraging and insulin genes.

I will show that a concept related to but importantly different from the classical analysis of variance based GEI better explains much of the observed differences between foraging strains. This concept, Relative Nutrient Sensitivity (RNS), captures the differences in sensitivity of the rover and sitter genotypes to changes in food availability. Such differential sensitivity to food is seen in human disorders such as metabolic syndrome and diabetes (Schwartz and Porte 2005; Marshall 2006; O'Rahilly and Farooqi 2006). The demonstration that foraging interacts with insulin-related genes and that rovers show generally higher RNS (are more responsive to food changes) than sitters suggests fruitful future directions of research into whether rovers and sitters may represent a model system in which to study metabolic disorders.

Since the foraging polymorphism is naturally occurring and stably maintained in the wild and in the lab (Sokolowski 1980; Sokolowski and Hansell 1983; Fitzpatrick, Feder et al. 2007), the RNS dichotomy between rovers and sitters may have ecological and evolutionary implications. I demonstrate that rovers and sitters have consistent differences in allocation of energy stores between carbohydrates and lipids and discuss potential ecological consequences of this difference. These differences between lipids and carbohydrates are mirrored in a wider analysis of metabolites which reveals broad differences between rovers and sitters which follow the RNS model.

Although metabolic gene pathways show some of the strongest, clearest signals of rover-sitter GEI and RNS differences, the question remains of whether and how this influences behaviour. Using genomic data I show that pathways of synthesis of “secretory” proteins differ strongly between rovers and sitters. Two important classes of secretory proteins are neurotransmitters and peptide hormones. I show that genes associated with secretory vesicle release, axonal transport, and other subsystems of neural function show consistent differences between rovers and sitters. In the Discussion I relate these patterns to results from other researchers on behavioral plasticity and learning in rovers and sitters.

5

Chapter Four will therefore take us into details of foraging and PKG, and into what is known about the insulin signaling system and its interaction with nutrition. Thus the last two sections of the Introduction will look at foraging and insulin in depth.

Unifying Theoretical Perspective What theoretical perspective unifies the several areas of enquiry in this thesis? These principles are important in my approach: -parsimony (dimension reduction) -prior knowledge of biological pathways -conservation of mass and energy -Bayesian inference -Pattern testing of fitted models

Parsimony. Parsimonious explanations of complex phenomena attempt to find a small number of factors which explain a large number of observed effects. In concrete biological terms, finding one enzyme or one transcription factor which causally controls phenotypes under study is the primary goal of many investigations. However, if a single causal factor for a complex system can’t be determined, we can still look for simple observable patterns. One of the issues when dealing with datasets recording dozens to thousands of phenotypes simultaneously is finding simple patterns to explain the many observed results.

Many statistical techniques such as Principal Components Analysis and Principal Coordinates provide condensed views of the many variables (phenotypes) in the dataset chosen so that the view (the PC axes, for example) explains the maximum amount of variation possible with one or more linear factors (Nash, Als et al. 2008). These techniques are useful as starting points for analysis, but need to be linked back to concrete biological hypotheses that are explanatory and predictive, rather than the PC axes which are simply descriptive. A second kind of statistical technique is similar to PC analysis: Factor Analysis (FA) both tries to identify one or several linear “factors” explaining variation in data, and tries to test the adequacy of a linear model to explain the data given the underlying factors (Lawley, Maxwell et al. 1971; Bartholomew and Knott 1999).

In all of these cases, we are doing what a mathematician would call dimension reduction. If each phenotype or variable represents a dimension, something like an Affymetrix Drosophila Genome microarray dataset has about 14,000 dimensions. This large dimensionality can be daunting! However it’s possible that a smaller dimension “subspace” – that is, several linear factors – may explain much of the variation.

6

In each of the complex datasets I analyse, I look for the smallest number of dimensions which explains the data well. Having determined this, I then look for a testable biological hypothesis about which biological mechanism in the organism may produce the significant dimensions.

Prior knowledge. A drawback of applying PC analysis to something as complex as the entire genome or circadian CH phenotype of a fly is that there are many different causal factors at work in the organ systems and biochemical pathways of the organism. These causal factors may respond differently to the same environmental factors, so each PC axis may explain a small amount of variation over the whole dataset. However, given the knowledge accumulated in the scientific literature and encoded in systems such as Gene Ontology (GO) (Consortium 2000; Consortium 2001) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) (Kanehisa, Araki et al. 2008; Okuda, Yamada et al. 2008). Applying parsimony discovery tools such as PCA or FA to entities (such as genes) sharing a common role is more likely to yield meaningful, interpretable factors which explain larger amounts of variation. Where there are not predefined GO or KEGG groupings for things like metabolites or CH compounds, we can use common chemical characteristics or common roles in a known pathway to group entities for analysis. This approach combines the prior research done by the biological community with the data at hand to give a more biologically reasonable segmentation or partitioning of analysis. This guided or a priori analysis contrasts with popular unguided techniques such as hierarchical clustering, which can fail to provide biological insight. Combining parsimony and prior knowledge of pathways and gene groups leads to an approach which I call “Constrained Hierarchical Parsimony ANOVA” (CHPA). This acronym captures four aspects together: - Using ANOVA at each level of analysis to fit general linear models - Replacing less parsimonious ANOVA models (for example, one model per gene or metabolite) with more parsimonious ones (for example, one model for a gene group or pathway) - Building hierarchies of parsimonious ANOVA models, so that effects best explained at a higher level are captured by the parameters of that model while those capturing a local subsystem’s deviation from the general pattern are captured at the lower level. - Constraints on the set of all possible hierarchies are derived from a priori hierarchies such as those given by Gene Ontology. This enforces a degree of biological meaningfulness on CHPA models that general hierarchical clustering does not provide.

Conservation. One day our models of cell biochemical function will be so detailed and reliable that systems biologists will routinely apply mass and energy conservation principles to constrain and analyse whole organisms. At this point in time the most complex organism to yield in part to this approach is E. coli (Almaas, Kovacs et al. 2004). However conservation principles can still be applied approximately to yield insight. I provide several applications of this, which will be presented in the Discussion. In Chapter 1 mass and energy conservation arguments are applied. A pathway analysis and mass conservation argument is used in Chapter 3, and an energy conservation method is inherent in Chapter 4.

7

Bayesian inference. Although complex statistical techniques of Bayesian inference now exist (Baldi and Hatfield 2002; Efron and Tibshirani 2002; Tadesse and Ibrahim 2004; Edwards, Page et al. 2005; Tadesse, Ibrahim et al. 2005), my use of Bayesian principles has been simpler. Just as prior knowledge of biological pathways helps make data analysis more powerful, a Bayesian prior assumption that many disparate genes or metabolites share common distributional properties makes statistical analysis of large datasets immensely more powerful (avoiding in part the dreaded “Bonferroni insignificance” trap (Moran 2003; Nakagawa 2004)). Since it is often quite expensive or impossible to get large enough numbers of replicates of gene array or GC datasets to meet the minimum replicates assumptions of most statistical methods (e.g. N greater than 15-20), a Bayesian “sharing” of information between multiple independent variables within a dataset effectively multiplies degrees of freedom and statistical power. Here I have applied this to generalize Bayesian methods of gene-array analysis such as Cyber-T (Baldi and Hatfield 2002) to a simple, extremely powerful form. I have used it in conjunction with pathway or gene group ANOVA analysis using a simple CHPA model to provide very powerful methods of detecting changes which are very small by traditional methods. The primary applications of this are in Chapter 4 and are further developed in the Discussion.

Pattern testing. The final aspect of my theoretical approach which deserves mention is to always look for meaningful biological pattern in the fitted parameters of models. In this my debt to the school of Tukey is clear (Tukey 1977). When we fit a model with multiple parameters to a high-dimensional dataset such as a gene array or a set of CH compounds, the model parameters (the slopes and intercepts of the linear equations) become an alternative representation of the data. Just as it is always important to plot the data itself (to look for errors and outliers as well as to see patterns the eye is better than the computer at detecting), so it is always important to look for patterns in fitted model parameters. In the publications resulting from the work in this thesis, I have minimized the use of three dimensional (3-D) plots because of feedback from advisers and reviewers that many readers have difficulty following them. However, when dealing with complex models, I personally view such plots as essential. Often these reveal patterns which lead to new hypotheses about the biology. These must then be tested using an appropriate statistical method. The Discussion returns to this topic with specific examples from the four chapters.

Genotype by Environment Interactions This section discusses genotype by environment interactions (GEI) from several points of view, in an attempt to point to a number of generalizations of the concept which have recently become more prominent in the literature. After introducing “classical” GEI, I look at (a) dynamic models and GEI, (b) geographical mosaics and GEI, (c) social interactions and GEI, (d) pleiotropy and GEI, and finally an overarching concept called (e) genotype by context interactions.

8

Genotype by environment interactions (GEI) arise whenever the response of two or more genotypes to two or more environments is unequal (Lynch and Walsh 1998). GEI interactions are of interest for practical applications in medicine and plant and animal breeding, and theoretical reasons in genetics and evolutionary theory.

The advent of “personalized medicine” has greatly increased interest in GEI, because an individual’s genotype may predict variant risk for disease or unusual reaction to treatments (Alm, Sanjeevi et al. 2002; Martin, Kissebah et al. 2003; Eley, Sugden et al. 2004; Hunter 2005; Levin 2006; Ghebranious, McCarty et al. 2007). GEI in human genetic diseases are perhaps best known from cases such as sickle cell disease (SCD) and β-thalassemia, where the gene is β A globin (HBB) and the environmental factor is the presence or absence of malaria parasites. Experimental therapies for SCD are based on the genetic and biochemical characterization of the disease and range from drugs which increase transcription of fetal hemoglobin genes (Perrine, Ginder et al. 1993; Dover, Brusilow et al. 1994; Perrine, Olivieri et al. 1994) to gene therapy (Pawliuk, Westerman et al. 2001; Imren, Payen et al. 2002; Imren, Fabry et al. 2004). Further, models of the “adaptive landscape” (Wright 1932) induced by the normal HBB A allele and mutant S and C alleles in the presence of malaria helped explain why the homozygous near-lethal S allele is maintained in African populations when the C allele, which is homozygous viable with far higher fitness than AS heterozygotes in malarial regions, is rare ( (Cavalli-Sforza and Bodmer 1971), quoted in (Freeman and Herron 1998)). Different mutations in the same gene cause β-thalassemia. Thalassemic patients have different clearance rates for analgesic drugs such as paracetamol (Acetaminophen) (Tankanitlert, Howard et al. 2006). These drugs are metabolized by UDP-glucuronosyltransferases (UGT), and naturally occurring polymorphisms in genes UGT1A1 and UGT1A6 vary rates of drug clearance in thalassemics (Tankanitlert, Morales et al. 2007). Thus there is not only a GEI between HBB alleles and malaria resistance, but there are additional GEI for drug metabolism which themselves may be subject to epistatic interactions with polymorphisms in UGT genes. This suite of naturally occurring alleles of the HBB locus in SCD and βthalassemia therefore represents two ways in which humans have evolved malaria resistance, each causing its own GEI, and each requiring genotype-specific treatment modalities.

Nutrition, an environmental factor with large effects on health and disease, is known to interact with genotype in metabolic syndrome, obesity, and diabetes (Bell, Walley et al. 2005; Schwartz and Porte 2005; O'Rahilly and Farooqi 2006). Changes in phenotypes such as body fat or blood sugar in response to changes in food are examples of phenotypic plasticity; when genotype affects plasticity, GEI result. Some individual genes involved in GEI with nutrition are known for humans, such as the interaction between naturally occurring alleles of PPARγ and the amount or types of dietary fats (Luan, Browne et al. 2001; Memisoglu, Hu et al. 2003), or the interaction of beta(2)-adrenoceptor alleles and carbohydrate intake (Martinez, Corbalan et al. 2003); the phenotype measured in both cases above is total body mass.

9

Genotype by nutritional environment interactions are also important in the study of evolution. Desert Drosophila in the western hemisphere live in cactus rots, and different species and subspecies have specialized on different species of cactus whose nutritional qualities differ. D. mohavensis populations in Baja California and mainland Mexico differ in preferred cactus species; by raising F2 males from crosses of the Baja and mainland populations in the laboratory on two cactus species, Etges et al. were able to identify 20 quantitative trait loci (QTL) affecting courtship song or mating success behaviours (Etges, de Oliveira et al. 2007). They found significant interactions between locus and cactus (food) in at least one behaviour for 10 of the 20 QTL. This study illustrates the importance of GEI in evolutionary theory, with a primary environmental variable (the nutritional qualities of different host cactus rots) driving many adaptations. Etges et al. point out that GEI in traits important in sexual selection can maintain genetic variation in these traits while slowing the reduction in genetic variation due to sexual selection in this Drosophila which may be undergoing the early stages of incipient speciation.

Etges et al consider traits important to prezygotic isolation; GEI has also been shown to be a potential contributor to postzygotic isolation in the Dobzhansky-Muller model of speciation (Bordenstein and Drapeau 2001). Considerable experimental and theoretical attention has been paid to signalling traits and sexual selection, with GEI considered one important way in which genetic variation is maintained in spite of the reduction which sexual selection should cause (Brandt and Greenfield 2004; Greenfield and Rodriguez 2004; Danielson-Francois, Kelly et al. 2006; Kokko and Heubel 2007).

An experimental study used 26 E. coli strains with single mutations all derived from a common background, exposed to a factorial design with 2 levels each of nutrition and temperature (Remold and Lenski 2001). GEI for temperature was not significant, but 11 of 26 strains showed significant GEI for nutrition. The authors suggest that this demonstrates GEI may be quite common in de novo mutations and hence may be important in selection on new mutant alleles.

GEI can maintain genetic variance in populations subject to environmental heterogeneity, but only under certain conditions (Levene 1953). Levene’s insight has been explored in a variety of theoretical models (Gillespie and Turelli 1989; Turelli and Barton 2004). More recent experimental studies have emphasized that GEI in varying environments will not always lead to maintenance of variation (Prout and Savolainen 1996). Nonetheless, adaptation of local populations to different environments has frequently been shown to lead to GEI, with many reciprocal rearing studies in plants and animals showing that local races or varieties have higher fitness in the location where they are found than those from other sites, and lower fitness when raised in other locales (Ferrari, Via et al. 2008; Nash, Als et al. 2008).

10

An example of behavioural GEI in local populations occurs in a study of Daphnia magna preferred feeding depths in two environments: water which had contained fish, and fish-free water (De Meester 1996). Some clonal lines (genotypes) showed no response to presence of fish while others shifted to feeding in much deeper waters, showing that there was GEI. Clones originally collected from lakes with fish tended to show a larger response to fish-scented water than those collected originally from fish-free lakes. Here we see that a complex behavioural phenotype can exhibit strong GEI (Sokolowski and Wahlsten 2001). Awareness of this tendency for behaviours to respond strongly to varying environments is essential for good design and interpretation of behaviour genetics experiments (Crabbe and Wahlsten 2003; Wahlsten, Metten et al. 2003).

GEI feedback In the examples cited thus far, the environments with which genotypes interact can be viewed as constant external influences, although the reality is that food species, parasites, and predators can themselves evolve or change behaviour. When such change in the environment is slow compared to the phenotype change induced by the environment, it can be ignored in GEI analysis, but what happens when the environmental factor is itself responding to an individual focal organism’s phenotype with a similar rate of change?

In such cases the environment itself is a function of the interaction. Then the dynamically changing phenotype of the focal organism and the state of the environment may be modeled as a two variable differential equation, and the observed phenotype of the focal individual at the end of an experimental period must be measured along with its covariate, the environmental state. If the differential equation relating phenotype and environment is linear then there may be an equilibrium solution reached after a characteristic time (mathematically this is called the “relaxation time” (Glass and Mackey 1988; Lewis and Othmer 1997)}). If the experimental period is longer than this time, each phenotypic and environmental value pair will be approximately constant and standard GEI analyses may still make sense. If the experimental period is shorter than the relaxation time, or if the differential equation is unstable or exhibits limit cycles, the observed phenotype and environmental value will vary depending on initial conditions and the dynamics of the system. If the equation of change exhibits a sensitive dependence on initial conditions (e.g. a Lyapunov exponent greater than zero), the dynamic behaviour of the system may be chaotic (Glass and Mackey 1988). In these cases replicate experiments may give very different results.

Let us call the first kind of experiment, where phenotype and environment stably converge to approximately constant equilibria “dynamically stable GEI”, and the second kind “dynamically unstable GEI”. Below I give examples of actual GEI studies which may fall into the first or second category. These

11

comments apply to cases where plasticity of phenotype within a generation causes both focal organism phenotype and the environmental state to change rapidly. When viewed over the intergenerational time scale at which evolution acts, these interactions can lead to what Thompson has called GxGxE, resulting in a “geographical mosaic” of interacting co-adapted genotypes (Thompson 2005). The role of GxGxE interactions in maintaining or reducing genetic variation is reviewed in (Wade 2007) and is an exciting new application of classical genetic methods to community ecology and co-evolution.

Dynamically stable GEI: There is a large body of literature on genome expression and fitnessrelated traits in organisms such as yeast or E. coli grown in chemostat culture (Gasch, Spellman et al. 2000; Causton, Ren et al. 2001). A chemostat provides an environment in which a continuous inflow of nutrients, constant mixing of the medium, and dilution-mediated loss of cells in the outflow eventually results in a constant equilibrium cell number and cell mass and constant nutrient concentration in the chamber. Absent mutation, a single genotype of cells started in chemostat culture will go through a period of growth until the loss of cells in outflow is balanced by gain through division, and the nutrient levels are determined by a balance between inflow and utilization by growth and respiration. The relaxation time is the number of generations for the population of cells to grow from initial density to near equilibrium density.

When the continuous flow regime of the chemostat is replaced by periodic replenishment of resources in batch culture, there is a fluctuating environment, which may nonetheless be stable when averaged over the period of several replenishments. An example is given for Daphnia pulex consuming the alga Chlamydomonas reinhardtii, where the overall population dynamics was stable over longer periods even though algal concentrations fluctuated considerably from beginning to end of each batch (Nisbet, E. et al. 1997). A study using D. pulex explicitly compared batch versus chemostat culture and the effect of varying food quantities and qualities, using genetically distinct clones (Weider, Makino et al. 2005) feeding together in the same cultures. They found a clear GEI with nutrient quality in the batch culture experiment, with different clones driving the other to extinction depending on the nutrient inputs. This experiment had fluctuating environments in the short term (food varied based on replenishment schedule, and the level of the competing clone varied over time) but converged reproducibly to the same genotype in each replicate run. In their second experiment, the two genotypes competed within a continuous-flow chemostat where different light levels produced algal food quality changes. Again the results were dynamically stable with GEI, but the population dynamics differed; one clone drove the other to extinction in the high light/food environment, but the clones coexisted stably at the low light/food level. Thus in this last case the equilibrium environment for one genotype included not only equilibrium food levels but an equilibrium concentration of the other genotype.

Unstable GEI: As a counter-example to the general rule that chemostat cultures approach a steady state, researchers in yeast have documented robust 40 minute metabolic cycles (detectable as fluctuations in

12

dissolved oxygen in the medium) in cultures maintained under slow, but not zero, growth conditions. Recent results have shown that during these cycles transcription of gene groups is coordinated, 3 separate clusters of genes being transcribed at 3 different points in the 40-minute cycle (Klevecz, Bolen et al. 2004; Tu, Kudlicki et al. 2005; Lloyd and Murray 2006). As the doubling time under these conditions is ~ 8 hours, this is not a simple manifestation of transcription dependent on cell cycle phase. Thus a stable “limit cycle” in gene expression is established; it has been shown that key metabolites also cycle in characteristic ways (Tu, Mohler et al. 2007). Strains with alterations in the cycle period will show differing phenotypes at a given point in time if gene expression or metabolite level is measured. Interestingly, mutants where the cycle time varies have been isolated and in some cycle time is dependent on nutrient conditions (Lloyd and Murray 2006). Therefore if these mutants are compared to wild type strains under two nutrient conditions, profound GEI in gene expression or metabolites will be seen at some points in time when the cycles are out of phase, but not at others when they are in phase. These authors also present evidence that chaotic components may be present in these cycling systems (Murray and Lloyd 2006).

I have introduced the ideas of dynamically stable and unstable GEI and of GxGxE to illustrate how the presence of “genes in the environment” (Wolf 2003; Wolf, Brodie III et al. 2004) can change interpretation and analysis of even classical single species GEI experiments. As soon as we shift our viewpoint from genotype response to environment, to looking at environment response to genotypes, a wealth of ideas new and old can be seen to be related to GEI theory. A partial list of these includes geographical mosaics, social interactions, epistasis, and pleiotropy. Definitions of these, examples of studies, and discussion of their relevance to this thesis are given below.

Geographical mosaics and GxE “Although it would seem probable that the adaptive value of a genotype is a function of the presence and relative frequencies of other genotypes coexisting with it, the study of this dependence has been neglected.” – Lewontin 1955 “Geographical mosaics” is a phrase introduced by Thompson (Thompson 2005) to capture the fact that prey, predator, or parasite species may be genetically structured over space or time. Then the biotic environment a focal species encounters in different locations varies due to genetic causes as well as abiotic causes such as microclimates - see review by Wade (Wade 2007).

The relevance to fruit flies is direct: each life cycle stage encounters patches whose biotic constitution is a key determinant of viability and fecundity. For instance eggs are laid in bunches due to gregarious female choice of oviposition sites, influenced by a host of factors including type of food and the choices of other females as indicated by aggregation pheromones (Wertheim 2001; Rohlfs and Hoffmeister 2003; Rohlfs and Hoffmeister 2004). Here the primary biotic variable is the species and decomposition

13

state of the fruit chosen, which is itself partly determined by the microflora inhabiting the site of decomposition. The presence of parasitoids is an additional factor (Rohlfs and Hoffmeister 2004). Most Drosophila scientists have had the experience of media with too few eggs laid per volume of food producing fewer emerging adults per egg laid than those with a higher initial concentration of eggs. For example, in Lewontin’s classic 1955 IGE study quoted above, 11 of 22 strains tested showed an optimal intermediate density for larval survival (Lewontin 1955; Lewontin and Matsuo 1963). Drosophila inoculate food with yeasts and bacteria, influencing the food environment for larvae (Begon and Shorrocks 1978; Gilbert 1980). Increasing larval densities may increase survivorship due to reduction of competing fungi by larval feeding (Hodge, Mitchell et al. 1999). Larvae are mobile, and can move between patches of food; the extent of this is influenced by larval condition and genotype (Graf and Sokolowski 1989; Kaun, Riedl et al. 2007). However genotypes of fruit substrates or microflora contribute to patch discrimination and larval viability, for instance by affecting sugar content of the fruit or amounts of toxic byproducts produced by microflora making this an example of GxGxE.

Less is understood about biotic effects on pupal site choice and survivorship, where abiotic factors such as humidity are reflected in genotypic differences between fly strains in pupal site height (Riedl, Riedl et al. 2007). Adults oviposition site choice in turn is influenced by environmental signals, signals from conspecifics, and genetic differences between females (Wertheim 2005; Wertheim, van Baalen et al. 2005). Differences among natural fly breeding sites in parasitoid and bacterial pathogen species have been documented, and genetic variance for concomitant differences in adult fly susceptibility to these pathogens and parasitoids shown (Graf and Sokolowski 1989; Hughes and Sokolowski 1996; Green, Kraaijeveld et al. 2000; Lazzaro, Sceurman et al. 2004; Corby-Harris and Promislow 2008). Indeed immune genes are among the fastest evolving in 12 sister species of Drosophila (Sackton, Lazzaro et al. 2007), suggesting that coevolutionary arms races between pathogens and their hosts are one driving force in evolution (Wade 2007). In these examples, the genes in the environment are those of host fruits, microflora (both as food sources, competitors, and pathogens), parasitoids, and other fruit eating species. In addition the environment contains genes of conspecifics which influence larval success through competition (Wolf 2003) and frequency-dependent selection (Fitzpatrick, Feder et al. 2007) and influence adult success via competition for mates.

I cite two further examples of GxGxE studies, this time in the emerging area of community genomics. In a recent review Whitham et al. (Whitham, DiFazio et al. 2008) discuss how the sequencing of model organisms which are themselves hosts to large communities can contribute to the study of GxGxE and geographical mosaics. His primary example is the poplar tree Populus angustifolia which can reproduce clonally by root runners, producing stands of genetically identical trees. The sequencing of the poplar combined with experimental plots in which replicate plantings of trees are surveyed for community composition and other characters, has allowed quantification of the degree to which community species

14

composition and ecosystem traits are affected by poplar genotype. This can be captured by the community heritability measure H2C, (Whitham, Bailey et al. 2006) which has values a high as 0.60 for canopy arthropod community, 0.70 for soil microbial community, and 0.80 for community trophic interactions (Whitham, DiFazio et al. 2008). A project to sequence the genomes of major soil microbes and fungi in poplar communities has already yielded the genome of a symbiotic mycorrhizal fungus (Moya, Pereto et al. 2008); as this project proceeds the tools for directly identifying GxGxE interactions through sequencing of microbial partners, pathogens, and community members will become available.

A laboratory study using a 3-level trophic system (food plants, aphids, and ladybird beetles) investigated whether genetic differences at one trophic level could interact via an intermediate level with genetic differences at a third level, a phenomenon they called “genotype-by-indirect environmental effects” (Astles, Moore et al. 2005). They found significant genetic variation in the response of ladybird phenotypes to the indirect effect of host plant on their aphid prey. This laboratory demonstration of GxGxE interactions using split-family experimental designs shows how tools of quantitative population genetics and experimental ecology can be combined.

Wolf et al. (Wolf, Brodie III et al. 2004) show how the Price equation (Price 1970) can be applied in cases where there is a geographic mosaic in which the genotype found in a location is correlated with the environment in that location. In this case selection acts based not only on the direction of the GEI, but also contains a term related to the product of the GxE direction and the correlation between genotype and environment. Thus the correlation between genotype and environment affects the direction of selection, and interacts with the direction of GEI effects.

Social interactions and GEI Behaviours are a rich source of GEI (Sokolowski and Wahlsten 2001), and behaviours which vary (are plastic) in response to the phenotypes of conspecifics are common. We may call such behaviours “social behaviours” sensu lato. If several genotypes’ social behaviours respond differently to the social group in which they exist, the potential exists for selection on social behaviours (Hamilton 1964). When the phenotype we measure is a component of a social behaviour, and there is heritable genetic variation for this phenotype, and one individual’s phenotype affects another’s, we have the preconditions for indirect genetic effects (IGE) on social behaviours, which will be introduced at greater length in a later section. Here, I wish to note several points relevant to GEI.

First, since behaviours often vary quite quickly in response to stimuli, in some cases of IGE for social behaviours we may be looking at what I have called dynamically unstable GEI. Thus we may not be

15

measuring equilibrium states of the dynamic social system in our experiments, and we may need to modify experimental designs and analytic tools to correctly determine the existence, magnitude and direction of IGE interactions. In addition, if abiotic environmental factors (such as light cycles) interact with the behaviours under study (e.g. if there is circadian regulation of behaviour) the dynamic social system may be driven by externally varying factors as well as its internal social dynamic. In such cases we must be alert for multiple levels of interactions such as G (genotype) x S (social), G x L (light cycle), S x L, and G x S x L (Chapter 2, (Kent, Azanchi et al. 2008)).

Second, if we detect significant IGE interactions (symbolized by path coefficients ψa,b between genotypes a and b), we may ask two GEI-related questions about the pattern of social interactions. For both questions the phenotype we look at is an IGE interaction strength ψ: The first question defines the environment to be the genotype of the fly a focal individual is interacting with, when flies are in a mixed environment. We may then ask if ψma,a - ψma,b ≠ ψmb,a - ψmb,b, where ψma,a is the interaction of a with a in a mixed group. If so there is a GEI in IGE coefficients within a mixed group. The second question defines the environment to be whether the social group is mixed or unmixed, and the phenotype to be the interaction strength (IGE ψ) for encounters with other flies of one’s own genotype. We may then ask if ψua,a - ψma,a ≠ ψub,b - ψmb,b, where ψu is the interaction in unmixed groups. This tests whether the change in interaction strengths between one social context (one environment) and another is genotype dependent. If so, then there is GEI in IGE coefficients between mixed and unmixed groups.

Although these two questions may be unfamiliar at first, they examine the genetic basis of indirect genetic effects in an important new way.

The first question asks whether the average strength (over genotypes) of IGE interactions with one’s own genotype (“self interactions”) differs from the average strength of interactions between genotypes, within one particular social group. To see this rearrange the test condition in (a) to be ψma,a + ψmb,b ≠ ψma,b + ψmb,a . To rephrase, on average are “kin” (same genotype) treated the same as “non-kin” (different genotype)? In (a) we have simply written down the numerator of the test statistic for a GEI in ψ, when E is the second genotype.

BUT, this is NOT what an experienced behavioural ecologist might think the most relevant question! If there were a simple neural circuit governing perception of “kin” versus “non-kin”, our behaviourist’s null hypothesis might be that all genotypes differentiate in the same way between “kin” and “non-kin”. In other words, does the difference between interacting with “kin” and “non-kin” depend on the focal individual’s genotype? But this is not equation (a); it is instead (c) ψma,a - ψma,b ≠ ψmb,b - ψmb,a . Here

16

we see that the population geneticist’s GEI equation (a) and the behaviourist’s equation (c) both make sense in their own domains but test very different hypotheses. What has shifted is the definition of “environment”. For the population geneticist in (a), environment is the genotype of the second fly in an interaction. For the behaviourist in (c), environment is whether the second fly is kin or non-kin.

The second question (b) asks whether genotypes have similar or differing shifts in response to “kin” as the environment changes from an unmixed to a mixed group. In this case, the population geneticist and the behavioural ecologist agree on the question to ask, so we don’t need an equation (d)!

We will examine questions a-c in the context of the host-visitor social group design (Levine, Funes et al. 2002) in the Discussion. Briefly, this design involves comparing two groups of flies. The first group has a single “wild-type’ genotype. The second group has a few of the wild type flies (or “hosts”) replaced by flies of a differing genotype (“visitors”). One asks whether the presence of the visitors modifies behaviours of the hosts. In (Levine, Funes et al. 2002) the visitors somehow caused a change in the circadian clock phase of the hosts.

Third, social interactions may depend on social context – that is, the type of social group in which interactions happen.

(a)

In the host-visitor paradigm (Levine, Funes et al. 2002) the nature and strength of

interactions depend not only on the mix of genotypes present but also on their frequency (Levine, J.D. – pers. comm.). When this happens there is frequency dependence of the interactions, and of any GEI found.

(b)

Social context may depend on other factors such as nutrition. A frequency-dependent

GxE social interaction also occurs in the social amoebae Dictyostelium discoideum (Shaulsky and Kessin 2007). Wild-type and “cheater” genotypes exist. When food is abundant they have similar fitnesses. When food is exhausted starving cells emit cyclic AMP (cAMP) aggregation signals, form into an aggregate which locomotes to a new location, where some cells differentiate into a stalk which holds up the fruiting body containing the spores formed by remaining cells. Stalk cells do not reproduce and die, spores cells can disperse and start new colonies. One cheater genotype does not produce stalk cells and “tricks” wild-type cells into producing disproportionate numbers of stalk cells (Ennis, Dao et al. 2003). Mixed genotype aggregates do exist in nature, and in the lab cheaters rise in frequency over generations of selection. However, Gilbert et al. (Gilbert, Foster et al. 2007) calculate that when cheaters represent more than 25% of the cells in aggregates they are no longer favored by selection due to deficient stalk formation. Thus, there is a GEI with fitness as the phenotype, nutritional state as the environment, and a frequencydependent GEI effect size.

17

(c)

If the presence of genetic diversity within groups of organisms increases the mean fitness

of the group, diversity can be maintained if the population is split into demes with differing gene frequencies, an effect called “social heterosis” by Nonacs and Kapheim (Nonacs and Kapheim 2007). Negative frequency dependent selection such as demonstrated for rover and sitter D. melanogaster larvae is a special case of social heterosis (Fitzpatrick, Feder et al. 2007). Nonacs and Kapheim cite numerous cases which they maintain support the existence of social heterosis, but few of their citations actually involve any form of social interaction; the majority in fact cite better mixed-group fitness under pathogen, predator, or parasite attack. What is of interest from their analysis are studies such as that of Clark et al (Clark, Schweikert et al. 2007) where deep sequencing of 20 wild-collected strains of Arabidopsis thalia demonstrated that 4% of the genome was highly dissimilar to the reference genome. They report “patterns of polymorphism are highly nonrandom among gene families, with genes mediating interaction with the biotic environment having exceptional polymorphism levels”. Clearly this is of interest for understanding GEI in wild populations, but it may also lead to an understanding of how social heterosis may be a cause of context-dependent IGE. Mutic and Wolf directly addressed this by raising recombinant inbred (RI) A. thaliana lines with a single control line, that is, in a mixed-genotype group (Mutic and Wolf 2007). They recorded morphological and fitness traits and performed QTL and genetic correlation analysis. Of interest here is that they found that 13 of 15 QTLs with direct genetic effects (DGE) also showed indirect IGE effects, and that as opposed to a model of neighbour-neighbour competition (which would produce opposite signs in direct and indirect effects), these pleiotropic DGE/IGE QTLs in many cases show signs in the same direction (facilitative rather than competitive). Thus in this model plant species, the conditions for social heterosis may exist, in that there are facilitative interactions between neighbours of differing genotypes. Whether the QTLs found by Mutic and Wolf map to genes involved in interaction with the environment (as seen in the Clark et al. study) remains to be tested. If so an interesting model system would exists in which genetic variation for GEI was associated with facilitative “social” interactions.

To summarize, in this section we have looked at some of the ways in which social behaviour phenotypes can have GEI. As these are some of the most complex phenotypes organisms exhibit, they are prone to interaction with the environment (Sokolowski and Wahlsten 2001) in ways that include dynamic instability and higher-order interactions, context-dependent effects on social behaviours such as IGE’s, and frequency dependent effects. There is much room for both theory and experiment to clarify these complex patterns of GEI.

Epistasis and GEI Genes can interact with other genes in the social or biotic environment as described above. When genes interact with other genes within one individual, we call this epistasis. Epistasis is not uncommon (Gurganus, Nuzhdin et al. 1999; Anholt, Dilda et al. 2003; Lazzaro, Sceurman et al. 2004; Spencer and

18

Promislow 2005; van Swinderen and Greenspan 2005). A quick summary of some ways in which epistasis and GEI overlap includes the points below:

Epistasis is detected using ANOVA techniques formally identical to those used to find GEI (Falconer and Mackay 1996; Mackay 2001; Anholt, Dilda et al. 2003; Gibson and Dworkin 2004; Gibson, Riley-Berger et al. 2004).

Epistasis variance in developmental traits is more commonly found in extreme environments which stress developmental homeostasis (Blows and Sokolowski 1995), so epistasis and environment interact. Epistatic interactions may contribute to speciation through hybrid incompatibility; this seems to be accentuated in extreme environments (Demuth and Wade 2007). These variations on Waddington’s original insight (Waddington 1942) all suggest that many more epistatic interactions exist than are visible normally, due to canalization or network homeostasis. Environments at the edge of tolerance levels diminish canalization and allow epistatic relationships to be seen.

The comments by Friedman and Perrimon on signaling network robustness (Friedman and Perrimon 2007) and epistasis reiterate the view that epistasis is common but often masked by network redundancy or “robustness”. However, van Swinderen and Greenspan present a contrasting view in their study of mutant alleles reacting epistatically with a Syntaxin1A (Syx1A) behavioural phenotype (van Swinderen and Greenspan 2005). Their article begins with the sentence “Context is the hallmark of biological processes” and proceeds to determine whether networks of epistatis interactions are sensitive to genetic context. They started with a temperature-sensitive mutant of Syx1A and located mutations in 16 other genes which modified the Syx1A mutant phenotype. They then tested pairs of the 16 interacting genes and found many epistatic interactions among these genes. However, which pairs of genes interacted depended dramatically on whether the original Syx1A mutant was present or absent. They warn that patterns of network interaction seem to be sensitive to the genetic background of the network.

Proulx et al studied robustness to both genetic and environmental perturbation of yeast gene expression and protein levels (Proulx, Nuzhdin et al. 2007). They found that the largest proportion of variation in robustness to genetic background was explained by robustness to environmental changes, but that other factors such as network position of a gene also are significant. Their “genetic robustness” is in a sense the inverse of epistasis: it measures how little a change at other loci affects one locus. Thus the patterns they find suggest that a portion of the tendency of a gene to have epistatic interactions is determined by functional factors such as its position in networks.

Epistatic interactions of a gene may be the cause of multiple associated phenotypes (pleiotropy). If some have positive and some negative fitness effects antagonistic pleiotropy results, which is hypothesized

19

to be an evolutionary source of late-life senescence (Spencer, Howell et al. 2003; Spencer and Promislow 2005; Wolf, Leamy et al. 2005; Wolf, Pomp et al. 2006; Scarcelli, Cheverud et al. 2007). One connection to GxE studies is in the area of caloric restriction (CR), where interactions between life-extending or shortening alleles of genes and nutrition are common (Clancy, Gems et al. 2001; Lin, Hsin et al. 2001; Clancy, Gems et al. 2002; Kapahi, Zid et al. 2004; Iser 2007).

Interestingly, although gene arrays performed on two genotypes provide one of the simplest methods of detecting epistasis en masse (in transcript levels), some authors have found that epistatic interactions are more common in metabolic traits in a pathway than in the transcript levels for the enzymes (Wentzell, Rowe et al. 2007), suggesting levels of epistasis found from transcript studies underestimate true levels.

One kind of IGE is a maternal effect. A study which quantified maternal and direct effect loci in mice showed that epistasis was 2.5 times more common in maternal effects than in direct effect (Wolf, Vaughn et al. 2002), so there is a tantalizing suggestion of relationship between IGEs and epistasis. Wolf has called epistatic interactions between genes in conspecifics “GxG epistasis” (Wolf 2000). If the two genes are at the same loci, this is a synonym for a non-linear IGE, but if the loci are different coevolution of the two loci can occur across individuals (Wolf, Brodie III et al. 2004). In the more classical case of two genes in the same individual, Wolf et al. show how the same concept of the correlation of genotype to environment across locations mentioned above under geographic mosaics can effect evolution of linkage disequilibrium within individuals (Wolf, Brodie III et al. 2004).

To sum up this section, epistasis can be viewed as GEI where the environment is other genes in the internal milieu. More traditional GEI where the environment is external to the organism may well be due to internal epistatic interactions (for instance, when the focal gene modifies the effects of a gene network whose job is to respond to environmental variables such as changes in nutrition). Internal epistatic interactions may be masked by canalization, robustness, or redundancy in the gene networks involved, and revealed when environmental stresses (such as starvation, heat shock, or oxidative stress) disrupt the normal homeostasis. Thus, traditional external GEI and epistatic internal GEI may be closely linked. An internal form of epistasis x environment interactions, where the environment is the presence/absence of a mutation at a third locus, demonstrated that for a fly behaviour phenotype networks of interactions may not be stable to the presence of mutants at additional loci.

Pleiotropy and GEI “Whereas some fear and loathe pleiotropy, we increasingly have to bite the bullet and accept it.” -J.C. Hall (Chan, Villella et al. 2002; Hall 2002)

20

Pleiotropic effects of genes are common, both at the molecular level where network analysis shows many if not a majority of genes or proteins are linked to at least one or more others in a functional relationship (Featherstone and Broadie 2002; Giot, Bader et al. 2003; Fitzpatrick 2004), and at the level of behaviours (Anholt and Mackay 2004).

Antagonistic pleiotropy Why is pleiotropy important? As I discuss below, several theories compete to explain evolution of life history traits such as ageing. One theory involves pleiotropy. Medical researchers have generalized from this to apply negative or “antagonistic” pleiotropic effects to the explanation of many human diseases, including mental illnesses. Enormous efforts have been and are being expended on trying to map susceptibility loci for mental illnesses in the human genome, often with disappointing results, even when heritability of the disease is high. A fascinating paper (Keller and Miller 2006) asks which evolutionary theory best fits the human data on mental illnesses, and determines that negative pleiotropy is not a good fit and that large projects attempting to find major effect loci may be therefore wasting their time. They delve deeply into some of the theoretical literature discussed below as well as providing a meta-analysis of many medical genetics studies. Results on gene by environment interactions in mental illness are important to the argument in several places. The journal issue in which this paper appears has over 20 responses and critiques by practitioners as diverse as psychiatrists and medical geneticists. This lively discussion is an example of real-world impacts in human health research of what might otherwise seem rather academic studies.

If pleiotropy is common, how does it affect evolution? A number of authors have addressed this issue, amongst whom Barton and Turelli have written influential theoretical papers (Turelli 1985; Barton 1990; Gavrilets and de Jong 1993; Turelli and Barton 2004). Pleiotropy can maintain genetic variation in traits under selective pressure via distinct mechanisms such as balancing selection or mutational load (Barton 1990; Turelli and Barton 2004). The evolutionary impact of pleiotropy is not a subject of this thesis (thank goodness!) except where connections between pleiotropy and GxE may be derived from evolutionary arguments. I have already mentioned the study of Mutic and Wolf (Mutic and Wolf 2007) which examined IGE in pairs of genetically diverse A. thaliana plants grown in the same pot. A high degree of pleiotropy was detected using multiple-trait composite interval mapping, and direction or sign of pleiotropic effects was the same more often than different. This was surprising because a dominant school of thought has followed G.C. Williams in focusing on cases where an allele has opposite effects on two traits, which is called “antagonistic pleiotropy” (Williams 1957). Williams highlighted particular cases where traits were related to fitness and a negative effect occurs later in life than a positive effect. In these cases selection for the positive impact could lead to senescence. The theory of antagonistic pleiotropy is one of several which competes to explain ageing from an evolutionary viewpoint (Hughes and Reynolds

21

2005). Although much debated, there appears to be evidence of antagonistic pleiotropy in fitness traits at different ages (Vieira, Pasyukova et al. 2000; Scarcelli, Cheverud et al. 2007), but there is also negative evidence (Curtsinger and Khazaeli 2002), and evidence favoring the mutation accumulation theory (Leips, Gilligan et al. 2006).

It is notable, and cautionary, that T.F.C. Mackay’s group found evidence of antagonistic pleiotropy at QTLs in D. melanogaster lifespan, and that these were associated with GEI and genotype by sex interactions (Vieira, Pasyukova et al. 2000), but when they examined fecundity at different ages they found little evidence of pleiotropy (Leips, Gilligan et al. 2006). It is unclear whether this is due to methodological differences or whether these two classic fitness traits, survivorship and fecundity, are differently controlled at evolutionary time scales.

Using genetic manipulation in a third study, they examined the effect of four P-element insertions in the region between two genes, one coding for a gustatory receptor for trehalose sensing (Rollmann, Magwire et al. 2006). They examined nutrient intake, lifespan, starvation resistance, and heat stress. The effects on lifespan and stress resistance were often positively correlated (positive pleiotropy). There were GEI in food choice (a behavioural phenotype) depending on trehalose levels. When two insertions in similar locations and orientations but in different backgrounds were compared, they had opposite effects on 3 of 4 traits (gene by background interaction). As the insulin signaling pathway is known to mediate effects on lifespan, they tested the insertions for interaction with mutations of genes at 3 points in the insulin pathway, using both starvation resistance and lifespan as phenotypes. Strikingly, although significant epistasis was found with each of the 3 insulin pathway genes, it was found in different genes, phenotypes, and directions for each of the insertions tested. Some insertions interacted with some insulin genes only in one phenotype, other insertions showed interactions with other genes, in opposite directions or in the other phenotype.

Thus in this controlled genetic manipulation, GEI and pleiotropic effects were found, there was an epistatic interaction with a major pathway governing lifespan and growth, but the effects were extraordinarily sensitive to the nature of the mutation and the genetic background. As this research group is sophisticated in quantitative genetic techniques, their differing results in 3 studies examining similar phenotypes in D. melanogaster suggest that no single theory of pleiotropy and ageing will explain all biological variation in senescence. They also emphasize that polymorphisms in genes affecting sensory traits such as trehalose gustatory response and food choice may interact with downstream metabolic pathways such as insulin, so sensory, behavioural, metabolic, and life history traits can be affected by single loci. Interactions between sensory neurons and the insulin pathway affecting lifespan have been found in C. elegans as well (Alcedo and Kenyon 2004), and Alcedo et al. have suggested that the connection between sensing food and insulin secretion may be a taxonomically widespread feature. This

22

provides a connecting link between a large class of gene by food or sensory environment interactions and metabolic and life history phenotypes, such that association of GEI and pleiotropy may be common in this cluster of food-related traits.

The most-studied theoretical view of pleiotropy relates to antagonistic pleiotropy, based on William’s insights over 50 years ago. One mechanism leading to negative pleiotropy is allocation of a fixed pool of resources between two different uses, for example reproduction versus stress resistance (“resource allocation”). However, genes work in networks and pathways; if variation in an upstream gene governing getting resources (“resource acquisition”) causes the size of the pool of resource to be changed, stress resistance and fecundity might become positively correlated (Houle 1991; Worley, Houle et al. 2003). Thus depending on where in a pathway a genetic polymorphism is found, it may induce positive or negative pleiotropy. Gardner and Latta performed a careful meta-analysis of 27 QTL studies and showed that both patterns may be found (Gardner and Latta 2007). In a number of cases, both positive and negative pleiotropy occurred for the same trait at different loci, consistent with the resource acquisition versus allocation model. In addition, they found that genetically correlated traits shared common QTLs much more often than expected by chance; this is important for mapping loci contributing to pleiotropy by methods such as multiple-trait mapping (Chesler, Lu et al. 2005; Mutic and Wolf 2007).

Resource allocation versus acquisition may thus determine whether mutations show positive or negative pleiotropic effects. In this thesis, I trace effects of genes at different points in synthetic pathways leading to production of cuticular hydrocarbons (CH). Effects relating to the total pool of metabolites available to be partitioned (resource acquisition) are extremely important, and Chapter 1 deals extensively with this. Once the pool is determined, it still needs to be partitioned into the 24 different CH compounds we detect (resource allocation). Chapter 2 shows how to use dimension reduction to map the observed patterns to underlying branch points in the synthetic pathways known to be due to enzymes called desaturases and elongases. The mapping allows us to detect large GEI effects in each putative enzyme class. Chapter 2 also shows that the social environment is one of the most important sources of GEI. Chapter 3 shows variation in one allocation “decision” map to social and other environmental effects on expression of the gene for a key desaturase. In chapter 4, variation at the foraging locus, previously shown to affect resource acquisition (Kaun, Riedl et al. 2007), is shown to produce large-scale positive and negative pleiotropy in gene expression, and allocation “decisions” which shift energy stores dramatically between carbohydrates and lipids. This same genetic polymorphism is known to regulate behavioural pleiotropy in food acquisition, absorption, and other phenotypes. This shows that foraging flies may be a model system for a single locus with both resource acquisition and allocation effects and with strong GEI that are widely pleiotropic. Given all this, it seems worthwhile to dig a bit deeper into connection between GxE and pleiotropy.

23

Pleiotropy and networks The enterprise of mechanistically connecting traditional phenotypes such as behaviour or lifespan to genes has advanced enormously with the increase of molecular and genetic information about the networks genes and proteins form and how they work. Saccharomyces cerevisiae yeast has been a model organism for this in the past decade. Much has been learned about the structures of yeast gene and protein networks, including how pleiotropy of genes relates to their location in networks, their essentiality, and other traits (Featherstone and Broadie 2002; Hahn and Kern 2005).

Promislow (Promislow 2004) notes that proteins associated with senescence in yeast have more connections in protein networks than average, and that a protein’s connectivity is correlated with its degree of pleiotropy in annotated traits. He suggests this is consistent with models of evolution of antagonistic pleiotropy. I noted in the section on epistasis and GEI that genes connected with ageing often show gene x nutrition or gene x stress GEI. This raises the question: are genes with higher protein connectivity associated with more GEI? The connectivity of the gene regulatory network (as determined by the number of transcription factors affecting a gene), is indeed positively associated with more variation in transcript levels (Promislow 2005). This suggests but does not prove that GEI variation may be higher for such genes, a topic for future research.

A group of researchers associated with the University of Tennessee have assembled a series of databases of genomic and behavioural data on recombinant inbred (RI) mice whose parent strains are fully sequenced (WebQTL) (Chesler, Lu et al. 2004; Chesler and Williams 2004; Li, Chen et al. 2006; Rosen, Chesler et al. 2007). One of the early studies leading to the development of WebQTL (Chesler, Wilson et al. 2002) was motivated specifically by the problem of lab to lab differences in behavioural phenotypes highlighted by Wahlsten as GxE variation (Crabbe and Wahlsten 2003; Wahlsten, Metten et al. 2003). In this early computational study, Chesler et al. identified numerous environmental variables (E) affecting (inadvertently!) a nociception phenotype. They showed that variation could be partitioned into G (27%), E (42%) and GEI (18%), so GEI accounted for 2/3 as much variation as genotype (Chesler, Wilson et al. 2002). The association of such GEI with network structure and pleiotropy has been based on QTL studies (below).

WebQTL tools allow independent discovery of quantitative trait loci (QTLs) for behavioural traits (Chesler, Lu et al. 2005) and molecular aspects of neurogenesis (Kempermann, Chesler et al. 2006) and the cross-comparison of these to QTLs for gene expression (Chesler and Williams 2004; Baldwin, Chesler et al. 2005; Li, Lu et al. 2005). The combination of the large panel of RI lines, the many phenotypes surveyed across these lines, and the independent genomic data allows much higher power in QTL studies. A relationship which might be ambiguous in a QTL study for a single trait becomes much clearer when

24

attention is restricted to overlapping QTLs for multiple phenotypes, which in turn have direct genomic evidence of association with the phenotypes under study. Using these techniques, genes such as the D2 dopamine receptor (Drd2) were identified whose expression correlates very significantly with 5 separate behavioural phenotypes, including two involving ethanol preference or response (Chesler, Lu et al. 2005). The Drd2 transcript is not polymorphic, but the gene contains multiple SNPs in its promoters. A search using behaviour (ethanol-induced open-field activity) and Drd2 transcript levels identified a QTL on chromosome 9 which is associated both with behaviour and with dopamine receptor gene expression. It therefore seems that one or more chromosome 9 loci are involved in a pleiotropic system which encompasses several behaviours and may have dopamine receptor expression as one causative factor. A second result from this study found a “clique” of highly-coregulated genes some of whose members had very high connectivity to multiple other genes (a transcriptional “hub” gene). Interestingly, this clique had significant correlation between clique gene expression and several locomotor behaviours, and included genes for proteins known to form a complex important in synaptic function.

Some enzymes are “hubs” in the sense that they have many effects in the cell by direct interaction with many other proteins. Enymes such as Protein Phosphatase 1 (PP1*) or cyclic AMP dependent protein kinase (PKA) can interact with many proteins. High interaction specificity comes from being part of clusters of catalytic, regulatory, and binding proteins. Given multiple isoforms of each protein, the number specific protein interactions can be much larger than the number of genes (Janssens and Goris 2001; Zhang, Ma et al. 2001; Bauman and Scott 2002; Bennett, Lyulcheva et al. 2006). Many of these enzymes are involved in the regulation of processes such as metabolism. Since metabolic processes are often responsive to environmental factors such as nutrition or stress, variation in genes for these enzymes can produce GEI. This would be a specific instance of the general trend noted by Promislow between connectivity and GEI. I mention signaling enzymes in particular since much of Chapter 4 will deal with GEI due to the kinase PKG which is encoded by the foraging gene.

Pleiotropy, signaling cascades, and stress When a gene encodes a crucial enzyme or transcription factor near the beginning of a cascade of responses to changing environments, mutations in that gene will often affect many downstream phenotypes, leading to large-scale pleiotropy. An example is the well-studied transcriptional response of yeast Saccharomyces cerevisiae. A number of studies have detected “modules” or clusters of yeast genes whose transcription is up- or down-regulated in response to several kinds of environmental stress, such as low nutrition, hypoxia, or osmotic stress (Gasch, Spellman et al. 2000; Causton, Ren et al. 2001; Gasch and Eisen 2002; Herrgård, Covert et al. 2003). Several modules occurring in the largest number of stress conditions seem to be regulated either by the YAP1 or Msn2/Msn4 genes, and mutants in these genes

25

correspondingly block many kinds of stress response (Gasch, Spellman et al. 2000; Causton, Ren et al. 2001). If each type of stress response is counted as one trait, variants of these core regulatory genes will reveal a large number of GEI, or pleiotropy in GEI at these loci.

Downstream functional modules associated with glycolysis and cellular respiration tend to be upregulated by stresses in yeast, while ribosome components and genes for RNA processing tend to be downregulated by many stresses (Gasch, Spellman et al. 2000; Causton, Ren et al. 2001). Thus under these conditions there will tend to be negative correlations between expression levels of genes in these modules, so variation in upstream regulators common to all modules such as MSN2/MSN4 will cause GEI showing negative correlation between the pleiotropic traits such as ribosome and glycolysis gene expression.

The GEI measurement in ANOVA may be thought of as a difference in responses to environmental changes between two or more genotypes. Thus each trait’s GEI has a sign, negative or positive, depending on which genotype shows the larger response to environments. I call this sign the “direction of GEI” in this thesis. Two traits with different signs of GEI have negative genetic correlation. In one study mutants of the transcription factor HAP4 (which controls transcription of respiration genes in downstream of the central response hub) were used as the variant genotypes. The normal negative correlation between ribosomal genes (where the RAP1 transcription factor transmits the stress response) and respiratory gene modules was broken (Herrgård, Covert et al. 2003; Tanay, Steinfeld et al. 2005) because the causal link (signals from upstream MSN2/MSN4) is missing due to the HAP4 mutation. So, genetic variation in key signal transduction genes may cause widespread GEI, whose direction or sign will change in different downstream modules. There will be widespread pleiotropy of GEI in such cases, with the direction of GEI of the pleiotropic traits governed by their position in various response modules. This provides us with a molecular and genetic model for understanding widespread, patterned pleiotropic GEI resulting from variation at a single locus.

In some cases a gene regulation mechanism is partly understood for such negative correlations. For instance, genes with several promoter motifs including the well-known TATA box and regulated by TBP (Tata binding protein) have been shown to be upregulated in yeast in response to stresses. Genes with other promoter motifs regulated by NC2 and MOT1 transcription factors are downregulated by stress and upregulated in growth (Boorsma, Lu et al. 2008). An amazing r = -0.85 correlation between expression of two groups of genes with motifs from the two groups was observed over a set of 936 experiments. Mutations of NC2 or TBP affected expression of the two sets of genes in opposite ways (Boorsma, Lu et al. 2008). Interestingly, TATA, TBP, NC2, MOT1 all have homologs in flies and humans. The fly motif equivalent DPE (downstream promoter element) of one of the yeast growth-related motifs (rRPE) is known to be associated with regulation by the NC2 genes and Hel89B, the fly homolog of MOT1 (Willy, Kobayashi et al. 2000). In flies these promoters and transcription factors have been shown to participate in

26

a “regulatory circuit” which results in genes with TATA promoters being upregulated when genes with DPE promoters are downregulated, and vice versa (Hsu, Juven-Gershon et al. 2008).

Are there “general stress-response” genes in flies? In D. melanogaster clusters of genes transcriptionally co-regulated by several stresses as well as ageing have been reported (Pletcher, Macdonald et al. 2002; Girardot, Monnier et al. 2004; Landis, Abdueva et al. 2004; Qin, Neal et al. 2005; Sorensen, Nielsen et al. 2005; Sorensen, Nielsen et al. 2007). These authors and those studying yeast note there is a smaller core set of genes regulated under most stress conditions, plus a larger stress-specific set (such as heat shock genes). Sorensen et al. 2007 used lines selected for resistance to cold, several kinds of heat stress, starvation, desiccation, and longevity (Sorensen, Nielsen et al. 2007). Genes differentially expressed from selected lines showed high overlap between starvation and desiccation lines, some overlap with longevity lines, and much less overlap with heat and cold selected lines. Note that these were genes differentially expressed in unstressed flies raised under normal conditions, after generations of selection. So, in this case we see the residual effects on normal expression of selection for multiple different stresses. Interestingly, the Gene Ontology categories of differentially expressed genes common to all selection lines were mitochondrial respiration and cytosolic ribosome, while most other groups were specific to one or more selection regimes. This suggests that correlated patterns of expression change after selection for stress resistance are most robustly seen in the protein synthesis and respiration pathways.

Bergmann et al. looked at coexpression patterns of genes in many varying environments across 6 organisms ranging from E. coli to H. sapiens (Bergmann, Ihmels et al. 2004). There were many withinspecies significant correlations between modules, but across species the patterns were not consistent in direction. For instance, ribosomal genes and heat-shock genes were significantly correlated within each of the 6 taxa, but S. cerevesiae and D. melanogaster had negative correlations while humans, worms, Arabidopsis, and E. coli all had positive correlations. Some relationships were common to non-plant eukaryotes (yeast, worms, flies, people) such as the positive relationship between ribosome and RNA processing modules, while secreted protein genes and ribosomes were positively correlated in worms, flies, and people but not yeast. The stress modules heat shock and peroxide-response genes were positively correlated in 5 of 6 organisms (but not C. elegans). The negative correlation seen within the yeast stress response datasets of ribosomal and RNA response genes to glycolysis and respiration genes was not present in this study, where many environments included non-stress conditions.

These disparate results relate back to the question of whether there are common responses to stress whose direction explains positive and negative pleiotropy. There seems to be no universal answer to this question valid across all eukaryotes, although positive correlations across a number of modules in higher animals means there may be partial answers valid among certain groups (such as the high association between heat and oxidative stress modules).

27

To conclude this section, I cite an example of a stress-related signaling cascade with pleiotropic effects involving oxidative stress and use of reactive oxygen species (ROS). NO is an important signaling molecule for transducing environmental stimuli. NO can cause oxidative damage in large amounts. It’s interesting that clinical researchers have hypothesized that ROS signaling is an example of antagonistic pleiotropy, with benefits at all ages but with gradually accumulating free radical damage connected to senescence (Lambeth 2007). Since PKG is an important downstream link in some branches of the NO signaling pathways, the relevance to foraging of this hypothesis is clear.

Pleiotropy and life history traits Two traits (survival to pupation and adult weight) in a set of isofemale lines derived from Denmark and Costa Rica showed strong negative correlation, consistent with antagonistic pleiotropy (Bochdanovits and de Jong 2004). In an array analysis of transcriptional covariation with these two traits, two sets of genes correlating with the opposite extremes of the trait distributions seemed to be distinguished by different metabolic roles (Bochdanovits and de Jong 2004). As survivorship and adult weight are two traits known to be associated positively with fitness, the authors interpret this temperate/tropical lines contrast as examples of local adaptation to two extremes of a temperature cline acting on a pair of negatively correlated traits. In this case, local adaptation plus negative genetic correlation sets up a geographical mosaic with pleiotropic effects: variations in genotype that affect survivorship also tend to affect adult weight. In companion studies these authors ask whether local populations are adapted differently to temperature and food deprivation in the lab and discover GEI and pleiotropy (Bochdanovits and De Jong 2003; Bochdanovits and de Jong 2003; Bochdanovits, van der Klis et al. 2003). Although these results are consistent with the theory of antagonistic pleiotropy and tradeoffs for fitness, other studies have found that genetic correlations can interact with the environment (reviewed in Sgro et al. (Sgro and Hoffmann 2004)) Thus, the genetic correlations which are associated with pleiotropy are themselves subject to GEI.

Experimental selection studies reveal whether tradeoffs hypothesized to underlie antagonistic pleiotropy in fact can affect evolution. Studies by Kawecki and Mery shows that there seem to be tradeoffs between learning in flies and life history traits such as stress resistance (Mery and Kawecki 2003; Mery and Kawecki 2005) or longevity (Burger, Kolss et al. 2008) and that the foraging locus can regulate learning (Mery, Belay et al. 2007). Thus tradeoffs consistent with antagonistic pleiotropy involve learning, a prototypical plasticity trait.

28

I discuss one final fly study to draw attention to pleiotropic “side effects” of adaptation to an environmental stress in neural function. Passador-Gurgel et al. 2007 compared transcript levels in 108 lines of flies from two locations which varied in their survivorship when fed nicotine. Flies were fed normal or nicotine-laced food (Passador-Gurgel, Hsieh et al. 2007). By regressing transcript levels against line survivorship a small number of candidate genes were identified. One gene (CG8745) converts ornithine to glutamate. This led the authors to a downstream gene, Gad1, which catalyses conversion of glutamate into the inhibitory neurotransmitter GABA, and which showed highly significant GEI variation among lines. Crosses of a loss of function P-element mutant of CG4785 showed significant main and genotype x line effects on nicotine survivorship, while direct metabolite measurements of ornithine and GABA levels in heads of flies exposed to nicotine showed a significant regression of nicotine survival time on levels of these compounds. The authors suggest nicotine survival may depend in part on pre-exposure ratios of the stimulatory glutamate to inhibitory GABA, and on regulation of this ratio once exposed to nicotine stress. Although this study did not test behavioural phenotypes, a hypothesis for future research would be that lines with genetic variation in glutamate -GABA ratios may well express multiple behavioural differences. For example, nicotine is an appetite suppressant in mammals, so assays of feeding behaviours might be fruitful. This system would then represent a new model of natural variation in metabolic genes affecting behavioural phenotypes in a GEI-related manner.

Pleiotropy is a gene-centric view of phenotypes: when does one gene affect multiple phenotypes? Genetic correlation on the other hand is a phenotype-centric view of genes: when is there a heritable genetic component to observed correlations between two phenotypes? Population geneticists tend to study genetic correlation or covariance matrices among multiple phenotypes (the G matrix) since the structure of the G matrix determines the direction of selection when selection acts simultaneously on multiple traits of an individual (Lande 1979; Lande and Arnold 1983). Actual measurement of genetic correlations is demanding, requiring large, carefully structured experiments (Lynch and Walsh 1998). As a result, determining when two G matrices differ is even more demanding. Roff and Simons found that genetic correlations between morphological traits were similar under laboratory and field conditions. However life history traits fecundity and developmental time had negative genetic correlation in a variable field environment but no correlation in the lab (Simons and Roff 1996; Roff and Simons 1997). Arnold studied garter snake feeding responses to the taste/odors of multiple prey species from inland and coastal populations, finding convincing evidence of differences in the two populations in genetic covariances between response to different prey types (Arnold 1981). As these examples demonstrate, genetic correlation is a twin concept to pleiotropy and GEI in individual traits may lead to GEI in genetic correlations or G matrices. I return to these points in the Discussion.

To summarize this section, I have looked at pleiotropy and GEI in a variety of contexts. There is evidence from genomic and molecular studies that genes and proteins in “hub” positions in networks or

29

near the beginning of regulatory response cascades can affect a wide variety of downstream modules, resulting in changes in many phenotypes. When such genes have a GEI upstream, this can result in a large number of individual GEI downstream. Since some downstream modules may be upregulated and some suppressed by the same regulatory system, there can be negative or positive correlations among the downstream phenotypes. Hence, there can be strongly patterned pleiotropic GEI “directions” of response over multiple traits. When we encounter a gene with such patterned pleiotropic GEI, the pattern of GEI directions may be related to underlying regulatory patterns. These ideas will be relevant in Chapter 2, in the context of multiple cuticular hydrocarbon phenotypes exhibiting patterned GEI, and again in Chapter 4 where transcript, metabolite, and behaviour phenotypes show patterned GEI pleiotropy. In the Discussion I will return to the question of what contribution investigations of patterned pleiotropic GEI in laboratory strains of model organisms can make to the larger debate on antagonistic pleiotropy and evolution.

Genotype by Context interactions I have now discussed the relationship of GEI to four different contexts in which they may occur: geographical mosaics, social interactions, epistasis, and pleiotropy. Wolf and colleagues have called the collection of such GEI in varying contexts GxC (Genotype by Context) (Wolf, Brodie III et al. 2004). They describe a general theory of evolution of GxC, derived from the Price equation (Price 1970), which unifies these disparate areas in a single theoretical framework. In the Discussion I will relate some of the disparate GEI results of Chapters 1-4 to GxC and its implications.

Cuticular Hydrocarbons Each of Chapters 1-3 contains introductory material on fly cuticular hydrocarbons (CH), so I focus here on the “why” question: why is the study of CH important to understanding fly biology? Biological roles of CH are relatively poorly understood compared to some other insects, so I draw on what has been studied elsewhere as well as the fly CH literature. CH compounds have several known roles: as communication signals (pheromones) (Wyatt 2003); as components of the cuticle’s defense against water loss (Gibbs, Chippindale et al. 1997; Gibbs 1998; Gibbs, Fukuzato et al. 2003); and as defense against pathogens (Lecuona, Clement et al. 1997; Napolitano and Juarez 1997; Juarez, Crespo et al. 2000), although this latter seems less studied, and overlaps with the first (Richard, Aubert et al. 2008).

The focus on CH in this thesis relates to their roles as pheromones, where they affect male-male, male-female, and female-female interactions (Jallon 1984; Ferveur 2005). Pathways of synthesis of CH are being elucidated (see Chapter 2, Figure 1) but are better characterized in other insects than in Drosophila

30

(Howard and Blomquist 2005). Three main classes of CH are found in D. melanogaster: straight-chain, fully saturated alkanes; mono-unsaturated (monoenes) or di-unsaturated (dienes) (collectively, alkenes); and branched methyl-alkanes, (“methyls”). CH are synthesized in oenocytes. They are transported and secreted to the cuticle by processes still under investigation. The primary enzymes responsible for the “allocation decisions” as to where precursor fatty acids will end up are the well characterized delta-9 desaturase encoded primarily by the desat1 gene (Labeur, Dallerac et al. 2002; Marcillac, Bousquet et al. 2005), and unknown SAM-transferases which produce methyl CHs. Other desaturases exist and are important in females for production of dienes after the initial monoene desaturation step (Dallerac, Labeur et al. 2000; Wicker-Thomas and Jallon 2001; Chertemps, Duportets et al. 2006). These “allocation” steps appear to occur early in the synthesis pathway, usually acting on a C16 fatty acid precursor (Jallon, Kunesch et al. 1997), after which elongase enzymes lengthen compounds in each of the categories (Chertemps, Duportets et al. 2005; Chertemps, Duportets et al. 2007). Thus the distribution of lipid precursors into final CH products is governed at allocation steps by desaturase and methyl-transferase enzymes, and then subdivided within each compound class depending on how much the carbon chain is lengthened by action of various elongases. Recent phylogenetic studies are adding further understanding of the evolution of hydrocarbon production enzymes (Hashimoto, Yoshizawa et al. 2008) and it seems likely candidate genes for most synthetic steps will be identified in D. melanogaster in the next few years.

Details of the known effects of CH on D. melanogaster behaviour have recently been reviewed (Ferveur 2005), and are discussed in Chapters 1-3. Here I highlight additional research both on D. melanogaster and on Australian drosophilids which is rapidly broadening the set of compounds known to influence mating behaviours. A series of population genetic studies from the lab of Mark Blows using CH and mating as phenotypes and full- or half-sib mating designs have elucidated CH roles in Drosophila serrata (Blows and Allan 1998; Hine, Lachish et al. 2002; Chenoweth and Blows 2003; Blows, Chenoweth et al. 2004; Hine, Chenoweth et al. 2004; Rundle, Chenoweth et al. 2005; Skroblin and Blows 2006) and have begun to do the same for D. melanogaster (Foley, Chenoweth et al. 2007). Some of the findings from this research have included the essential role of male methyl CHs in female choice of mates, and the existence of genetic constraints in the G-matrix (Lande and Arnold 1983) which appear to prevent sexual selection by females from reducing the expressed diversity of CH signals in males in wild populations (Blows, Chenoweth et al. 2004; Hine, Chenoweth et al. 2004; Rundle, Chenoweth et al. 2005; Chenoweth, Rundle et al. 2008).

Thus quite a bit is known about the genetics of CH synthesis and the role of CH compounds in fly mate choice, but some questions are still open.

For instance, Levine showed that a putative odor-borne signal caused resetting of circadian clocks due to social interactions between male flies (Levine, Funes et al. 2002), but the extent to which social

31

interactions modulate male CH signaling was not known, nor how important this might be compared to other environmental signals such as light cycle, temperature, or food. It is known that temperature appears affects elongase activity, as deduced from a shift in monoene chain length in male flies (Rouault, Capy et al. 2000), while food type can shift CH compound mix (Rundle, Chenoweth et al. 2005; Rundle, Chenoweth et al. 2006) and importantly that male D. serrata can vary CH levels quickly in response to encounters with females (Petfield, Chenoweth et al. 2005).

Therefore drosophilid male CH signaling is plastic with respect to several kinds of environments, leading to the question of whether there are GEI in CH traits, a subject of Chapter 2. As Petfield et al. also showed the existence of IGEs in interactions between female and male D. serrata, the question of whether male-male interactions in D. melanogaster cause IGEs is also discussed in Chapter 2. Finally, returning to questions about relationships between GEI, pleiotropy, and fitness we may ask whether our existing knowledge of the synthetic pathways of CH matches patterns of correlation in CH levels and whether these indicate anything novel about the control of synthesis. These points are addressed in Chapters 2 and 3 and in the Discussion.

Circadian rhythms Results presented in chapters 1-3 are based on looking at fly phenotypes under natural lighting cycles (LD, 12 hours light followed by 12 hours dark) and artificial absence of light (DD, continuous darkness). Many phenotypes of flies vary with time of day and presence/absence of light, and the yper0w flies used as visitors in our host-visitor experiments are mutants for clock function. Hence some background on the clock is presented here.

Circa 24 hour rhythms can be maintained for weeks or even months in many organisms without external environmental inputs, as a result of the genetic/molecular oscillators collectively known as the circadian clock. Recent reviews cover the state of understanding in 2008 of this system in Drosophila and other organisms (Nitabach and Taghert 2008; Zheng and Sehgal 2008). Briefly, transcription, translation, phosphorylation, and trafficking of several genes which encode transcription factors are important in the clock. The pair period (per) and timeless (tim) are synchronized and inhibit their own transcription after a delay, which is regulated by cytoplasmic-nuclear trafficking and other events. In the nucleus, per/tim cause transcription of a second pair of genes, cycle (cyc), and Clock (Clk). Clk and cyc in turn repress their own transcription and activate per and tim transcription, completing the cycle. One of the first circadian gene mutants, per01, disrupts clock function showing per is essential to the running of the clock (Konopka and Benzer 1971; Konopka 1979). Recent studies show latitudinal clines in genetic variation in the wild for per alleles which adjust clock function to northern and southern environments (Kyriacou, Peixoto et al. 2008).

32

Many other genes modulate clock function (Zheng and Sehgal 2008), and more are discovered yearly (Koh, Zheng et al. 2006), but cycling of transcription of the “core four” above in a tissue is a high-probability indicator that a clock is running. There is a central brain clock but also separate clocks in several tissues, notably including sensory neurons and receptors for olfaction and sight (Krishnan, Dryer et al. 1999; Hardin, Krishnan et al. 2003; Nagoshi, Saini et al. 2004; Krishnan, Dryer et al. 2005; Pyza and GorskaAndrzejak 2008), so the clock (whether central or peripheral) may modulate sensory activity. Conversely, sensory input from light, food, and temperature can reset or entrain clock cycles (Krishnan, Levine et al. 2001; Glaser and Stanewsky 2005). Clock cycling can regulate neuronal plasticity (Pyza and GorskaAndrzejak 2008) and is of course intimately tied to daily sleep rhythms (Huber, Hill et al. 2004; Koh, Evans et al. 2006; Zimmerman, Naidoo et al. 2008), immune status (Shirasu-Hiza, Dionne et al. 2007; Williams, Sathyanarayanan et al. 2007), and other physiological rhythms.

The circadian clock is therefore responsive to many environmental stimuli and has genetic variation in the wild for such responses. Since it in turn drives the activity of many other systems, it is a potential core or hub source of GEI which in turn may generate pleiotropic effects. In Chapters 1-3 we investigate whether there are circadian rhythms in CH production and how the rhythms may interact with light and social environment.

Characterizing a circadian phenotype involves fitting models of rhythmic behaviour (typically 24 hour curves) to data. However, the mathematical methods which do so can account for variation at other periods, allowing for a more complete characterization of the diurnal cycling (Levine, Funes et al. 2002; Kent, Azanchi et al. 2007). Long period (24 hour) clocks may be driven by shorter-period oscillators (Dowse and Ringo 1987; Lin, Han et al. 2002; Lloyd and Murray 2005). No mechanism as elegant as that of the circadian clock has been found to account for ultradian rhythms, but there is no shortage of hypotheses. A number of these focus around metabolic cycling, often studied in yeast, where impressive synchronization of gene expression has been demonstrated in short-term, cell cycle-related rhythms (Lloyd and Murray 2005; Tu, Kudlicki et al. 2005; Murray and Lloyd 2006; Tu and McKnight 2006; Tu, Mohler et al. 2007). One fascinating aspect of these studies is a negative correlation between DNA replication and oxidative activities (Chen, Odstrcil et al. 2007). A similar association of cell division in cancer patients with the sleep period, and proposals for optimizing chemotherapy based on this (Pogue, Morre et al. 2000; Wang, Pogue et al. 2001; Morre, Chueh et al. 2002; Morre and Morre 2003; Morre and Morre 2003; Orczyk, Morre et al. 2005), suggest that even though yeast chemostat cell growth may not generalize to multicellular organisms, some of the metabolic principles related to cell cycle in yeast may be regulated by circadian cycle in higher organisms. This compartmentalization of cellular functions by time of day, regulate by the clock, could then lead to extensive GEI patterned pleiotropy, driven by GEI emanating from the clock.

33

Mating Mating is central to understanding the contribution of behavioural genes to fitness in adult flies, especially for males. Drosophila species have stereotyped courtship and mating behaviours whose genetic basis is partially understood (Greenspan and Ferveur 2000; Hall 2002). I quoted Hall earlier on pleiotropy, from his review of courtship behaviour genetics, where pleiotropic effects of the cacophony ion channel gene on courtship song, vision, and song x temperature interactions are discussed (see also (Chan, Villella et al. 2002)). Mutants such as white or yellow also change both physical characteristics of flies and their mating success (Sturtevant 1915; Bastock 1956; Bastock 1967; Drapeau, Cyran et al. 2006). Thus mating, the endpoint of a complex series of courtship behaviours, is part of many pleiotropic or genetic correlation gene pairs.

In this thesis, although we demonstrate one direct effect of social environment on mating (Chapter 3), mating is not a central theme. Our study of mating and social environment is related to a series of studies on female choice including those of Ehrman (Ehrman, Wissner et al. 1973; Ehrman 1989) and of Chapman (Chapman and Partridge 1996; Chapman 2006), in the sense that we study the effect of different genotypic mixtures of males on mating. The Chapman and Partridge paper has dual relevance: it discusses remating frequency and the male environment (in their case the frequency of encounters with males) and the food environment (female nutrition). Thus both social and nutritional factors interact with female remating; but whether there is a heritable genetic basis for such interactions seems not to have been studied.

The theoretical debate about whether sexual selection depends on “reliable” signals of mate fitness is extensive and largely peripheral to this thesis, so I do not review it here. However, I mention it because the studies in this thesis contribute to mechanistic understanding of issues in fitness signalling and the nature of “condition” by elucidating how allocation of energy to lipids, and among lipids to compounds such as CH, is regulated. Although a prominent model suggests traits tied to male signalling such as CH should be strongly genetically linked to condition (Rowe and Houle 1996) recent work in Drosophila bunnanda shows the multivariate direction of female sexual selection is perpendicular to the direction of male variation in CH levels, so that even though there is condition dependence of male CH, females cannot select upon it (Van Homrigh, Higgie et al. 2007). In other insects, such as the german cockroach, raising nymphs on lower quality food causes them to maintain CH production at the expense of adult size and internal reproductive lipid stores (Young, Bachmann et al. 1999), thus breaking a link between condition and CH signals. It therefore seems the actual allocation “decisions” made by insects need a more mechanistic understanding, before links to theoretical models of the evolution of sexual selection can be fully linked to signals such as CH.

34

Social groups “All of life is social” – Steven Frank (Frank 2007) I have already summarized some important issues in GEI when the environment is social. Studies on flies, slime molds, spiders, mice, and Arabidopsis show social effects are important. The review by Frank quoted above mentions that even in bacteria social effects are found and have genetic bases. A look at a few groups of organisms illustrates the kinds of social effects found.

Social group effects with genetic bases have been studied for several fish species in the family Poecilidae, where dominance interactions between males cause shifts in maturation rate and size. Strong genotype by social pairing interactions on maturation were found in an experimental study on male platy swordtails Xiphophorus maculatus (Sohn 1977). A single locus polymorphism on the Y chromosome pleiotropically regulates a negative correlation in maturation size and rate in platy swordtails and is subject to a GEI with predator pressure (Basolo 2008), hence genotype, social environment, and predation environment interact. In green swordtails Xiphophorus helleri a full-sib rearing design with males from the same families raised in isolation or in groups found high heritability of size at maturity in isolated males but no heritability (due to very large within family variation) in group-reared males, even though family means in isolated and grouped conditions were strongly correlated (Campton 1992). This is an example of a second kind of GEI, where mean genotypic levels do not interact with environment but variances may. Green swordtails modify age at maturation in response to purely visual cues about the social environment (male coloration) without direct agonistic interactions (Walling, Royle et al. 2007). A study on the poecilid mosquitofish Gambusia affinis showed strong social effects on size but no GEI (Campton and Gall 1988). The guppies of Trinidad Poecilia reticulata have been shown to have interactions including (a) parental location by age and by social group density in male mating behaviour (Rodd and Sokolowski 1995), (b) offspring size and number for females by location of origin of social group, (c) male size by density of social group, in males derived from location but not another, and (d) male growth rate and final mass by social group composition by density (b-d from (Rodd, Reznick et al. 1997)). Experiments in Trinidad show that there is negative frequency dependent survival of males based on the local population distribution of coloration patterns (Olendorf, Rodd et al. 2006), e.g. an interaction between survival and social group composition. Thus, in the poecilid fishes the composition of the social group interacts with genotype to affect life history traits of group members.

Within the insects, the word “social” most often brings to mind the complex societies of the eusocial bees, wasps, ants, and termites, but many other “pre-social” species show effects of the group on the individual. For example, in some species of moths, caterpillars forage or defend against enemies collectively (Costa 1997). In the forest tent caterpillar Malacosoma disstria, a series of studies by Emma Despland and colleagues has shown that caterpillar foraging strategies depend on age (younger ones benefit

35

more from group membership (Despland and Le Huu 2007)) and from “leader following” – following pheromone and silk trails laid down by others in the group (Despland and Hamzeh 2004; Colasurdo and Despland 2005). Further, colonies show spontaneously emerging group foraging bouts with plasticity in this emergent group behaviour to light and temperature variables (Peters and Despland 2006). Individuals show stably different amounts of time spent foraging and quiescent (Nerniroff and Despland 2007), and experimental colonies formed from different proportions of individuals of the two foraging types exhibited different emergent colony foraging behaviours when confronted with suboptimal quality food (Dussutour, Nicolis et al. 2008). That is, caterpillar colonies showed frequency-dependent effects on social foraging. It is not known whether there is a genetic basis for this individual foraging behaviour polymorphism, but if so there would be a frequency-dependent GEI with both food type and the composition of the colony.

Despland and colleagues have found that when individuals chose between two food sources, one balanced and one unbalanced in nutritional content, the first choice was random, but they would leave unbalanced food more often and strongly preferred the balanced food. However when groups of similar individuals were placed in the same arena, the leader-following behaviour of the group meant that all its members would follow the first caterpillar and then stay on whichever food was randomly first investigated for 24 hours or more (Dussutour, Simpson et al. 2007). Thus group foraging behaviour cannot be predicted from foraging behaviour of individuals in all cases.

The effect of being in a group is found in several species of roaches. Blatella germanica females eat and reproduce more when paired than isolated (Holbrook, Armstrong et al. 2000), with juvenile hormone mediating some effects of pairing, while Diploptera punctata nymphs reared in pairs molted fewer times and became smaller adults than those reared alone (Holbrook and Schal 2004). Roaches can make collective “decisions” about where to shelter, which can be influenced by roach-sized robots impregnated with a roach cuticular hydrocarbon extract (Halloy, Sempo et al. 2007).

Theoretical models of group aggregation suggest that interactions between individuals which amplify individual choices can lead to group behaviours as an emergent outcome (Ame, Halloy et al. 2006; Ame, Millor et al. 2006; Millor, Ame et al. 2006). Such models are an outgrowth of studies originally undertaken by physicists and computer scientists into when and how simple models of individual behaviours yielded complex group behaviours, the prototypical example being the computer “boids” whose algorithms of movement yielded visually realistic flock motions (Reynolds 1987). Powerful mathematical results have been proven showing that groups of individuals with identical behavioural rules can undergo “phase transitions” from random, uncorrelated behaviours to highly organized, synchronized behaviours as density or interaction rates increase (Okubo 1986; Vicsek, Czirok et al. 1995; Czirok, Stanley et al. 1997). Small changes in individual behaviours can lead to large changes in groups, and groups can exhibit “memory” - present group behaviours influenced by past group membership (Couzin, Krause et al. 2002).

36

These theoretical insights have led to experiments showing that phase transitions can be found in groups of desert locusts Schistocerca gregaria in response to changes in density (Buhl, Sumpter et al. 2006) and that environmental features such as clumping or dispersion of food patches can, by increasing or decreasing local densities of locusts, mediate the phase transition (Despland and Simpson 2006).

In a series of studies, Kim and Ehrman looked at how rearing in isolation or in groups affected phenotypes of Drosophila paulistorum such as mating (Kim, Ehrman et al. 1992; Kim, Ehrman et al. 1996; Kim, Koepfer et al. 1996). One result pertinent to this thesis is that flies reared in isolation had higher CH levels than those reared in groups (Kim, Phillips et al. 2004). The question of how presence in groups may affect CH levels can be approached in several ways. In Chapter 1, and in the Discussion, I will examine how probabilistic models of interacting individuals in groups can make predictions about the distribution of a phenotype such as CH levels and its mean values. This represents an effort to take IGE-style models to a more granular or detailed level.

IGE models were discussed earlier in this Introduction, so here I focus on what some of the open issues are in the theories of indirect genetic effects and of “interacting phenotypes” (Moore, Brodie et al. 1997; Moore and Pizzari 2005). The IGE equations presume linear effects of interactions between individuals of different genotypes on the resulting phenotypes. As we have just seen, in models of group behaviours, phase transitions between modes of behaving can occur, and which behaviour mode the group ends up in may be sensitively dependent on small changes in individual behaviours. Figures 3 and 4 in Couzin et al (Couzin, Krause et al. 2002) illustrate this clearly – changes on the order of 10% for individual behaviour phenotypes can result in 100% plus changes in group behaviours in some cases. These non-linearities complicate the solution of IGE models and have not to my knowledge been addressed theoretically. A second insight comes when we look at linear stochastic models of time series events, where adding a small amount of variation to the interaction parameters (as would be reasonable in most groups of organisms) can result in non-intuitive “breakout” patterns of group response when critical parameter boundaries are approached (Nicholls and Quinn 1982). So even when the model of individual-individual interactions is linear, nonlinearities in mean group behaviour can result. More, in some cases the variation in group behaviour can become very large, making normal statistical techniques for analysis of behaviour infeasible. Thus adoption of “individual based modeling” or IBM (DeAngelis and Mooij 2005) instead of population mean modeling uncovers new modes of group behaviour which may deviate from underlying linear interactions of individuals.

These emergent group behaviours are an opportunity for the study of social behaviours, not an embarrassment, since the sensitivity of the group phenotype to small changes in individual phenotype provides a point of leverage for the experimenter. In the discussion I will review this issue in the light of

37

the results presented and ask what future directions are suggested for theoretical study and experimental investigation.

foraging The lower case title to this section makes clear that here we discuss the D. melanogaster gene foraging (for) rather than food search behaviours in general, although of course the topics overlap! This large section is split into several parts: I introduce first the phenotypic effects of for in flies, then other organisms, then some of the biochemical mechanisms by which phenotypes are affected, and finally what changes in gene transcription are due to for and its homologs.

Phenotypes due to for in flies Sokolowski identified a behavioural polymorphism in larvae where “rovers” covered larger distances than “sitters” while feeding. The polymorphism was localized to the 2nd chromosome (Sokolowski 1980) and later to the gene foraging (de Belle, Hilliker et al. 1989). The biochemical function of the gene product was identified as cGMP-dependent protein kinase (PKG) (Kalderon and Rubin 1989; Pereira, Burgess et al. 1995; Osborne, Robichon et al. 1997), a kinase enzyme acting in a second messenger cascade initiated in many cases by nitric oxide (NO) (Davies 2000). There are three genes annotated in FlyBase as having PKG function, all on the second chromosome, with for at 24A3-5, Pkg21D at 21E2, and CG4839 at 31A3. Neither Pkg21D nor CG4839 have been shown to have behavioural effects, so are not discussed further here. Rovers possess the forR allele and natural sitters are homozygous fors (forR is dominant to fors in many larval phenotypes). Mutagenesis produced a number of for alleles, of which the fors2 mutant is a strong sitter but is on the same background as the rover forR stocks, controlling for genetic background effects (de Belle, Sokolowski et al. 1993; Osborne, Robichon et al. 1997).

A body of work over more than two decades has described behavioural attributes of rovers and sitters and their genetic basis. Larval crawling movement and path lengths (measured by the length of the trace a larva leaves in a yeast slurry) on food are greater in rovers than sitters (Sokolowski 1980; Sokolowski and Hansell 1983), while the two do not differ on agar substrates (Sokolowski, Kent et al. 1983). The difference in path lengths is due to more rover crawling movements, not larger ones (Sokolowski and Hansell 1992). Differences are evident between the two forms in 2nd and 3rd larval instars (Graf and Sokolowski 1989) until the onset of the pre-pupation “wandering” behaviour in late 3rd instar, when the larvae cease to feed (Sokolowski, Kent et al. 1984). Although rover-sitter differences in path length persist after four hour starvation (Graf and Sokolowski 1989), in larvae raised throughout their life on 15% food rovers behave like sitters (Kaun, Riedl et al. 2007). Well fed sitters ingest more food than rovers but absorb glucose less efficiently than rovers; the ingestion difference is eliminated by 15% food

38

but not the absorption difference (Kaun, Riedl et al. 2007). Thus there are GEI differences in larval path length (when E varies between yeast paste and agar), and there are GEI effects on ingestion when full food is compared to 15% food but not on absorption (Kaun, Riedl et al. 2008).

Adult behaviours also differ between rovers and sitters. Studies by Bell et al. showed that search strategies after feeding on a drop of sucrose differed between rovers and sitters, with sitters spending more time in “intensive search” behaviour (characterized by high frequency of turning movement) and thus moved a shorter distance from the drop than rovers (Bell and Tortorici 1987; Nagle and Bell 1987). Several variables affected intensive search: more time in starvation before testing increased intensive search behaviour of both strains, while higher concentration of the sucrose drop caused rover trails to become more like those of sitters (Bell and Tortorici 1987). There is a parallel to larval GEI: adult rovers respond to sucrose concentration with greater phenotypic changes than sitters (GEI), but short-term starvation did not cause major GEI. Pereira et al. replicated the food versus non-fed results in a genetic cross design which verified the role of for in the differences between fed rovers and sitters, and which in addition found non-significant differences in adult trail lengths in the absence of food (Pereira and Sokolowski 1993).

Sensory responses to food were investigated in adult rovers and sitters. Food odors proved more attractive to sitters than rovers (Shaver, Varnam et al. 1998) while rovers were more responsive to sucrose in a gustatory response test (Scheiner, Sokolowski et al. 2004). Notably, in the latter the difference between rover and sitter responses was larger after 2 hours of starvation than after 24 hours (GEI with food deprivation).

Responsiveness to food can change over time; processes like habituation to sensory stimuli and satiation can be involved. In adult gustatory response, rovers showed less habituation to sucrose than sitters (Scheiner, Sokolowski et al. 2004), while in a sensorimotor habituation assay sitters also showed more rapid habituation (Engel, Xie et al. 2000). Habituation can happen at the level of individual sensory neurons; at the level of more complex integrative learning such as short- and long-term memory, there are rover-sitter differences but the story is different. Here adult rovers show better short term memory but worse long term memory, and increasing for expression in mushroom body (MB) makes sitters rover-like in both forms of memory (Mery, Belay et al. 2007). Larval training to olfactory stimuli produced higher learning scores (after 3 training trials) in rovers than sitters, but after a larger number of trials no differences were found (Kaun, Hendel et al. 2007). Thus, there is GEI in learning when E is the number of trials.

Rovers and sitters differ in responses to stresses in the larval stage. Rovers leave food medium under anoxic conditions more quickly than sitters, a difference related to nitric oxide signaling (Wingrove

39

and O'Farrell 1999). When heat is applied to food medium containing 3rd instar larvae, rovers leave sooner than sitters and travel faster to cooler parts of the medium (Kent, Sheffield, Dawson-Scully, unpublished data) . Rover larvae placed directly on hot agar travel further than sitters (Kent, unpublished data) and adult responses to heat and anoxia follow similar patterns and are mediated by the PKG
PP2A
ion channel pathway (Dawson-Scully et al., submitted).

Rovers have higher amounts of daytime “sleep” (inactivity for defined periods) than sitters, matching a pattern seen in C. elegans (Raizen, Zimmerman et al. 2008).

PKG may be, and cGMP is involved in steps preceding ecdysis or molting (Kingan, Cardullo et al. 2001; Ewer and Reynold 2002). The increase of cGMP during moulting in a well-defined network of neurons associated with crustacean cardioactive peptide (CCAP) is documented for many invertebrates (Ewer, Devente et al. 1994; Ewer and Truman 1996), although whether PKG is involved is not documented.

Metabolic and physiological effects of PKG are less studied in flies, although there are many indications of these in other organisms. Kaun et al. showed that in rover larvae, ingested C14 from glucose is stored more in lipid than in sitter larvae, although there were not overall body lipid store differences (Kaun, Riedl et al. 2008). After 3 hour food deprivation, rovers decreased lipids but not sitters. By contrast, total body carbohydrates are higher in fed sitter larvae than in rovers, but hemolymph glucose levels don’t differ, while after 3 hour food deprivation rover hemolymph glucose levels are lower than sitters. mRNA levels of adipokinetic hormone (Akh) are lower in fed rovers than in sitters. Thus there is evidence of GEI effects in larval hemolymph glucose and total lipids. PKG from both for and Pkg21D affected renal function (Malphigian tubule secretion rates) (MacPherson, Lohmann et al. 2004), but as this study was done using artificial overexpression of these genes it’s not clear to what extent each has a physiological role in the tubules.

In summary, numerous behavioural and physiological phenotypes have been demonstrated to be affected by PKG in flies. These include movement in the presence of food, food consumption, habituation to components of food, and learning. In some of these phenotypes GEI is found, most often when E is food/non-food. The question I address next is what neural mechanisms are known that may contribute to these phenotypes.

Neural mechanisms for PKG phenotypes in flies

40

for appears to exert its effects in multiple neural tissues. Recent autoimmune imaging of the adult brain using PKG antibody has demonstrated that there are multiple sites of PKG expression including neurons that project to the central complex, axon bundles of the antennal nerve and of certain subesophageal-ganglion nerves, and the medulla optic lobe (Belay, Scheiner et al. 2007).

At the level of neural function, reduced PKG activity in sitters is associated with hyperexcitability and enhanced nerve terminal sprouting at larval neuromuscular junctions and with reduced K+ currents and increased membrane excitability in a significant population of neurons in dissociated embryonic cultures (Renger, Yao et al. 1999). Similar patterns of K+ currents in locust neurons following heat-treatment (Ramirez, Elsen et al. 1999) suggested the hypothesis that similar heat-response differences might be found comparing sitters to rovers, and indeed adult sitters were found to tolerate higher temperatures before neural failure (Dawson-Scully, Kent et al. 2005). Prior mammalian studies of the connection between PKG and K+ currents suggested a pathway in which PKG phosphorylates and activates protein phosphatase 2A (PP2A), which in turn dephosphorylates and activates K+ channels (White, Lee et al. 1993; Zhou, Ruth et al. 1996; White 1999). Using pharmacological tests of this hypothesis performed using PKG activators and inhibitors, PP2A inhibitors, and heat shock, we demonstrated this pathway is operative in both locust and fly neurons (Dawson-Scully, Armstrong et al. 2007) and thus provided one mechanistic explanation for the effect of PKG on nervous system function. Interestingly, the group of Chun-Fang Wu has shown similar neural effects using mutants of K+ channels or of cAMP metabolism (Engel and Wu 1992; Zhong, Budnik et al. 1992; Zhong and Wu 1993; Engel and Wu 1996; Engel and Wu 1998; Zhong and Wu 2004). This connects the PKG-neural function story (via common neural phenotypes) to the larger and older literature on behavioural mutants due to potassium channels such as Hyperkinetic and Shaker, and to mutants such as dunce and rutabaga which also impact associative leaning and memory. Several papers from this group emphasize effects of these mutants on neuronal plasticity: “These genetic alterations are thought to perturb mechanisms relevant to activity-dependent neural plasticity, in which neuronal activity activates the cAMP pathway, and consequently affect nerve terminal arborization by regulating expression of adhesion molecules.” (Zhong and Wu 2004). The genetic effects are not limited to cAMP metabolism mutants. In double mutant trials with cAMP mutants and K+ channel mutants, neural phenotypes related to plasticity were effected by K+ channel mutations (Zhong, Budnik et al. 1992). Thus PKG, which from our previous work affects K+ channel activity in fly and locust neurons, may engage in cross-talk with a cAMP-based system with effects on neuronal plasticity.

Phenotypes due to PKG in other organisms

41

Turning to other organisms, the effects of PKG are many and varied. Behavioural effects related to PKG in multiple species and their neurogenetic basis have been reviewed (Fitzpatrick and Sokolowski 2004; Douglas, Dawson-Scully et al. 2005; Kaun and Sokolowski 2009).

PKG in worms (C. elegans) is interesting, as the homolog of for (egl-4, egg-laying defective 4) was identified as being ancestral in form to that in vertebrates and insects (Fitzpatrick and Sokolowski 2004). egl-4 affects a polymorphism in behaviours (roaming vs. dwelling) but in a different direction to fly PKG – a decrease in PKG activity in egl-4 mutants results in an increase in roaming behaviours, opposite to that found in sitter larvae (Fujiwara, Sengupta et al. 2002). egl-4 also regulates olfactory adaptation, in the short term through phosphorylation of the cGMP-gated ion channel tax-2, and longer-term through translocation to the nucleus (L'Etoile, Coburn et al. 2002). Mutants in egl-4 and tax-2 showed abnormal responses to a temperature gradient, paralleling our rover-sitter results (Yamada and Ohshima 2003). Mutants in tax-2 have abnormal CO2 avoidance behaviour, an effect shown to be mediated by cGMP signaling and calcineurin; CO2 avoidance is also modulated by nutritional status, where insulin and TGF-β pathways are involved (Hallem and Sternberg 2008). Interestingly, a “quiescence” state in worms has been identified and hypothesized to be due to satiety: it is induced by prior feeding, regulated by insulin and TGF-β, and egl-4 functions downstream of these pathways in sensory neurons to control quiescence in response to food (You, Kim et al. 2008). As mentioned before, egl-4 mutants change worm sleep-like states (Raizen, Zimmerman et al. 2008), and a novel gain-of-function egl-4 mutant (higher PKG activity) showed lower locomotion in food, smaller body size, increased fat deposits, and reduced propensity to enter dauer (Raizen, Cullison et al. 2006). Loss of function egl-4 mutants were identified in a screen for increased adult body size, and had longer lifespan, with the insulin pathway mediating lifespan extension and TGF-β mediating increased body size (Hirose, Nakano et al. 2003).

This list of phenotypes and pathways of PKG effects in worms is instructive to compare to flies. As with flies, PKG affects sensory responsiveness and behaviours related to food, heat, and anoxic conditions. Sleep is PKG-affected in both organisms. Ion channels are involved in these responses in both. The roles of calcineurin, insulin, and TGF- β have not been demonstrated in flies in conjunction with PKG, but I will review evidence from Chapter 4 that they may function in flies in the Discussion.

PKG is involved in the maturation from nurse to forager in honeybees, a food related developmental change with behavioural impacts as well (Ben-Shahar, Robichon et al. 2002; Ben-Shahar, Leung et al. 2003). In locusts, an NO-sGC-cGMP-PKG pathway modulates rhythmic movements of the ovipositor during egg laying (Newland and Yates 2007), and is involved in the gustatory response to sucrose for this phenotype, but not in the response to salt (Newland and Yates 2008). In adult female cockroach Diploptera punctata, cyclic changes in corpus allatum are associated with juvenile hormone synthesis changes. During the reduction phase of these cycles, cGMP is elevated, and an inhibitor of PKG

42

blocked normal increases in autophagosomes and a mimic of cGMP caused reductions in protein content, cell size, and JH levels in long term, but not shorter treatments (Chang, Tsai et al. 2005).

A great deal has been learned about PKG function in mice through use of the Cre/Lox knockout system, which allows the gene to be selectively inactivated in various tissues. Knockout phenotypes of the mouse homolog cGKI of for are reviewed in (Feil, Hofmann et al. 2005) and (Hofmann, Feil et al. 2006). They include effects on (1) the development and sensitization of nociceptive neurons, (2) synaptic plasticity and learning, (3) axon guidance (by interaction with semaphorins sema1,3), (4) dendrite orientation and midline development (Demyanenko, Halberstadt et al. 2005), (5) impaired protein synthesis-dependent CREB-requiring LTP (long-term potentiation) in hippocampus, (6) impaired cerebellar purkinje cell LTD (long term depression), (7) reduced cell proliferation in tissues such vascular smooth muscle, and (8) impaired intestinal function. The biochemical mechanisms by which PKG produces some of these effects are the subject of the next section.

Biochemical effects of PKG PKG has been found to be involved in many biochemical pathways, especially in mammals where the majority of research is done. As discussed above, some of the effects of PKG are known to be mediated by its impacts on ion channels. One specific instance of this is PKG regulation of the BKCa big conductance calcium and voltage regulated potassium channel, whose fly homolog is slowpoke (Barman, Zhu et al. 2004; Gragasin, Michelakis et al. 2004; Werner, Zvara et al. 2005). This pathway transduces the effects of the anti-impotence drug sildenafil (Viagra); cAMP/PKA also regulates this (Zhou, Arntz et al. 2001; Zhou, Wang et al. 2002). Other PKG effects are mediated by pathways similar in nature to the PP2A route described above, in that phosphorylation of a phosphatase by PKG is the first step downstream of PKG in the pathway. Examples of this include (a) the inhibition of soluble guanylyl cyclase (sGC) activity via dephosphorylation which depends on PKG activity (Ferrero, Rodriguez-Pascual et al. 2000; Murthy 2001; Murthy 2004), and (b) activation of myosin light chain phosphatase by PKG phosphorylation induces dephosphorylation of myosin light chain (Mlc), leading to muscle relaxation (Surks, Mochizuki et al. 1999; Lincoln, Dey et al. 2001; Huang, Fisher et al. 2004; Given, Ogut et al. 2007). In other pathways, PKG acts more directly through phosphorylation of target proteins.

Inhibition of sGC activity by PKG is but one example of several cases where PKG is involved in a negative homeostatic feedback mechanism. PKG phosphorylation of phosphodiesterases (PDE’s) is another important homeostatic role. For instance, PDE5 is a cGMP-specific PDE which is inhibited by sildenafil (Viagra) (Corbin and Francis 1999) but stimulated by PKG phosphorylation (Gopal, Francis et al. 2001; Murthy 2001). Thus there is a negative feedback loop in which increased sGC activity increases cGMP, which increases PKG activity, which in turn decreases both sGC activity and cGMP levels. Further,

43

crosstalk through cAMP-specific and cGMP-specific PDEs occurs via both PKG and PKA (Zaccolo and Movsesian 2007). So, PKG activity can change cAMP levels as well as cGMP, and therefore PKG can affect the PKA pathway and its downstream targets such as lipolysis (Beavo, Rogers et al. 1971) indirectly as well as directly (Lafontan 2005; Lafontan 2008).

The involvement of PKG in such homeostatic regulation can make it important to look at time scales of action, since the short term phenotypes associated with cGMP and PKG activity may be inversely related to middle-term and long-term phenotypes. As an example, PGC-1α upregulates reactive oxygen species (ROS) response genes via NO, cGMP, and PKG. NO transiently upregulates, but PKG is involved in longer-term downregulation of gene expression (Borniquel, Valle et al. 2006).

The role of PKG in smooth muscle relaxation described above (via myosin light chain phosphatase) has been studied for its medical applications, but is still incompletely understood, as it lies at the intersection of many different homeostatic mechanisms of actin/myosin control. In addition to myosin phosphatase, PKG binds to Troponin T and phosphorylates Troponin I on the myosin light chain (Yuasa, Michibata et al. 1999). BKCa channels are modulated in smooth muscle by G protein coupled receptors which signal through cGMP and PKG to produce relaxation (Begg, Mo et al. 2003). Other channels which allow Ca++ inflow include the TRP family, some of which are inhibited by PKG phosphorylation in a negative feedback loop regulating cytosolic Ca++ levels (Yao, Kwan et al. 2005). In one case PKG phosphorylation of VASP (vasodilator-stimulated phosphoprotein, whose fly homolog is enabled, ena) was required to affect the TRPC4 channel; VASP-PKG interactions are important in many other pathways (below). Still other Ca++ sources include the internal stores in endoplasmic reticulum (ER) and mitochondria; PKG has been shown to regulate IP3–gated ER Ca channels (Schlossmann, Ammendola et al. 2000; Murthy and Zhou 2003). Mitochondrial BKCa and KATP channels regulate mitochondrial Ca++ balance mediated by PKG (Cuong, Kim et al. 2006; Kang, Park et al. 2007). Since mitochondrial Ca++ levels are potent regulators of oxidative metabolism, PKG may modify metabolism via calcium regulation. Thus in smooth muscle, PKG is intimately connected with K+ and Ca++ channels and their effects on Ca++ signaling in the cell. Some of the downstream effects of Ca++ signaling are mediated through a cascade involving calmodulin, calcineurin (Li and Sun 2005), and the transcription factor NFAT, with gene transcription consequences reviewed in the next section. The modulation of calcineurin activity is especially important in neural tissue - as its name implies, this Ca++-sensitive phosphatase is abundant in nerves and has many effects. One of particular interest is its association with the insulin cascade kinase Akt, which ties calcineurin activity to the insulin pathway. Both PP2A and calcineurin are associated with Akt and reduce its activity and thus reduce insulin signaling; foxo, an insulin cascade member downstream of Akt, affects both PP2A and calcineurin activity and thus modulates Akt (Ni, Wang et al. 2007). This

44

provides one potential pathway by which the proven PKG modulation of calcineurin and PP2A activity may interact with the insulin pathway.

In addition to its direct effects on neural function via modulation of ion channel activity, PKG exerts indirect effects important to plasticity and learning. LTP and LTD require long term changes in synaptic activity. One model for how PKG affects LTD involves presynaptic release of NO stimulated by calcium influxes. NO diffuses to postsynaptic neurons where it activates sGC to produce cGMP. Activated PKG then phosphorylates G-substrate, an inhibitor of PP1 and PP2 phosphatases. Reduced phosphatase activity results in higher phosphorylation of AMPA receptors resulting in their removal from the synapse and reduced responsiveness to glutamate (Hofmann, Feil et al. 2006). For LTP in mouse hippocampal neurons a different mechanism has been proposed. Here NO release is postsynaptic and affects cGMP in presynaptic cells, where PKG appears to be involved in a late-phase, protein synthesis and CREBdependent part of LTP. Thus for LTD PKG seems to be involved in a phosphorylation-dephosphorylation cascade while in LTP it may be involved in transcriptional regulation via CREB (Hofmann, Feil et al. 2006).

PKG effects in neural development are associated with modulation of axonal growth cones. In grasshopper embryonic neurons PKG enhances migration, and PKA inhibits it (Haase and Bicker 2003). In untreated cells, actin bundles are mostly in cell projections, but in PKG-inhibited cells a dense network of actin bundles spans the cell body (Haase and Bicker 2003). Here PKG seems to be regulating the aggregation of non-muscle myosin and actin to modify cell motility; the same VASP phosphoprotein which mediates PKG smooth muscle relaxation is also associated with NO-mediated LTP (Wang, Lu et al. 2005) and with modulation of TRP channel activity (Wang, Pluznick et al. 2007). The fly homolog of VASP, enabled (ena), is associated with actin filament development and cytoskeleton organization, axonogenesis, dendrite morphogenesis, and dorsal closure (Flybase). In the snail Helisoma trivolis neurite growth requires NO and PKG to modulate filopodial growth (Welshhans and Rehder 2005). Synapse formation is affected by the formation of focal adhesion complexes, which join cells together via the extracellular matrix (Rico, Beggs et al. 2004). VASP is an important member of the focal adhesion complex, and phosphorylation of VASP by PKG regulates adhesion in a number of cell types (Smolenski, Poller et al. 2000; Lawrence and Pryzwansky 2001). VASP is phosphorylated at two different sites by PKA and PKG with different effects on its actin-related and other roles; PKA seems to promote adhesion formation while PKG promotes dissolution (Chen, Daum et al. 2004). We see therefore that an important theme in PKG regulation of neural cells has to do with the role of non-muscle actin/myosin and adhesion complexes in the direction of growing neurons and the formation of synapses.

One area of PKG regulation of ion channel activity is deferred to the section on insulin. Briefly, it concerns the role of the KATP ATP-stimulated potassium channel, which plays significant roles in several

45

tissues and in mitochondrial regulation. KATP interacts with PKG, providing a link between PKG and control of energy storage and metabolism.

This overview of PKG effects has focused on known roles of PKG. A cautionary note about the role of NO and cGMP in long-term memory is sounded by a study in crickets, where both NO and cGMP, but not PKG, were required to induce PKA/CREB-dependent long-term memory. In this system, cGMP exerted its effect on memory formation via a cyclic nucleotide gated ion channel, much like tax-2 in worms (Matsumoto, Unoki et al. 2006). There are many studies demonstrating roles for NO and cGMP in various kinds of neural function in insects, and perhaps 1/10 of them test whether PKG plays a role. As the example in crickets shows, it does not always do so, but given the large number of NO-cGMP dependent neural effects demonstrated without tests of PKG, it is likely that some fraction of them will show a role for PKG when tested. Thus, it is probable that many more neural roles for PKG remain to be discovered in insects.

PKG regulation of gene expression The regulation of gene expression by PKG, and regulation of PKG gene expression, has been reviewed (Lincoln, Dey et al. 2001; Pilz and Casteel 2003; Pilz and Broderick 2005). One of the caveats in this area is that many genes are regulated by NO and cGMP; not all of these require PKG for control of expression. Since differences in PKG activity are often tied to changes in cGMP abundance, the effects of the two can be confounded. One of the important discoveries is that, in human cells, type I PKG (the homolog of for) stimulated by cGMP can translocate to the nucleus, where it interacts with the fos promoter (Gudi, Lohmann et al. 1997; Gudi, Casteel et al. 2000). However, type II PKG also regulates fos in neuronal cells but does not translocate to the nucleus, indicating there are multiple pathways of PKG control of gene expression (Gudi, Hong et al. 1999). PKG-I reduces nuclear factor AT (NFAT) transcriptional signaling via a reduction of Ca++ concentration. Ca++ binds to calmodulin, which activates calcineurin (a Ca++-sensitive phosphatase), which dephosphorylates NFAT, which then translocates to the nucleus (Fiedler, Lohmann et al. 2002; Li and Sun 2005). The mechanism of Ca++ reduction by PKG is not fully worked out (Klein 2000). One source of Ca++ is release from stores in ER or mitochondria, triggered in the ER case by phospholipase C (PLC) hydrolysis of phosphatidylinositol 4,5-bisphosphates (PIP2) into diacylglycerol (DAG) and inositol 1,4,5-triphosphate (IP3). IP3 signals receptors on the ER cytoplasmic surface to release Ca++. In this pathway PKG is thought to intervene in at least two places. PKG phosphorylates and inhibits PLC-β2 and PLC-β3 (two isoforms) and thus inhibits IP3 generation and so reduces Ca++ release (Xia 2001). Alternatively, PKG phosphorylates IP3 receptor IP3R-1, reducing its Ca++ release potential (Murthy and Zhou 2003; Sergeant, Johnston et al. 2006). Whatever the detailed mechanism, the general mechanism is clear: by interfering with Ca++ increases in the cell, the calmodulin-calcineurin pathway is reduced as is NFAT-driven transcription of target genes. These include complexes of the mitochondrial oxidative phosphorylation pathway and TCA

46

cycle control genes such as pyruvate dehydrogenase (Bushdid, Osinska et al. 2003; Bonnet, Rochefort et al. 2007). There is therefore a link between PKG and metabolic control via Ca++ and NFAT, with higher PKG activity potentially reducing mitochondrial gene transcription.

PKG interacts in some mammalian tissues with MAPK and Pi3K/Akt pathways to increase vascular cell proliferation (Hofmann, Feil et al. 2006), and interacts with the transcriptional regulator TFIII (Casteel, Zhuang et al. 2002). A number of other transcriptional effects are outlined in (Pilz and Broderick 2005) but are not discussed here, either because they have no fly homologs or are not conclusively identified with PKG-I.

Nutrition, Insulin signaling, and stress vs. growth Many interlocking pathways regulate the balance between growth (anabolism, buildup of proteins and other products), energy production (catabolism, breakdown of nutrients), and quiescent states in which both anabolism and catabolism may be damped down to promote surviving stresses such as lack of food or harsh environmental conditions. Multicellular organisms have several key pathways triggered by extracellular ligands such as insulin or insulin-like peptides on the one hand (Figure I.1) and glucagon or adipokinetic hormone on the other (Figure I.2). Although individual genes, proteins, and hormones differ between species, a surprising number of components of these pathways are conserved between flies, worms, and humans. Indeed, these hormone-triggered pathways are built upon still more ancient components present in yeast and other unicellular organisms. Examples of these include the phosphatidylinositol lipid kinases and phosphatases such as Pi3K and Pten embedded near the core of the insulin signaling system, the Tor signaling system which responds to nutrient availability through signals such as amino acid abundance, and the cellular energy state sensor AMPK which responds to the AMP/ATP ratio (Figure I.3).

These core signaling pathways are strongly conserved in evolution, so species-specific changes in regulating catabolism, anabolism, and stress response are often achieved through modifications in peripheral inputs to or modifiers of the core pathways. Thus to understand how GEI between food or stress and the genome are structured, we need to look at these core systems for regulating metabolism.

47

Figure I.1 Major insulin signaling pathway components. Drosophila insulin-like peptides (dilps) are secreted into hemolymph and act as ligands for the insulin receptor (InR), a receptor tyrosine kinase. Ligand binding activates the Pi3K kinase (Pi3K92E, Dp110) which adds a phosphate group to phosphatidyl inositol (3,4) (PIP2) to produce PIP3. Pten is a phosphatase which can convert PIP3 back into PIP2. Thus the balance between PIP2 and PIP3 is one of the early “readouts” of insulin signaling. PIP3 in turn affects localization to the membrane of dPKB (Akt1) and dPDK1 (Pk61C); the latter phosphorylates and activates the former. Akt1 acts upon many targets; only foxo is shown here. Phosphorylated foxo binds to melted and is held in the cytoplasm in inactive form.

Insulin is well known for stimulating the uptake of glucose from blood or hemolymph. Failures of insulin secretion or responsiveness to insulin result in diabetes or metabolic syndrome. In humans there is insulin and its receptor, and numerous insulin related hormones such as the three insulin-like growth factors

48

(IGF), relaxin, and others with their own receptors. In flies, there are 7 drosophila insulin-like peptides (dilps) all of which stimulate the single insulin receptor InR (Broeck 2001; Brogiolo, Stocker et al. 2001; Cao and Brown 2001)(Figure I.1). Secretion of dilps is nutrient-dependent (Ikeya, Galic et al. 2002) and ablation of dilp-secreting cells and genetic knockdowns produce starvation or diabetes-like phenotypes (Ikeya, Galic et al. 2002; Rulifson, Kim et al. 2002). In the hemolymph, dilps are bound in a complex with two proteins, Imp-L2 and dALS (Andersen, Hansen et al. 2000; Arquier, Geminard et al. 2008), both of which have analogues with related function in mammals. On arrival at the cell surface, dilps bind to InR and trigger changes to the cytoplasmic portion of the receptor which modify binding of dIRS1 (Insulin Receptor Substrate 1, chico) and of the Pi3K kinase (Pi3k92E, Dp110) (Leevers, Weinkove et al. 1996; Oldham, Stocker et al. 2002). Mutations in InR (Chen, Jack et al. 1996; Tatar, Kopelman et al. 2001), chico (Bohni, Riesgo-Escovar et al. 1999; Oldham, Stocker et al. 2002) or in Dp110 (Leevers, Weinkove et al. 1996; Britton, Lockwood et al. 2002) impair growth and can phenocopy starvation. Other traits such as diapause can be affected by Dp110 manipulations (Williams, Busto et al. 2006).

The lipid kinase Pi3k92E (abbreviated Pi3K) is a key member of this pathway. Pi3K adds a phosphate to phosphatidyl-inositol (PIP2) to produce PIP3, and is part of a phospholipid signaling module in the cell membrane. It is activated by InR and chico, but can require input from the Ras pathway to achieve maximum activity (Orme, Alrubaie et al. 2006). Key enzymes in the pathway are regulated at multiple sites by different kinases and phosphatases; for instance the closely related Pi3K Pi3k59F is regulated by Tor signaling (Nobukuni, Joaquin et al. 2005). Pi3k92E is also bound to a regulatory unit, Pi3k21B (dp60) which sometimes stimulates and other times inhibits its function (Weinkove, Leevers et al. 1997; Weinkove, Neufeld et al. 1999; Britton, Lockwood et al. 2002).

49

Figure I.2 Insulin signaling and glycogen This sketch shows regulation of glycogen production and breakdown by the insulin and adipokinetic hormone (Akh) pathways; see Figure I.1 for more detail on insulin signaling. The kinase Akt1 (dPKB) acts on at least three downstream pathways; interactions not yet demonstrated in Drosophila are dashed gray. First, Akt1 inhibits shaggy (GSK3, Glycogen synthase kinase 3) relieving its inhibition of CG6904 (Glycogen synthase) and increasing glycogen deposition. Second, Akt1 activates protein phosphatases including PP1 which dephosphorylate and inhibit the rate limiting glycogen breakdown enzyme GlyP, glycogen phosphorylase. Third, Akt1 activates cAMP phosphodiesterases; reduced cAMP inhibits PKA, which would otherwise stimulate GlyP activity. Akh stimulates GlyP activity through the PKA pathway.

50

The mammalian paralogue PIK3R1 of Pi3k21B was shown to regulate Pten activity, suggesting that Pi3K, Pten, and regulatory units are found on the membrane in a Pi3k / Pten cassette (Taniguchi, Tran et al. 2006). Another member of such cassettes in flies is Susi which binds Pi3k21B and reduces its activity (Wittwer, Jaquenoud et al. 2005). Pi3K is a key gene in cancer (Cully, You et al. 2006). Most oncogenic mutations in Pi3K are the boundaries of kinase domains or between the kinase and the p85 domains (Huang, Mandelker et al. 2008). This is an instance of the major theme that localization and trafficking of insulin/Tor pathways components is an important determinant of function. Pten, the dual-specificity lipid and protein phosphatase which opposes Pi3K by reducing PIP3 levels, is also well known as a cancer gene (it is the second most commonly mutated gene in tumors after p53 (Yin and Shen 2008)). Since Pten inhibits the insulin pathway, loss-of-function mutations in Pten can disinhibit growth in tumors (Marsh, Dahia et al. 1998; Goberdhan and Wilson 2003; Yin and Shen 2008). Deletions of Pten on human chromosome 10 lead to Cowden’s Disease, an autosomal dominant syndrome with a high incidence of cancers (Marsh, Dahia et al. 1998). Pten is a tumor suppressor and inhibitor of cell proliferation in flies (Goberdhan, Paricio et al. 1999; Gao, Neufeld et al. 2000; Goberdhan and Wilson 2003). Together, the involvement of Pi3K and Pten in cell proliferation in cancer and normal growth tell us that insulin signaling has important effects on both growth in cell number (examples above) and growth in cell mass (viewed as an increase in cell size or as deposition of fats and other energy stores). Pten is also a protein phosphatase, but has been less studied in this role in flies. In mammals, Pten has numerous roles: an essential component of hippocampal LTD (long term depression, a component of memory formation/persistence) (Wang, Cheng et al. 2006); regulating cell-cell adhesion through direct protein phosphatase action on FAK (Focal adhesion kinase) (Tamura, Gu et al. 1998); and modulation of MAPK signaling through interactions with several components (Yin and Shen 2008). One of the least studied roles of Pten involves regulation of DNA repair and cell cycle checkpoint proteins in the nucleus, where it helps control G0/G1 transitions (Yin and Shen 2008). Notably, this role has not been supported in RNAi screens in fly cell culture (Chen, Archambault et al. 2007).

51

Figure I.3 Regulation of Insulin and Tor signaling This sketch shows a few of the most important regulatory circuits for insulin and Tor signaling and their downstream effects on translation and transcription. In general terms both insulin and Tor signaling increase translation through positive effects on mRNA cap-binding protein eIF-4E and on kinase S6K, which phosphorylates RpS6. Three paths stimulate translation (through Pk61C/S6K, Akt1/Tor, and ERK/MNK) and one inhibits (through Thor). Reduced insulin signaling allows foxo to migrate to the nucleus where it stimulates transcription of numerous gene including InR and Thor (shown in blue). Inhibitory interactions are shown in red for clarity.

52

The presence of PIP3 in the membrane, determined by a balance between Pi3K and Pten activity, recruits the next two important members of the cascade, PDK1 (Pk61C) and Akt/PKB (Akt1). PDK1 both phosphorylates and activates Akt (Cohen, Alessi et al. 1997; Rintelen, Stocker et al. 2001), and has direct downstream effects on targets such as S6k (Rintelen, Stocker et al. 2001; Radimerski, Montagne et al. 2002) (Figure I.3) and PKC. PDK1 in the nucleus associates with PPARgamma and increases expression of adipogenic genes in mice (Yin, Yuan et al. 2006). This has not been shown in flies.

Akt is a focal point of insulin signaling and the intersection of many regulatory inputs (Figures I.2, I.3). Akt is important as a regulator of glycogen metabolism (Figure I.2), transcription via foxo (Figure I.3), and Tor signaling (Figure I.3). Akt localizes to the plasma membrane in the presence of PIP3 where phosphorylation by PDK1 (Cohen, Alessi et al. 1997; Rintelen, Stocker et al. 2001) and the Tor complex (Sarbassov, Guertin et al. 2005) activate it. Akt phosphorylates foxo (see below), GSK3 (Glycogen synthase kinase 3; shaggy=sgg) (Cross, Alessi et al. 1995; Cohen, Alessi et al. 1997) thus stimulating glycogen production, and the Tor-regulating Tsc2 complex (Gao, Zhang et al. 2002; Hietakangas and Cohen 2007).

Akt is regulated by several phosphatases, including calcineurin and PP2A (Ni, Wang et al. 2007; Vereshchagina, Ramel et al. 2008). A feedback loop within insulin signaling results from activated foxo (which exists when insulin signaling is low) reducing calcineurin and PP2A inhibition of Akt (Figure I.3). Interestingly, Akt which normally is found at the membrane can be found in the cytoplasm in fly nurse cells where it associates with lipid droplets and changes their morphology from small to large droplets (Vereshchagina and Wilson 2006). A PP2A regulatory unit reduces this lipid regulation in nurse cells (Vereshchagina, Ramel et al. 2008). So, Akt localization is an important determinant of function. As mentioned above, Akt phosphorylates GSK3/sgg, inhibiting it. GSK3 has many effects beyond glycogen regulation, including in Wnt signaling and neurogenesis; thus it is interesting that treatment of mouse neurospheres (precursor cells of neural tissue) with sildenafil (a PDE5 inhibitor) increased neurogenesis in a PKG-, Pi3K- and Akt-dependent way, and that GSK3 phosphorylation was increased (Wang, Zhang et al. 2005). This study did not test specifically whether GSK3 was causal in PKGstimulated neurogenesis, but it did show the requirement for Akt. Further, the alternative pathway through ERK (Figure I.3) was shown not to be involved. Thus, PKG changes neurogenesis in mice via a central portion of the insulin pathway.

Downstream of Akt insulin signaling branches into at least three major pathways. Glycogen and neurogenesis via GSK3 has been discussed above; next I discuss foxo and transcriptional effects, and after that I will continue to the third pathway, the Tor signaling system.

53

Akt phosphorylates foxo which is then retained in the cytoplasm bound to melted and 14-3-3 proteins (Junger, Rintelen et al. 2003; Teleman, Chen et al. 2005). If insulin signaling through Akt is reduced, foxo is dephosphorylated by phosphatases including PP2A translocates to the nucleus. However, foxo is the target of other kinases, and is also regulated by acetylation and ubiquitination state (Calnan and Brunet 2008). Thus insulin signaling inputs to foxo are only one of many signals which are integrated at this point. In particular, stress-activated signaling through JNK or AMPK can also cause foxo translocation (Wang, Bohmann et al. 2005; Calnan and Brunet 2008). Caloric restriction and sirtuins can affect foxo acetylation independently of insulin, partly accounting for the observation that some foxo effects are not dependent on insulin signaling (Berdichevsky and Guarente 2006; Min, Yamamoto et al. 2008).

Once in the nucleus, foxo acts as a transcription factor or transcriptional coactivator or corepressor for many genes, notably including InR and Thor (Puig, Marr et al. 2003; Puig and Tjian 2005). In experiments in Drosophila cell culture activation was somewhat more common than repression (Gershman, Puig et al. 2007). The transcriptional feedback homeostasis via InR helps maintain insulin sensitivity, while feed-forward through Thor reduces protein translation when insulin signaling is low (Figure I.3).

The Tor (target of rapamycin) gene is found in yeast, plants, and animals (Wullschleger, Loewith et al. 2006) and appears to be an ancient point for regulation of growth versus stress response. Upstream of Tor are many genes of which only a few can be shown in Figure I.3. The Tsc2 gene (gigas in flies) is important as a node of several signaling pathways. Insulin signals arrive here in the form of Akt inhibitory phosphorylation (Gao, Zhang et al. 2002; Sarbassov, Ali et al. 2004) while input from the AMPK energysensing pathway also enters here as an activating phosphorylation (Inoki, Zhu et al. 2003) which is followed by further TSC2 activation by GSK3 (Inoki, Ouyang et al. 2006). The Wingless pathway intersects here by inhibiting the GSK3 phosphorylation of TSC2 (Inoki, Ouyang et al. 2006). TSC2 is ultimately inhibitory to Tor signaling (via its inhibitory phosphorylation of Rheb (Zhang, Gao et al. 2003)) so activation of TSC2 by AMPK or GSK3 inhibits Tor while inhibition of TSC2 by Akt or Wnt signaling activates Tor. The next component downstream is Rheb, which has either a direct stimulatory effect on Tor (Saucedo, Gao et al. 2003) or on a Tor-binding protein (FKBP38, CG5482) (Bai, Ma et al. 2007).

The effects of activated Tor depend on its location. Tor can be bound in two different complexes called TORC1 and TORC2. Tor in TORC1 stimulates protein translation and regulates cell growth and is rapamycin sensitive, while Tor in TORC2 regulates the cytoskeleton, and is rapamycin insensitive (Inoki and Guan 2006). Interestingly, the TORC2 complex stimulates Akt phosphorylation (Sarbassov, Guertin et al. 2005) while the TORC1 complex indirectly phosphorylates the insulin pathway at an earlier point in an inhibitory manner (Inoki and Guan 2006), so the two complexes have different feedback effects on insulin

54

signaling. I concentrate here on the better understood effects of the rapamycin-sensitive TORC1 complex. As shown in Figure I.3, Tor in this complex affects protein translation via at least two different routes.

First, Tor stimulates the S6k kinase (Oldham, Montagne et al. 2000) which phosphorylates ribosomal protein RpS6, facilitating translation activity (Meyuhas 2000; Barcelo and Stewart 2002; Lizcano, Alrubaie et al. 2003). As with foxo and Akt, PP2A is associated with S6k in rats and yeast, and direct effects of Tor on PP2A activation can modulate S6k indirectly (Westphal, Coffee Jr et al. 1999; Duvel and Broach 2004; Nien, Dauphinee et al. 2007). However, these PP2A interactions have not been shown in flies so are omitted from Figure I.3.

Second, Tor affects the mRNA cap-binding complex by promoting activity of eIF-4E, which is rate-limiting for translation of many mRNAs (Lachance, Miron et al. 2002). Phosphorylation of eIF-4E is important for activity, and a primary source of this is the Ras/ERK pathway kinase known as LK6 in flies (Lachance, Miron et al. 2002; Hay and Sonenberg 2004; Arquier, Bourouis et al. 2005; Reiling, Doepfner et al. 2005). PP2A can dephosphorylate the LK6 target site, thus inhibiting eIF-4E, so Tor inhibition of PP2A is stimulatory to translation (shown as gray dashed line in Figure I.3) (Duvel and Broach 2004; Sun, Rosenberg et al. 2005; Nien, Dauphinee et al. 2007). Although ERK signaling also acts upstream in the insulin pathway, its interaction with Tor components has been shown to be essential in the protein translation-dependent part of hippocampal long-term potentiation (Tsokas, Ma et al. 2007), a result linking more fundamental anabolic pathways to behaviour and learning.

A second means of regulating eIF-4E activity is via the inhibitory binding protein variously known as 4E-BP or Thor in flies. Here Tor signaling and insulin/foxo signaling converge. Recall that foxo upregulates transcription of Thor when insulin signaling is low, thus inhibiting cap-dependent translation. Meanwhile Tor phosphorylates Thor, dissociating it from eIF-4E and relieving the inhibition (Miron, Lasko et al. 2003). Stress increases Thor inhibition of translation (Teleman, Chen et al. 2005).

Another less understood role of eIF-4E is mRNA export from the nucleus. Some mRNAs contain specific eIF-4E binding sites and are preferentially exported when so bound; these tend to be involved in growth - an example is cyclin D1 (Topisirovic, Ruiz-Gutierrez et al. 2004; Culjkovic, Topisirovic et al. 2006). Conversely, “housekeeping” genes tend not to use eIF-4E export. Hence, a second effect of regulation of eIF-4E may be to regulate export of growth-oriented mRNAs.

Finally, one effect of Tor signaling should be mentioned, although its mechanism is still being clarified. Recent results show that the TORC1 complex can be found associated with mitochondria, and that rapamycin-sensitive Tor activity regulates mitochondrial basal activity (Schieke, Phillips et al. 2006). In yeast, deletion of the Tor1 gene or rapamycin lead to enhanced translation of mitochondrial oxidative

55

phosphorylation components, increased respiration, and increased lifetime (Bonawitz, Chatenay-Lapointe et al. 2007). In human cells, mitochondrial respiration and delta-psi levels were correlated with cell cycle stage and levels of TORC1 complex (Schieke, McCoy et al. 2008), which provides another mechanism to account for known effects of Tor signaling on cell cycle and cell division (Oldham, Bohni et al. 2000; Zhang, Stallock et al. 2000; Hennig and Neufeld 2002; Wu, Cully et al. 2007).

This overview of the insulin and Tor signaling pathways has necessarily left out many additional genes and interactions. To recap, the primary points of importance are that insulin/Tor signaling strongly regulates a number of cellular processes associated with growth and energy stores. Because these pathways have a central role in such regulation, they are themselves regulated by many other inputs, especially those representing signals of cellular energy or homeostatic stress. Regulation is not just post-transcriptional, but also involves changes in transcription of many genes through factors such as foxo. Thus, when we look to understand how GEI with environmental factors such as food supply or stressors work mechanistically, interactions with the insulin/Tor pathway must be a primary point of investigation.

56

List of C. Kent contributions for material in each Chapter Chapter 1. Published in PLoS ONE (Kent, Azanchi et al. 2007). My Role: I conceived the multidimensional analyses, wrote the programs, performed the statistical analyses, identified new concepts describing the chemical patterns of CH variation, and derived models which led to hypotheses about and estimates of CH turnover rates, including a novel application of Random Coefficient Auto-Regressive (RCAR) models, critiqued performance of existing methods of CH normalization and derived new methods, and wrote the paper. RA, BPS, JDL and AC performed experiments designed by JDL, who rewrote the paper.

Chapter 2. Published in Current Biology (Kent, Azanchi et al. 2008). My Role: I conceived the multidimensional analyses, wrote the programs, performed the statistical analyses, identified new concepts describing the patterns of social CH variation, discovered the role of chemical cluster prediction of direction and magnitude of social change, created methods of analysis of IGE for temporal data and derived non-linear corrections for such methods and their power curves, created a hybrid Fourier theory-ANOVA analysis method for detection of temporally structured GEI effects, and co-wrote the paper with JDL. RA, BPS, and JDL performed host visitor experiments and collected GC data, AF performed isolates-communals experiments and collected GC data. JDL designed the experiments and co-wrote the paper.

Chapter 3. Published in Current Biology (Krupp, Kent et al. 2008). My Role: I performed analyses based on models and methods from Chapters 1 and 2 and developed and performed methods for extending qPCR analyses to multiple datasets, and developed methods for analysis of mating data.. JJK, RA, BPS, JDL performed experiments. JDL, JJK, and CK designed experiments. JJK and JDL wrote the paper with input from CK.

Chapter 4. A revised version was accepted by PLoS Genetics My Role: I performed behavior and some metabolite experiments, analysed these and Fourier Transform Mass Spec data, created methods for GEI analysis at pathway level for gene array data, generated hypothesis re insulin involvement, performed crosses, supervised testing of crosses, and wrote the paper. TD and LC performed FTMS analysis, and members of the RJG lab performed microarray hybridizations. Brandon Sheffield and Bianco Marco assisted with fly raising, behaviour experiments, and spectrophotometry assays. I wrote the paper with editing from MBS and RJG. Thanks to Karen Williams for critical reading of the paper prior to submission.

57

References

Alcedo, J. and C. Kenyon (2004). "Regulation of C. elegans longevity by specific gustatory and olfactory neurons." Neuron 41(1): 45-55. Almaas, E., B. Kovacs, T. Vicsek, Z. N. Oltvai and A. L. Barabasi (2004). "Global organization of metabolic fluxes in the bacterium Escherichia coli." Nature 427(6977): 839-843. Andersen, A. S., P. H. Hansen, L. Schaffer and C. Kristensen (2000). "A new secreted insect protein belonging to the immunoglobulin superfamily binds insulin and related peptides and inhibits their activities." J. Biol. Chem. 275(22): 16948-16953. Anholt, R. R., C. L. Dilda, S. Chang, J. J. Fanara, N. H. Kulkarni, I. Ganguly, S. M. Rollmann, K. P. Kamdar and T. F. Mackay (2003). "The genetic architecture of odor-guided behavior in Drosophila: epistasis and the transcriptome." Nat. Genet. 35(2): 180-184. Anholt, R. R. and T. F. Mackay (2004). "Quantitative genetic analyses of complex behaviours in Drosophila." Nat. Rev. Genet. 5(11): 838. Arnold, S. J. (1981). "Behavioral variation in natural populations. 1. Phenotypic, genetic, and environmental correlations between chemoreceptive responses to prey in the garter snake, Thamnophis elegans." Evolution 35(3): 489-509. Arquier, N., M. Bourouis, J. Colombani and P. Leopold (2005). "Drosophila Lk6 kinase controls phosphorylation of eukaryotic translation initiation factor 4E and promotes normal growth and development." Curr. Biol. 15(1): 19-23. Arquier, N., C. Geminard, M. Bourouis, G. Jarretou, B. Honegger, A. Paix and P. Leopold (2008). "Drosophila ALS regulates growth and metabolism through functional interaction with insulin-like peptides." Cell Metab. 7(4): 333-338. Astles, P. A., A. J. Moore and R. F. Preziosi (2005). "Genetic variation in response to an indirect ecological effect." Proc. R. Soc. Lond., Ser. B: Biol. Sci. 272(1581): 2577-81. Bai, X., D. Ma, A. Liu, X. Shen, Q. J. Wang, Y. Liu and Y. Jiang (2007). "Rheb Activates mTOR by antagonizing its endogenous inhibitor, FKBP38." Science 318(5852): 977-980. Baldi, P. and G. W. Hatfield (2002). DNA microarrays and gene expression : from experiments to data analysis and modeling. New York, NY, Cambridge University Press. Barcelo, H. and M. J. Stewart (2002). "Altering Drosophila S6 kinase activity is consistent with a role for S6 kinase in growth." Genesis 34(1-2): 83-85. Barman, S. A., S. Zhu and R. E. White (2004). "PKC activates BKCa channels in rat pulmonary arterial smooth muscle via cGMP-dependent protein kinase." Am. J. Physiol. Lung Cell Mol. Physiol. 286: L1275L1281. Bartholomew, D. J. and M. Knott (1999). Latent variable models and factor analysis. London, Arnold. Barton, N. H. (1990). "Pleiotropic models of quantitative variation." Genetics 124(3): 773-782. Basolo, A. L. (2008). "Evolution of pleiotropic alleles for maturation and size as a consequence of predation." Biol. Lett. 4(2): 200-203.

58

Beavo, J. A., N. L. Rogers, O. B. Crofford, C. E. Baird, J. G. Hardman, E. W. Sutherland and E. V. Newman (1971). "Effects of phosphodiesterase inhibitors on cyclic AMP levels and on lipolysis." Ann. N. Y. Acad. Sci. 185: 129-136. Begg, M., F.-M. Mo, L. Offertaler, S. Batkai, P. Pacher, R. K. Razdan, D. M. Lovinger and G. Kunos (2003). "G protein-coupled endothelial receptor for atypical cannabinoid ligands modulates a Ca2+dependent K+ current." J. Biol. Chem. 278(46): 46188-46194. Begon, M. and B. Shorrocks (1978). "Feeding Sites and Breeding Sites of Drosophila obscura and Drosophila subobscura " J. Nat. Hist. 12(2): 137-151. Belay, A. T., R. Scheiner, A. K. C. So, S. J. Douglas, M. Chakaborty-Chatterjee, J. D. Levine and M. B. Sokolowski (2007). "The foraging gene of Drosophila melanogaster: Spatial-expression analysis and sucrose responsiveness." J. Comp. Neurol. 504(5): 570-582. Bell, W. J. and C. Tortorici (1987). "Genetic and non-genetic control of search duration in adults of two morphs of Drosophila melanogaster." J. Insect Physiol. 33(1): 51-54. Ben-Shahar, Y., H. T. Leung, W. L. Pak, M. B. Sokolowski and G. E. Robinson (2003). "cGMP-dependent changes in phototaxis: a possible role for the foraging gene in honey bee division of labor." J. Exp. Biol. 206(14): 2507-2515. Ben-Shahar, Y., A. Robichon, M. B. Sokolowski and G. E. Robinson (2002). "Influence of gene action across different time scales on behavior." Science 296(5568): 741-744. Berdichevsky, A. and L. Guarente (2006). "A stress response pathway involving sirtuins, forkheads and 143-3 proteins." Cell Cycle 5(22): 2588. Bergmann, S., J. Ihmels and N. Barkai (2004). "Similarities and differences in genome-wide expression data of six organisms." PLoS Biol. 2(1): 85-93. Blows, M. W., S. F. Chenoweth and E. Hine (2004). "Orientation of the genetic variance-covariance matrix and the fitness surface for multiple male sexually selected traits." Am. Nat. 163(3): 329-340. Blows, M. W. and M. B. Sokolowski (1995). "The expression of additive and nonadditive genetic variation under stress." Genetics 140(3): 1149. Bochdanovits, Z. and G. de Jong (2003). "Experimental evolution in Drosophila melanogaster: interaction of temperature and food quality selection regimes." Evolution 57(8): 1829. Bochdanovits, Z. and G. De Jong (2003). "Temperature dependent larval resource allocation shaping adult body size in Drosophila melanogaster." J. Evol. Biol. 16(6): 1159. Bochdanovits, Z. and G. de Jong (2004). "Antagonistic pleiotropy for life-history traits at the gene expression level." Proc. Biol. Sci. 271 Suppl 3: S75-8. Bochdanovits, Z., H. van der Klis and G. de Jong (2003). "Covariation of larval gene expression and adult body size in natural populations of Drosophila melanogaster." Mol. Biol. Evol. 20(11): 1760-1766. Bohni, R., J. Riesgo-Escovar, S. Oldham, W. Brogiolo, H. Stocker, B. F. Andruss, K. Beckingham and E. Hafen (1999). "Autonomous control of cell and organ size by CHICO, a Drosophila homolog of vertebrate IRS1-4." Cell 97(7): 865-875.

59

Bonawitz, N. D., M. Chatenay-Lapointe, Y. Pan and G. S. Shadel (2007). "Reduced TOR signaling extends chronological life span via increased respiration and upregulation of mitochondrial gene expression." Cell Metab. 5(4): 265-77. Bonnet, S., G. Rochefort, G. Sutendra, S. L. Archer, A. Haromy, L. Webster, K. Hashimoto, S. N. Bonnet and E. D. Michelakis (2007). "The nuclear factor of activated T cells in pulmonary arterial hypertension can be therapeutically targeted." Proc. Natl. Acad. Sci. U.S.A. 104(27): 11418. Boorsma, A., X.-J. Lu, A. Zakrzewska, F. M. Klis and H. J. Bussemaker (2008). "Inferring conditionspecific modulation of transcription factor activity in yeast through regulon-based analysis of genomewide expression." PLoS ONE 3(9): e3112. Bordenstein, S. R. and M. D. Drapeau (2001). "Genotype-by-environment interaction and the DobzhanskyMuller model of postzygotic isolation." J. Evol. Biol. 14(3): 490-501. Borniquel, S., I. Valle, S. Cadenas, S. Lamas and M. Monsalve (2006). "Nitric oxide regulates mitochondrial oxidative stress protection via the transcriptional coactivator PGC-1{alpha}." The FASEB Journal 20(11): 1889. Britton, J. S., W. K. Lockwood, L. Li, S. M. Cohen and B. A. Edgar (2002). "Drosophila's insulin/PI3kinase pathway coordinates cellular metabolism with nutritional conditions." Dev. Cell 2(2): 239-249. Broeck, J. V. (2001). "Neuropeptides and their precursors in the fruitfly, Drosophila melanogaster." Peptides 22(2): 241-254. Brogiolo, W., H. Stocker, T. Ikeya, F. Rintelen, R. Fernandez and E. Hafen (2001). "An evolutionarily conserved function of the Drosophila insulin receptor and insulin-like peptides in growth control." Curr. Biol. 11(4): 213-221. Buhl, J., D. J. T. Sumpter, I. D. Couzin, J. J. Hale, E. Despland, E. R. Miller and S. J. Simpson (2006). "From disorder to order in marching locusts." Science 312(5778): 1402-1406. Burger, J. M. S., M. Kolss, J. Pont and T. J. Kawecki (2008). "Learning ability and longevity: A symmetrical evolutionary trade-off in Drosophila." Evolution 62(6): 1294-1304. Bushdid, P. B., H. Osinska, R. R. Waclaw, J. D. Molkentin and K. E. Yutzey (2003). "NFATc3 and NFATc4 are required for cardiac development and mitochondrial function." Circul. Res. 92(12): 1305. Calnan, D. R. and A. Brunet (2008). "The FoxO code." Oncogene 27(16): 2276-2288. Campton, D. E. (1992). "Heritability of body size of green swordtails, Xiphophorus helleri. 1. Sib analyses of males reared individually and in groups." J. Hered. 83(1): 43-48. Campton, D. E. and G. A. E. Gall (1988). "Effect of individual and group rearing on age and size at maturity of male mosquitofish, Gambusia affinis." J. Fish Biol. 33(2): 203-212. Cao, C. and M. R. Brown (2001). "Localization of an insulin-like peptide in brains of two flies." Cell Tissue Res. 304(2): 317-321. Casteel, D. E., S. Zhuang, T. Gudi, J. Tang, M. Vuica, S. Desiderio and R. B. Pilz (2002). "cGMPdependent protein kinase I beta physically and functionally interacts with the transcriptional regulator TFIII." J. Biol. Chem. 277(35): 32003-32014. Cavalli-Sforza, L. L. and W. F. Bodmer (1971). The Genetics of Human Populations. San Francisco, W.H. Freeman.

60

Chan, B., A. Villella, P. Funes and J. C. Hall (2002). "Courtship and other behaviors affected by a heatsensitive, molecularly novel mutation in the cacophony calcium-channel gene of Drosophila." Genetics 162(1): 135-53. Chang, L. W., C. M. Tsai, D. M. Yang and A. S. Chiang (2005). "Cell size control by ovarian factors regulates juvenile hormone synthesis in corpora allata of the cockroach, Diploptera punctata." Insect Biochem. Mol. Biol. 35(1): 41-50. Chapman, T. (2006). "Evolutionary conflicts of interest between males and females." Curr. Biol. 16(17): R744-54. Chapman, T. and L. Partridge (1996). "Female fitness in Drosophila melanogaster: an interaction between the effect of nutrition and of encounter rate with males." Proc. Biol. Sci. 263(1371): 755-9. Chen, C., J. Jack and R. S. Garofalo (1996). "The Drosophila insulin receptor is required for normal growth." Endocrinology 137(3): 846-856. Chen, F., V. Archambault, et al. (2007). "Multiple protein phosphatases are required for mitosis in Drosophila." Curr. Biol. 17(4): 293-303. Chen, L., G. Daum, K. Chitaley, S. A. Coats, D. F. Bowen-Pope, M. Eigenthaler, N. R. Thumati, U. Walter and A. W. Clowes (2004). "Vasodilator-stimulated phosphoprotein regulates proliferation and growth inhibition by nitric oxide in vascular smooth muscle cells." Atertio. Thromb. Vasc. Biol. 24(8): 1403-1408. Chen, Z., E. A. Odstrcil, B. P. Tu and S. L. McKnight (2007). "Restriction of DNA replication to the reductive phase of the metabolic cycle protects genome integrity." Science 316(5833): 1916. Chenoweth, S. F., H. D. Rundle and M. W. Blows (2008). "Genetic constraints and the evolution of display trait sexual dimorphism by natural and sexual selection." Am. Nat. 171(1): 22-34. Chesler, E. J., L. Lu, et al. (2005). "Complex trait analysis of gene expression uncovers polygenic and pleiotropic networks that modulate nervous system function." Nat. Genet. 37(3): 233-242. Chesler, E. J., S. G. Wilson, W. R. Lariviere, S. L. Rodriguez-Zas and J. S. Mogil (2002). "Identification and ranking of genetic and laboratory environment factors influencing a behavioral trait, thermal nociception, via computational analysis of a large data archive." Neurosci. Biobehav. Rev. 26(8): 907-23. Cheverud, J. M. (1984). "Evolution by kin selection - a quantitative genetic model illustrated by maternal performance in mice." Evolution 38(4): 766-777. Cobb, M. and J.-M. Jallon (1990). "Pheromones, mate recognition and courtship stimulation in the Drosophila melanogaster species sub-group." Anim Behav 39(6): 1058. Cohen, P., D. R. Alessi and D. A. Cross (1997). "PDK1, one of the missing links in insulin signal transduction?" FEBS Lett. 410(1): 3-10. Colasurdo, N. and E. Despland (2005). "Social cues and following behavior in the forest tent caterpillar." J. Insect Behav. 18(1): 77-87. Consortium, G. O. (2000). "Gene Ontology: tool for the unification of biology." Nat. Genet. 25: 25-29. Consortium, G. O. (2001). "Creating the Gene Ontology resource: design and implementation." Genome Res. 11: 1425-1433. Corbin, J. D. and S. H. Francis (1999). "Cyclic GMP phosphodiesterase-5: target of Sildenafil." J. Biol. Chem. 274(20): 13729-13732.

61

Corby-Harris, V. and D. E. Promislow (2008). "Host ecology shapes geographical variation for resistance to bacterial infection in Drosophila melanogaster." J. Anim. Ecol. 77(4): 768. Costa, J. T. (1997). "Caterpillars as social insects." Am. Sci. 85(2): 150-159. Couzin, I. D., J. Krause, R. James, G. D. Ruxton and N. R. Franks (2002). "Collective memory and spatial sorting in animal groups." J. Theor. Biol. 218(1): 1-11. Crabbe, J. C. and D. Wahlsten (2003). "Of mice and their environments." Science 299(5611): 1313. Cross, D. A., D. R. Alessi, P. Cohen, M. Andjelkovich and B. A. Hemmings (1995). "Inhibition of glycogen synthase kinase-3 by insulin mediated by protein kinase B." Nature 378(6559): 785-789. Culjkovic, B., I. Topisirovic, L. Skrabanek, M. Ruiz-Gutierrez and K. L. B. Borden (2006). "eIF4E is a central node of an RNA regulon that governs cellular proliferation." J. Cell Biol. 175(3): 415-426. Cully, M., H. You, A. J. Levine and T. W. Mak (2006). "Beyond PTEN mutations: the PI3K pathway as an integrator of multiple inputs during tumorigenesis." Nat. Rev. Cancer 6(3): 184-192. Cuong, D. V., N. Kim, et al. (2006). "Nitric oxide-cGMP-protein kinase G signaling pathway induces anoxic preconditioning through activation of ATP-sensitive K+ channels in rat hearts." Am. J. Physiol. Heart Circ. Physiol. 290(5): H1808. Curtsinger, J. W. and A. A. Khazaeli (2002). "Lifespan, QTLs, age-specificity, and pleiotropy in Drosophila." Mech. Ageing Dev. 123(2-3): 81-93. Davies, S. A. (2000). "Nitric oxide signalling in insects." Insect Biochem. Mol. Biol. 30(12): 1123-1138. Dawson-Scully, K., G. A. Armstrong, C. Kent, R. M. Robertson and M. B. Sokolowski (2007). "Natural variation in the thermotolerance of neural function and behavior due to a cGMP-dependent protein kinase." PLoS One 2(8): e773. Dawson-Scully, K., C. Kent, G. A. B. Armstrong, M. B. Sokolowski and R. M. Robertson (2005). Rapid neural thermoprotection using PKG inhibition. Cold Spring Harbor Neurobiology of Drosophila, Cold Spring Harbor, NY, Cold Spring Harbor Press. de Belle, J. S., A. J. Hilliker and M. B. Sokolowski (1989). "Genetic localization of foraging (for): a major gene for larval behavior in Drosophila melanogaster." Genetics 123(1): 157-164. de Belle, J. S., M. B. Sokolowski and A. J. Hilliker (1993). "Genetic analysis of the foraging microregion of Drosophila melanogaster." Genome 36(1): 94-101. De Meester, L. (1996). "Evolutionary potential and local genetic differentiation in a phenotypically plastic trait of a cyclical parthenogen, Daphnia magna." Evolution 50: 1293-1298. DeAngelis, D. L. and W. M. Mooij (2005). "Individual-based modeling of ecological and evolutionary processes." Annu. Rev. Ecol. Syst. 36(1): 147-168. Demuth, J. P. and M. J. Wade (2007). "Population differentiation in the beetle Tribolium castaneum. II. Haldane'S rule and incipient speciation." Evolution 61(3): 694. Demyanenko, G. P., A. I. Halberstadt, K. B. Pryzwansky, C. Werner, F. Hofmann and P. F. Maness (2005). "Abnormal neocortical development in mice lacking cGMP-dependent protein kinase I." Dev. Brain Res. 160(1): 1-8.

62

Despland, E. and S. Hamzeh (2004). "Ontogenetic changes in social behaviour in the forest tent caterpillar, Malacosoma disstria." Behav. Ecol. Sociobiol. 56(2): 177-184. Despland, E. and A. Le Huu (2007). "Pros and cons of group living in the forest tent caterpillar: separating the roles of silk and of grouping." Entomol. Exp. Appl. 122(2): 181-189. Despland, E. and S. J. Simpson (2006). "Resource distribution mediates synchronization of physiological rhythms in locust groups." Proc. R. Soc. Lond., B, Biol. Sci. 273(1593): 1517-1522. Douglas, S. J., K. Dawson-Scully and M. B. Sokolowski (2005). "The neurogenetics and evolution of foodrelated behaviour." Trends Neurosci. 28(12): 644-652. Dover, G. J., S. Brusilow and S. Charache (1994). "Induction of fetal hemoglobin production in subjects with sickle cell anemia by oral sodium phenylbutyrate." Blood 84(1): 339-43. Dussutour, A., S. C. Nicolis, E. Despland and S. J. Simpson (2008). "Individual differences influence collective behaviour in social caterpillars." Anim. Behav. 76: 5-16. Dussutour, A., S. J. Simpson, E. Despland and N. Colasurdo (2007). "When the group denies individual nutritional wisdom." Anim. Behav. 74: 931-939. Duvel, K. and J. R. Broach (2004). "The role of phosphatases in TOR signaling in yeast." Curr. Top. Microbiol. Immunol. 279: 19-38. Edwards, J. W., G. P. Page, G. Gadbury, M. Heo, T. Kayo, R. Weindruch and D. B. Allison (2005). "Empirical Bayes estimation of gene-specific effects in micro-array research." Funct.Integr.Genomics 5(1): 32-39. Efron, B. and R. Tibshirani (2002). "Empirical Bayes methods and false discovery rates for microarrays." Genet. Epidemiol. 23(1): 70-86. Ehrman, L. (1989). "Effects of lifelong experience on the rare-male mating advantage in Drosophila pseudoobscura." Behav. Genet. 19(6): 755-755. Ehrman, L., A. Wissner and J. Meinwald (1973). "Pheromone mediating rare male mating advantage in Drosophila pseudoobscura." Genetics 74(JUN): S69-S69. Engel, J. E. and C. F. Wu (1992). "Interactions of membrane excitability mutations affecting potassium and sodium currents in the flight and giant fiber escape systems of Drosophila." J. Comp. Physiol., A 171(1): 93-104. Engel, J. E. and C. F. Wu (1996). "Altered habituation of an identified escape circuit in Drosophila memory mutants." J. Neurosci. 16(10): 3486-3499. Engel, J. E. and C. F. Wu (1998). "Genetic dissection of functional contributions of specific potassium channel subunits in habituation of an escape circuit in Drosophila." J. Neurosci. 18(6): 2254-2267. Engel, J. E., X. J. Xie, M. B. Sokolowski and C. F. Wu (2000). "A cGMP-dependent protein kinase gene, foraging, modifies habituation-like response decrement of the giant fiber escape circuit in Drosophila." Learn. Memory 7(5): 341-352. Ennis, H. N., D. N. Dao, M. Y. Wu and R. H. Kessin (2003). "Mutation of the Dictyostelium fbxA gene affects cell-fate decisions and spatial patterning." Protist 154: 419-429.

63

Etges, W. J., C. C. de Oliveira, E. Gragg, D. Ortiz-Barrientos, M. A. Noor and M. G. Ritchie (2007). "Genetics of incipient speciation in Drosophila mojavensis. I. Male courtship song, mating success, and genotype x environment interactions." Evolution 61(5): 1106. Ewer, J., J. Devente and J. W. Truman (1994). "Neuropeptide induction of cyclic-GMP increases in the insect CNS - resolution at the level of single identifiable neurons." J. Neurosci. 14(12): 7704-7712. Ewer, J. and S. Reynold (2002). Neuropeptide control of molting in insects. Hormones, Brain and Behavior. D. W. Pfaff, A. P. Arnold, A. M. Etgen, S. E. Fahrbach and R. T. Rubin. San Diego, Academic Press. 3: 1-92. Ewer, J. and J. W. Truman (1996). "Increases in cyclic 3',5'-guanosine monophosphate (cGMP) occur at ecdysis in an evolutionarily conserved crustacean cardioactive peptide-immunoreactive insect neuronal network." J. Comp. Neurol. 370(3): 330-341. Feil, R., F. Hofmann and T. Kleppisch (2005). "Function of cGMP-dependent protein kinases in the nervous system." Rev. Neurosci. 16(1): 23-41. Ferrero, R., F. Rodriguez-Pascual, M. Miras-Portugal and M. Torres (2000). "Nitric oxide-sensitive guanylyl cyclase activity inhibition through cyclic GMP-dependent dephosphorylation." J. Neurochem. 75(5): 2029-2039. Ferveur, J. F. (2005). "Cuticular hydrocarbons: their evolution and roles in Drosophila pheromonal communication." Behav. Genet. 35(3): 279-295. Fiedler, B., S. M. Lohmann, A. Smolenski, S. Linnemuller, B. Pieske, F. Schroder, J. D. Molkentin, H. Drexler and K. C. Wollert (2002). "Inhibition of calcineurin-NFAT hypertrophy signaling by cGMPdependent protein kinase type I in cardiac myocytes." Proc. Natl. Acad. Sci. U.S.A. 99(17): 11363-11368. Fisher, R. A. (1958). The genetical theory of natural selection. N.Y., Dover. Fitzpatrick, M. J., E. Feder, L. Rowe and M. B. Sokolowski (2007). "Maintaining a behaviour polymorphism by frequency-dependent selection on a single gene." Nature 447(7141): 210-212. Fitzpatrick, M. J. and M. B. Sokolowski (2004). "In search of food: Exploring the evolutionary link between cGMP-dependent protein kinase (PKG) and behaviour." Integr. Comp. Biol. 44(1): 28-36. Foley, B., S. F. Chenoweth, S. V. Nuzhdin and M. W. Blows (2007). "Natural genetic variation in cuticular hydrocarbon expression in male and female Drosophila melanogaster." Genetics 175(3): 1465-1477. Frank, S. A. (2007). "All of life is social." Curr. Biol. 17(16): R648. Freeman, S. and J. C. Herron (1998). Evolutionary Analysis. Upper Saddle River, N.J., Prentice Hall. Friedman, A. and N. Perrimon (2007). "Genetic screening for signal transduction in the era of network biology." Cell 128(2): 225-31. Fujiwara, M., P. Sengupta and S. L. McIntire (2002). "Regulation of body size and behavioral state of C. elegans by sensory perception and the EGL-4 cGMP-dependent protein kinase." Neuron 36(6): 1091-1102. Gao, X., T. P. Neufeld and D. Pan (2000). "Drosophila PTEN Regulates Cell Growth and Proliferation through PI3K-Dependent and -Independent Pathways." Dev. Biol. 221(2): 404-418. Gao, X., Y. Zhang, P. Arrazola, O. Hino, T. Kobayashi, R. S. Yeung, B. Ru and D. Pan (2002). "Tsc tumour suppressor proteins antagonize amino-acid-TOR signalling." Nat. Cell Biol. 4(9): 699-704.

64

Gardner, K. M. and R. G. Latta (2007). "Shared quantitative trait loci underlying the genetic correlation between continuous traits." Mol. Ecol. 16(20): 4195-4209. Gershman, B., O. Puig, L. Hang, R. M. Peitzsch, M. Tatar and R. S. Garofalo (2007). "High-resolution dynamics of the transcriptional response to nutrition in Drosophila: a key role for dFOXO." Physiol. Genomics 29(1): 24-34. Gilbert, D. G. (1980). "Dispersal of yeasts and bacteria by Drosophila in a temperate forest." Oecologia 46(1): 135-137. Gilbert, O. M., K. R. Foster, N. J. Mehdiabadi, J. E. Strassmann and D. C. Queller (2007). "High relatedness maintains multicellular cooperation in a social amoeba by controlling cheater mutants." Proc. Natl. Acad. Sci. U.S.A. 104: 8913–8917. Gillespie, J. H. and M. Turelli (1989). "Genotype-Environment Interactions and the maintenance of polygenic variation." Genetics 121(1): 129-138. Given, A. M., O. Ogut and F. V. Brozovich (2007). "MYPT1 mutants demonstrate the importance of aa 888-928 for the interaction with PKGI{alpha}." Am. J. Physiol. Cell. Physiol. 292(1): C432. Glass, L. and M. C. Mackey (1988). From clocks to chaos. Princeton, NJ, Princeton University Press. Goberdhan, D. C., N. Paricio, E. C. Goodman, M. Mlodzik and C. Wilson (1999). "Drosophila tumor suppressor PTEN controls cell size and number by antagonizing the Chico/PI3-kinase signaling pathway." Genes Dev. 13(24): 3244. Goberdhan, D. C. and C. Wilson (2003). "PTEN: tumour suppressor, multifunctional growth regulator and more." Hum. Mol. Genet. 12 Spec No 2: R239. Gopal, V. K., S. H. Francis and J. D. Corbin (2001). "Allosteric sites of phosphodiesterase-5 (PDE5): A potential role in negative feedback regulation of cGMP signaling in corpus cavernosum." Eur. J. Biochem. 268(11): 3304-3312. Graf, S. A. and M. B. Sokolowski (1989). "The rover/sitter Drosophila foraging polymorphism as a function of larval development, food patch quality and starvation." J. Insect Behav. 2(3): 301-313. Gragasin, F. S., E. D. Michelakis, A. Hogan, R. Moudgil, K. Hashimoto, X. Wu, S. Bonnet, A. Haromy and S. L. Archer (2004). "The neurovascular mechanism of clitoral erection: nitric oxide and cGMP-stimulated activation of BKCa channels." FASEB J. 18(12): 1382-1391. Green, D. M., A. R. Kraaijeveld and H. C. Godfray (2000). "Evolutionary interactions between Drosophila melanogaster and its parasitoid Asobara tabida." Heredity 85 Pt 5: 450-458. Greenspan, R. J. and J. F. Ferveur (2000). "Courtship in Drosophila." Annu. Rev. Genet. 34: 205-232. Gudi, T., D. E. Casteel, C. Vinson, G. R. Boss and R. B. Pilz (2000). "NO activation of fos promoter elements requires nuclear translocation of G-kinase I and CREB phosphorylation but is independent of MAP kinase activation." Oncogene 19(54): 6324-6333. Gudi, T., G. K. Hong, A. B. Vaandrager, S. M. Lohmann and R. B. Pilz (1999). "Nitric oxide and cGMP regulate gene expression in neuronal and glial cells by activating type II cGMP-dependent protein kinase." FASEB J. 13(15): 2143-2152. Gudi, T., S. M. Lohmann and R. B. Pilz (1997). "Regulation of gene expression by cyclic GMP-dependent protein kinase requires nuclear translocation of the kinase: identification of a nuclear localization signal." Mol. Cell. Biol. 17(9): 5244-5254.

65

Gurganus, M. C., S. V. Nuzhdin, J. W. Leips and T. F. Mackay (1999). "High-resolution mapping of quantitative trait loci for sternopleural bristle number in Drosophila melanogaster." Genetics 152(4): 1585. Haase, A. and G. Bicker (2003). "Nitric oxide and cyclic nucleotides are regulators of neuronal migration in an insect embryo." Development 130(17): 3977-3987. Hall, J. C. (2002). "Courtship lite: a personal history of reproductive behavioral neurogenetics in Drosophila." J. Neurogenet. 16(3): 135-63. Hallem, E. A. and P. W. Sternberg (2008). "Acute carbon dioxide avoidance in Caenorhabditis elegans." Proc. Natl. Acad. Sci. U.S.A. 105(23): 8038-8043. Halloy, J., G. Sempo, et al. (2007). "Social integration of robots into groups of cockroaches to control selforganized choices." Science 318(5853): 1155-8. Hamilton, W. D. (1964). "The genetical evolution of social behavior." J. Theor. Biol. 7: 1-16. Hardin, P. E., B. Krishnan, J. H. Houl, H. Zheng, F. S. Ng, S. E. Dryer and N. R. Glossop (2003). "Central and peripheral circadian oscillators in Drosophila." Novartis Foundation symposium 253: 140. Hashimoto, K., A. C. Yoshizawa, S. Okuda, K. Kuma, S. Goto and M. Kanehisa (2008). "The repertoire of desaturases and elongases reveals fatty acid variations in 56 eukaryotic genomes." J. Lipid Res. 49(1): 18391. Hay, N. and N. Sonenberg (2004). "Upstream and downstream of mTOR." Genes Dev. 18(16): 1926-1945. Hennig, K. M. and T. P. Neufeld (2002). "Inhibition of cellular growth and proliferation by dTOR overexpression in Drosophila." Genesis 34(1-2): 107-110. Herrgård, M. J., M. W. Covert and B. O. Palsson (2003). "Reconciling gene expression data with known genome-scale regulatory network structures." Genome Res. 13: 2423-2434. Hietakangas, V. and S. M. Cohen (2007). "Re-evaluating AKT regulation: role of TOR complex 2 in tissue growth." Genes Dev. 21(6): 632-637. Hine, E., S. F. Chenoweth and M. W. Blows (2004). "Multivariate quantitative genetics and the lek paradox: Genetic variance in male sexually selected traits of Drosophila serrata under field conditions." Evolution 58(12): 2754. Hirose, T., Y. Nakano, Y. Nagamatsu, T. Misumi, H. Ohta and Y. Ohshima (2003). "Cyclic GMPdependent protein kinase EGL-4 controls body size and lifespan in C. elegans." Development 130(6): 1089-1099. Hodge, S., P. Mitchell and W. Arthur (1999). "Factors affecting the occurrence of facilitative effects in interspecific interactions: an experiment using two species of Drosophila and Aspergillus niger." Oikos 87: 166. Hofmann, F., R. Feil, T. Kleppisch and J. Schlossmann (2006). "Function of cGMP-dependent protein kinases as revealed by gene deletion." Physiol. Rev. 86(1): 1-23. Holbrook, G. L., E. Armstrong, J. A. S. Bachmann, B. M. Deasy and C. Schal (2000). "Role of feeding in the reproductive 'group effect' in females of the German cockroach Blattella germanica (L.)." J. Insect Physiol. 46(6): 941-949.

66

Holbrook, G. L. and C. Schal (2004). "Maternal investment affects offspring phenotypic plasticity in a viviparous cockroach." Proc. Natl. Acad. Sci. U.S.A. 101(15): 5595-5597. Howard, R. W. and G. J. Blomquist (2005). "Ecological, behavioral, and biochemical aspects of insect hydrocarbons." Annu. Rev. Entomol. 50: 371-393. Hsu, J.-Y., T. Juven-Gershon, M. T. Marr, II, K. J. Wright, R. Tjian and J. T. Kadonaga (2008). "TBP, Mot1, and NC2 establish a regulatory circuit that controls DPE-dependent versus TATA-dependent transcription." Genes Dev. 22(17): 2353-2358. Huang, C. H., D. Mandelker, S. B. Gabelli and L. M. Amzel (2008). "Insights into the oncogenic effects of PIK3CA mutations from the structure of p110 alpha/p85 alpha." Cell Cycle 7(9): 1151-1156. Huang, Q. Q., S. A. Fisher and F. V. Brozovich (2004). "Unzipping the role of myosin light chain phosphatase in smooth muscle cell relaxation." J. Biol. Chem. 279(1): 597-603. Hughes, K. and M. B. Sokolowski (1996). "Natural selection in the laboratory for a change in resistance by Drosophila melanogaster to the parasitoid wasp Asobara tabida." J. Insect Behav. 9(3): 477-491. Hughes, K. A. and R. M. Reynolds (2005). "Evolutionary and mechanistic theories of aging." Annu. Rev. Entomol. 50: 421-445. Ikeya, T., M. Galic, P. Belawat, K. Nairz and E. Hafen (2002). "Nutrient-dependent expression of insulinlike peptides from neuroendocrine cells in the CNS contributes to growth regulation in Drosophila." Curr. Biol. 12(15): 1293-1300. Imren, S., M. E. Fabry, et al. (2004). "High-level beta-globin expression and preferred intragenic integration after lentiviral transduction of human cord blood stem cells." J. Clin. Invest. 114(7): 953-62. Imren, S., E. Payen, et al. (2002). "Permanent and panerythroid correction of murine beta thalassemia by multiple lentiviral integration in hematopoietic stem cells." Proc. Natl. Acad. Sci. U.S.A. 99(22): 14380-5. Ingram, K. K., P. Oefner and D. M. Gordon (2005). "Task-specific expression of the foraging gene in harvester ants." Mol. Ecol. 14(3): 813-818. Inoki, K. and K. L. Guan (2006). "Complexity of the TOR signaling network." Trends Cell Biol. 16(4): 206-212. Inoki, K., H. Ouyang, et al. (2006). "TSC2 Integrates Wnt and energy signals via a coordinated phosphorylation by AMPK and GSK3 to regulate cell growth." Cell 126(5): 955-968. Inoki, K., T. Q. Zhu and K. L. Guan (2003). "TSC2 mediates cellular energy response to control cell growth and survival." Cell 115(5): 577-590. Jallon, J., G. Kunesch, L. Bricard and M. Pennanec'h (1997). "Incorporation of fatty acids into cuticular hydrocarbons of male and female Drosophila melanogaster." J. Insect Physiol. 43(12): 1111. Juarez, M. P., R. Crespo, G. C. Fernandez, R. Lecuona and L. F. Cafferata (2000). "Characterization and carbon metabolism in fungi pathogenic to Triatoma infestans, a Chagas disease vector." J. Invertebr. Pathol. 76(3): 198-207. Junger, M. A., F. Rintelen, H. Stocker, J. D. Wasserman, M. Vegh, T. Radimerski, M. E. Greenberg and E. Hafen (2003). "The Drosophila forkhead transcription factor FOXO mediates the reduction in cell number associated with reduced insulin signaling." J. Biol. 2(3): 2-17.

67

Kalderon, D. and G. M. Rubin (1989). "cGMP-dependent protein kinase genes in Drosophila." J. Biol. Chem. 264(18): 10738-10748. Kanehisa, M., M. Araki, et al. (2008). "KEGG for linking genomes to life and the environment." Nucleic Acids Res. 36(D): D480-4. Kang, S. H., W. S. Park, N. Kim, J. B. Youm, M. Warda, J.-H. Ko, E. A. Ko and J. Han (2007). "Mitochondrial Ca2+-activated K+ channels more efficiently reduce mitochondrial Ca2+ overload in rat ventricular myocytes." Am. J. Physiol. Heart Circ. Physiol. 293(1): H307-313. Kaun, K. R., T. Hendel, B. Gerber and M. B. Sokolowski (2007). "Natural variation in Drosophila larval reward learning and memory due to a cGMP-dependent protein kinase." Learn. Memory 14(5): 342-349. Kaun, K. R., C. A. L. Riedl, M. Chakaborty-Chatterjee, A. T. Belay, S. J. Douglas, A. G. Gibbs and M. B. Sokolowski (2007). "Natural variation in food acquisition mediated via a Drosophila cGMP-dependent protein kinase." J. Exp. Biol. 210(20): 3547-3558. Kaun, K. R., C. A. L. Riedl, M. Chakaborty-Chatterjee, A. T. Belay, S. J. Douglas, A. G. Gibbs and M. B. Sokolowski (2008). "Natural variation in plasticity of glucose homeostasis and food intake." J. Exp. Biol. 211(19): 3160-3166 Kaun, K. R. and M. B. Sokolowski (2009). "cGMP-dependent protein kinase: linking foraging to energy homeostasis." Genome 52(1): 1-7. Keller, M. C. and G. Miller (2006). "Resolving the paradox of common, harmful, heritable mental disorders: Which evolutionary genetic models work best?" Behav. Brain Sci. 29(4): 385-+. Kent, C., R. Azanchi, B. Smith, A. Chu and J. Levine (2007). "A model-based analysis of chemical and temporal patterns of cuticular hydrocarbons in male Drosophila melanogaster." PLoS One 2(9): e962. Kent, C., R. Azanchi, B. Smith, A. Formosa and J. D. Levine (2008). "Social context influences chemical communication in D. melanogaster males." Curr. Biol. 18(18): 1384-9. Kim, Y. K., L. Ehrman and H. R. Koepfer (1992). "Developmental isolation and subsequent adult behavior of Drosophila paulistorum. I. Survey of the six semispecies." Behav. Genet. 22(5): 545-556. Kim, Y. K., L. Ehrman and H. R. Koepfer (1996). "Developmental isolation and subsequent adult behavior of Drosophila paulistorum. II. prior experience." Behav. Genet. 26(1): 15-25. Kim, Y. K., H. R. Koepfer and L. Ehrman (1996). "Developmental isolation and subsequent adult behavior of Drosophila paulistorum. III. Alternative rearing." Behav. Genet. 26(1): 27-37. Kim, Y. K., D. R. Phillips, T. Chao and L. Ehrman (2004). "Developmental isolation and subsequent adult behavior of Drosophila paulistorum. VI. Quantitative variation in cuticular hydrocarbons." Behav. Genet. 34(4): 385-394. Kingan, T. G., R. A. Cardullo and M. E. Adams (2001). "Signal transduction in eclosion hormone-induced secretion of ecdysis-triggering hormone." J. Biol. Chem. 276(27): 25136-25142. Klein, G. (2000). "Protein kinase G reverses all isoproterenol induced changes of cardiac single L-type calcium channel gating." Cardiovasc. Res. 48(3): 367-374. Koh, K., X. Zheng and A. Sehgal (2006). "JETLAG resets the Drosophila circadian clock by promoting light-induced degradation of TIMELESS." Science 312(5781): 1809.

68

Krishnan, B., S. E. Dryer and P. E. Hardin (1999). "Circadian rhythms in olfactory responses of Drosophila melanogaster." Nature 400(6742): 375. Krishnan, P., S. E. Dryer and P. E. Hardin (2005). "Measuring circadian rhythms in olfaction using electroantennograms." Methods Enzymol. 393: 495. Krupp, J. J., C. Kent, J. C. Billeter, R. Azanchi, A. K. So, J. A. Schonfeld, B. P. Smith, C. Lucas and J. D. Levine (2008). "Social experience modifies pheromone expression and mating behavior in male Drosophila melanogaster." Curr. Biol. 18(18): 1373-83. Kyriacou, C. P., A. A. Peixoto, F. Sandrelli, R. Costa and E. Tauber (2008). "Clines in clock genes: finetuning circadian rhythms to the environment." Trends Genet. 24(3): 124-132. L'Etoile, N. D., C. M. Coburn, J. Eastham, A. Kistler, G. Gallegos and C. I. Bargmann (2002). "The cyclic GMP-dependent protein kinase EGL-4 regulates olfactory adaptation in C. elegans." Neuron 36(6): 10791089. Lachance, P. E., M. Miron, B. Raught, N. Sonenberg and P. Lasko (2002). "Phosphorylation of eukaryotic translation initiation factor 4E is critical for growth." Mol. Cell. Biol. 22(6): 1656-1663. Lafontan, M. (2005). "An unsuspected metabolic role for atrial natriuretic peptides - The control of lipolysis, lipid mobilization, and systemic nonesterified fatty acids levels in humans." Arteriosclerosis Thrombosis and Vascular Biology 25(10): 2032-2042. Lafontan, M. (2008). "Control of lipolysis by natriuretic peptides and cyclic GMP." Trends Endocrinol. Metab. 19(4): 130-137. Lambeth, J. D. (2007). "Nox enzymes, ROS, and chronic disease: An example of antagonistic pleiotropy." Free Radical Biol. Med. 43(3): 332-347. Lande, R. (1979). "Quantitative genetic analysis of multivariate evolution applied to brain: Body size allometry." Evolution 33: 402-416. Lande, R. and S. J. Arnold (1983). "The measurement of selection on correlated characters." Evolution 37: 1210-1226. Lawley, D. N., A. E. Maxwell and A. E. Maxwell (1971). Factor analysis as a statistical method: by D.N. Lawley and A.E. Maxwell. New York, American Elsevier Pub. Co. Lawrence, D. W. and K. B. Pryzwansky (2001). "The vasodilator-stimulated phosphoprotein is regulated by cyclic GMP-dependent protein kinase during neutrophil spreading." J. Immunol. 166(9): 5550-5556. Lazzaro, B. P., B. K. Sceurman and A. G. Clark (2004). "Genetic Basis of Natural Variation in D. melanogaster Antibacterial Immunity." Science 303(5665): 1873-1876. Lecuona, R., J. L. Clement, G. Riba, C. Joulie and M. P. Juarez (1997). "Spore germination and hyphal growth of Beauveria sp. on insect lipids." J. Econ. Entomol. 89: 119-123. Leevers, S. J., D. Weinkove, L. K. MacDougall, E. Hafen and M. D. Waterfield (1996). "The Drosophila phosphoinositide 3-kinase Dp110 promotes cell growth." EMBO J. 15(23): 6584-94. Leips, J., P. Gilligan and T. R. C. Mackay (2006). "Quantitative trait loci with age-specific effects on fecundity in Drosophila melanogaster." Genetics 172(3): 1595-1605. Levene, H. (1953). "Genetic equilibrium when more than one ecological niche is available." Am. Nat. 87: 311-313.

69

Levine, J. D., P. Funes, H. B. Dowse and J. C. Hall (2002). "Resetting the circadian clock by social experience in Drosophila melanogaster." Science 298(5600): 2010-2012. Levine, J. D., P. Funes, H. B. Dowse and J. C. Hall (2002). "Signal analysis of behavioral and molecular cycles." BMC Neurosci. 3(1): 1-25. Lewis, M. A. and H. G. Othmer (1997). Qualitative theory of ordinary differential equations. Mathematical modeling: Ecology, Physiology, and Cell Biology. H. G. Othmer, F. R. Adler, M. A. Lewis and J. C. Dallon. Upper Saddle River, NJ, Prentice Hall: 357-380. Lewontin, R. C. (1955). "The effects of population density and composition on viability in Drosophila melanogaster." Evolution 9: 27-41. Lewontin, R. C. and Y. Matsuo (1963). "Interaction of genotypes determining viability in Drosophila busckii." Proc. Natl. Acad. Sci. U.S.A. 49(2): 270-278. Li, S. J. and N. L. Sun (2005). "Regulation of intracellular Ca2+ and calcineurin by NO/PKG in proliferation of vascular smooth muscle cells." Acta Pharmacol. Sin. 26(3): 323-328. Lincoln, T. M., N. Dey and H. Sellak (2001). "Invited review: cGMP-dependent protein kinase signaling mechanisms in smooth muscle: from the regulation of tone to gene expression." J. Appl. Physiol. 91(3): 1421-1430. Lizcano, J. M., S. Alrubaie, A. Kieloch, M. Deak, S. J. Leevers and D. R. Alessi (2003). "Insulin-induced Drosophila S6 kinase activation requires phosphoinositide 3-kinase and protein kinase B." Biochem. J. 374(Pt 2): 297-306. Lloyd, D. and D. B. Murray (2006). "The temporal architecture of eukaryotic growth." FEBS Lett. 580(12): 2830. Luan, J., P. O. Browne, A. H. Harding, D. J. Halsall, S. O'Rahilly, V. K. Chatterjee and N. J. Wareham (2001). "Evidence for gene-nutrient interaction at the PPARgamma locus." Diabetes 50(3): 686. Lynch, M. and B. Walsh (1998). Genetics and analysis of quantitative traits. Sunderland, Ma., Sinauer. MacPherson, M. R., S. M. Lohmann and S. A. Davies (2004). "Analysis of Drosophila cGMP-dependent protein kinases and assessment of their in vivo roles by targeted expression in a renal transporting epithelium." J. Biol. Chem. 279(38): 40026-40034. Marcillac, F., F. Bousquet, J. Alabouvette, F. Savarit and J. F. Ferveur (2005). "A mutation with major effects on Drosophila melanogaster sex pheromones." Genetics 171(4): 1617-1628. Marcillac, F., B. Houot and J.-F. Ferveur (2005). "Revisited Roles of Drosophila Female Pheromones." Chem. Senses 30: i273. Marsh, D. J., P. L. Dahia, et al. (1998). "Allelic imbalance, including deletion of PTEN/MMACI, at the Cowden disease locus on 10q22-23, in hamartomas from patients with Cowden syndrome and germline PTEN mutation." Genes, chromosomes & cancer 21(1): 61-69. Martinez, J. A., M. S. Corbalan, A. Sanchez-Villegas, L. Forga, A. Marti and M. A. Martinez-Gonzalez (2003). "Obesity risk is associated with carbohydrate intake in women carrying the Gln27Glu beta2adrenoceptor polymorphism." J. Nutr. 133(8): 2549. Matsumoto, Y., S. Unoki, H. Aonuma and M. Mizunami (2006). "Critical role of nitric oxide-cGMP cascade in the formation of cAMP-dependent long-term memory." Learn. Memory 13(1): 35-44.

70

Memisoglu, A., F. B. Hu, S. E. Hankinson, J. E. Manson, I. De Vivo, W. C. Willett and D. J. Hunter (2003). "Interaction between a peroxisome proliferator-activated receptor gamma gene polymorphism and dietary fat intake in relation to body mass." Hum. Mol. Genet. 12(22): 2923. Mery, F., A. T. Belay, A. K. So, M. B. Sokolowski and T. J. Kawecki (2007). "Natural polymorphism affecting learning and memory in Drosophila." Proc. Natl. Acad. Sci. USA 104(32): 13051-13055. Mery, F., A. T. Belay, M. B. Sokolowski and T. J. Kawecki (2006). "Antagonistic effects of natural polymorphism at the foraging locus on short-term learning and long-term memory." J. Neurogenet. 20(34): 177-177. Meyuhas, O. (2000). "Synthesis of the translational apparatus is regulated at the translational level." Eur. J. Biochem. 267(21): 6321-6330. Min, K. J., R. Yamamoto, S. Buch, M. Pankratz and M. Tatar (2008). "Drosophila lifespan control by dietary restriction independent of insulin-like signaling." Aging Cell 7(2): 199-206. Miron, M., P. Lasko and N. Sonenberg (2003). "Signaling from Akt to FRAP/TOR targets both 4E-BP and S6K in Drosophila melanogaster." Mol. Cell. Biol. 23(24): 9117-26. Moore, A. J., E. D. Brodie, III and J. B. Wolf (1997). "Interacting phenotypes and the evolutionary process: I. Direct and indirect genetic effects of social interactions." Evolution 51(5): 1352(11). Moore, A. J. and T. Pizzari (2005). "Quantitative genetic models of sexual conflict based on interacting phenotypes." Am. Nat. 165 Suppl 5: S88-97. Moran, M. D. (2003). "Arguments for rejecting the sequential Bonferroni in ecological studies." Oikos 100(2): 403-405. Morre, D. J., P. J. Chueh, J. Pletcher, X. Tang, L. Y. Wu and D. M. Morre (2002). "Biochemical basis for the biological clock." Biochemistry 41(40): 11941. Morre, D. J. and D. M. Morre (2003). "Cell surface NADH oxidases (ECTO-NOX proteins) with roles in cancer, cellular time-keeping, growth, aging and neurodegenerative diseases." Free Radic. Res. 37(8): 795. Morre, D. J. and D. M. Morre (2003). "The plasma membrane-associated NADH oxidase (ECTO-NOX) of mouse skin responds to blue light." J. Photochem. Photobiol. B: Biol. 70(1): 7. Moya, A., J. Pereto, R. Gil and A. Latorre (2008). "Learning how to live together: genomic insights into prokaryote-animal symbioses." Nat. Rev. Genet. 9(3): 218-29. Murray, D. B. and D. Lloyd (2006). "A tuneable attractor underlies yeast respiratory dynamics." BioSyst. 90(1): 287-294. Murthy, K. S. (2001). "Activation of phosphodiesterase 5 and inhibition of guanylate cyclase by cGMPdependent protein kinase in smooth muscle." Biochem. J. 360(Pt 1): 199-208. Murthy, K. S. (2004). "Modulation of soluble guanylate cyclase activity by phosphorylation." Neurochem. Int. 45(6): 845-851. Murthy, K. S. and H. Zhou (2003). "Selective phosphorylation of the IP3R-I in vivo by cGMP-dependent protein kinase in smooth muscle." Am. J. Physiol. Gastrointest. Liver Physiol. 284(2): G221-30. Mutic, J. J. and J. B. Wolf (2007). "Indirect genetic effects from ecological interactions in Arabidopsis thaliana." Mol. Ecol. 16(11): 2371-2381.

71

Nagle, K. J. and W. J. Bell (1987). "Genetic control of the search tactic of Drosophila melanogaster: an ethometric analysis of rover/sitter traits in adult flies." Behav. Genet. 17(4): 385-408. Nagoshi, E., C. Saini, C. Bauer, T. Laroche, F. Naef and U. Schibler (2004). "Circadian gene expression in individual fibroblasts: cell-autonomous and self-sustained oscillators pass time to daughter cells." Cell 119(5): 693. Nakagawa, S. (2004). "A farewell to Bonferroni: the problems of low statistical power and publication bias." Behav. Ecol. 15(6): 1044-1045. Napolitano, R. and M. P. Juarez (1997). "Entomopathogenous fungi degrade epicuticular hydrocarbons of Triatoma infestans." Arch. Biochem. Biophys. 344(1): 208-14. Nash, D. R., T. D. Als, R. Maile, G. R. Jones and J. J. Boomsma (2008). "A mosaic of chemical coevolution in a large blue butterfly." Science 319(5859): 88-90. Nerniroff, L. and E. Despland (2007). "Consistent individual differences in the foraging behaviour of forest tent caterpillars (Malacosoma disstria)." Can. J. Zool. 85: 1117-1124. Newland, P. L. and P. Yates (2007). "Nitrergic modulation of an oviposition digging rhythm in locusts." J. Exp. Biol. 210(24): 4448-4456. Newland, P. L. and P. Yates (2008). "Nitric oxide modulates salt and sugar responses via different signaling pathways." Chem. Senses 33(4): 347-356. Ni, Y. G., N. Wang, D. J. Cao, N. Sachan, D. J. Morris, R. D. Gerard, M. Kuro-O, B. A. Rothermel and J. A. Hill (2007). "FoxO transcription factors activate Akt and attenuate insulin signaling in heart by inhibiting protein phosphatases." Proc. Natl. Acad. Sci. U.S.A. 104(51): 20517-20522. Nicholls, D. F. and B. G. Quinn (1982). Random coefficient autoregressive models : an introduction. New York, Springer-Verlag. Nien, W. L., S. M. Dauphinee, L. D. Moffat and C. K. L. Too (2007). "Overexpression of the mTOR alpha4 phosphoprotein activates protein phosphatase 2A and increases Stat1[alpha] binding to PIAS1." Mol. Cell. Endocrinol. 263(1-2): 10-17. Nisbet, R. M., M. E., W. S. C. Gurney, W. W. Murdoch and A. M. de Roos (1997). Simple representations of biomass dynamics in structured populations. Mathematical modeling: Ecology, Physiology, and Cell Biology. H. G. Othmer, F. R. Adler, M. A. Lewis and J. C. Dallon. Upper Saddle River, NJ, Prentice Hall: 61-80. Nitabach, M. N. and P. H. Taghert (2008). "Organization of the Drosophila circadian control circuit." Curr. Biol. 18(2): R84-R93. Nobukuni, T., M. Joaquin, et al. (2005). "Amino acids mediate mTOR/raptor signaling through activation of class 3 phosphatidylinositol 3OH-kinase." Proc. Natl. Acad. Sci. U.S.A. 102(40): 14238-14243. Nonacs, P. and K. M. Kapheim (2007). "Social heterosis and the maintenance of genetic diversity." J. Evol. Biol. 20(6): 2253-2265. Okuda, S., T. Yamada, M. Hamajima, M. Itoh, T. Katayama, P. Bork, S. Goto and M. Kanehisa (2008). "KEGG Atlas mapping for global analysis of metabolic pathways." Nucleic Acids Res. 36: W423-6. Oldham, S., R. Bohni, H. Stocker, W. Brogiolo and E. Hafen (2000). "Genetic control of size in Drosophila." Philos. Trans. R. Soc. Lond., Ser. B: Biol. Sci. 355(1399): 945-952.

72

Oldham, S., J. Montagne, T. Radimerski, G. Thomas and E. Hafen (2000). "Genetic and biochemical characterization of dTOR, the Drosophila homolog of the target of rapamycin." Genes Dev. 14(21): 26892694. Oldham, S., H. Stocker, M. Laffargue, F. Wittwer, M. Wymann and E. Hafen (2002). "The Drosophila insulin/IGF receptor controls growth and size by modulating PtdInsP(3) levels." Development 129(17): 4103-4109. Olendorf, R., F. H. Rodd, D. Punzalan, A. E. Houde, C. Hurt, D. N. Reznick and K. A. Hughes (2006). "Frequency-dependent survival in natural guppy populations." Nature 441(7093): 633-636. Orczyk, J., D. M. Morre and D. J. Morre (2005). "Periodic fluctuations in oxygen consumption comparing HeLa (cancer) and CHO (non-cancer) cells and response to external NAD(P)+/NAD(P)H." Mol. Cell. Biochem. 273(1-2): 161. Orme, M. H., S. Alrubaie, G. L. Bradley, C. D. Walker and S. J. Leevers (2006). "Input from Ras is required for maximal PI(3)K signalling in Drosophila." Nat. Cell Biol. 8(11): 1298-1302. Osborne, K. A., A. Robichon, E. Burgess, S. Butland, R. A. Shaw, A. Coulthard, H. S. Pereira, R. J. Greenspan and M. B. Sokolowski (1997). "Natural behavior polymorphism due to a cGMP-dependent protein kinase of Drosophila." Science 277(5327): 834. Passador-Gurgel, G., W.-P. Hsieh, P. Hunt, N. Deighton and G. Gibson (2007). "Quantitative trait transcripts for nicotine resistance in Drosophila melanogaster." Nat. Genet. 39(2): 264-268. Pawliuk, R., K. A. Westerman, et al. (2001). "Correction of sickle cell disease in transgenic mouse models by gene therapy." Science 294(5550): 2368-71. Pereira, H. S., E. Burgess, A. Campos, R. Greenspan, K. Osborne, A. Robichon and M. B. Sokolowski (1995). "Genetic and molecular analysis of foraging behavior in Drosophila melanogaster." Behav. Genet. 25(3): 300-300. Pereira, H. S. and M. B. Sokolowski (1993). "Mutations in the larval foraging gene affect adult locomotory behavior after feeding in Drosophila melanogaster." Proc. Natl. Acad. Sci. USA 90(11): 5044-5046. Perrine, S. P., G. D. Ginder, D. V. Faller, G. H. Dover, T. Ikuta, H. E. Witkowska, S. P. Cai, E. P. Vichinsky and N. F. Olivieri (1993). "A short-term trial of butyrate to stimulate fetal-globin-gene expression in the beta-globin disorders." New Engl. J. Med. 328(2): 81-6. Perrine, S. P., N. F. Olivieri, D. V. Faller, E. P. Vichinsky, G. J. Dover and G. D. Ginder (1994). "Butyrate derivatives. New agents for stimulating fetal globin production in the beta-globin disorders." Am. J. Pediatr. Hematol. Oncol. 16(1): 67-71. Peters, M. I. and E. Despland (2006). "Plasticity in forest tent caterpillar collective foraging schedules." Ethology 112(6): 521-528. Petfield, D., S. F. Chenoweth, H. D. Rundle and M. W. Blows (2005). "Genetic variance in female condition predicts indirect genetic variance in male sexual display traits." Proc. Natl. Acad. Sci. U.S.A. 102(17): 6045-6050. Pilz, R. B. and K. E. Broderick (2005). "Role of cyclic GMP in gene regulation." Front. Biosci. 10: 12391268. Pilz, R. B. and D. E. Casteel (2003). "Regulation of gene expression by cyclic GMP." Circul. Res. 93(11): 1034-1046.

73

Pogue, R., D. M. Morre and D. J. Morre (2000). "CHO cell enlargement oscillates with a temperaturecompensated period of 24 min." Biochim. Biophys. Acta 1498(1): 44. Price, G. R. (1970). "Selection and covariance." Nature 227: 520-521. Promislow, D. (2005). "A regulatory network analysis of phenotypic plasticity in yeast." Am. Nat. 165(5): 515. Promislow, D. E. (2004). "Protein networks, pleiotropy and the evolution of senescence." Proc. Biol. Sci. 271(1545): 1225-1234. Proulx, S. R., S. Nuzhdin and D. E. Promislow (2007). "Direct selection on genetic robustness revealed in the yeast transcriptome." PLoS ONE 2(9): e911. Prout, T. and O. Savolainen (1996). "Genotype-by-environment interaction is not sufficient to maintain variation: Levene and the leafhopper." Am. Nat. 148(5): 930-936. Puig, O., M. T. Marr, M. L. Ruhf and R. Tjian (2003). "Control of cell number by Drosophila FOXO: downstream and feedback regulation of the insulin receptor pathway." Genes Dev. 17(16): 2006-2020. Puig, O. and R. Tjian (2005). "Transcriptional feedback control of insulin receptor by dFOXO/FOXO1." Genes Dev. 19(20): 2435-2446. Pyza, E. and J. Gorska-Andrzejak (2008). "External and internal inputs affecting plasticity of dendrites and axons of the fly's neurons." Acta Neurobiol. Exp. 68(2): 322-333. Radimerski, T., J. Montagne, F. Rintelen, H. Stocker, J. van der Kaay, C. P. Downes, E. Hafen and G. Thomas (2002). "dS6K-regulated cell growth is dPKB/dPI(3)K-independent, but requires dPDK1." Nat. Cell Biol. 4(3): 251-255. Raizen, D. M., K. Cullison, A. I. Pack and M. V. Sundaram (2006). "A novel gain-of-function mutant of the cGMP-dependent protein kinase egl-4 affects multiple physiological processes in C. elegans." Genetics 173: 177-187. Raizen, D. M., J. E. Zimmerman, M. H. Maycock, U. D. Ta, Y. J. You, M. V. Sundaram and A. I. Pack (2008). "Lethargus is a Caenorhabditis elegans sleep-like state." Nature 451(7178): 569. Ramirez, J. M., F. P. Elsen and R. M. Robertson (1999). "Long-term effects of prior heat shock on neuronal potassium currents recorded in a novel insect ganglion slice preparation." J. Neurophysiol. 81(2): 795. Reiling, J. H., K. T. Doepfner, E. Hafen and H. Stocker (2005). "Diet-dependent effects of the Drosophila Mnk1/Mnk2 homolog Lk6 on growth via eIF4E." Curr. Biol. 15(1): 24-30. Remold, S. K. and R. E. Lenski (2001). "Contribution of individual random mutations to genotype-byenvironment interactions in Escherichia coli." Proc. Natl. Acad. Sci. USA 98(20): 11388-11393. Renger, J. J., W. D. Yao, M. B. Sokolowski and C. F. Wu (1999). "Neuronal polymorphism among natural alleles of a cGMP-dependent kinase gene, foraging, in Drosophila." J. Neurosci. 19(19): RC28. Reynolds, C. W. (1987). "Flocks, herds and schools: A distributed behavioral model." SIGGRAPH Comput. Graph. 21(4): 25-34. Richard, F. J., A. Aubert and C. M. Grozinger (2008). "Modulation of social interactions by immune stimulation in honey bee, Apis mellifera, workers." BMC Biol. 6: 50.

74

Rico, B., H. E. Beggs, D. Schahin-Reed, N. Kimes, A. Schmidt and L. F. Reichardt (2004). "Control of axonal branching and synapse formation by focal adhesion kinase." Nat. Neurosci. 7(10): 1059-1069. Riedl, C. A. L., M. Riedl, T. F. C. Mackay and M. B. Sokolowski (2007). "Genetic and behavioral analysis of natural variation in Drosophila melanogaster pupation position." Fly 1(1): 23-32. Rintelen, F., H. Stocker, G. Thomas and E. Hafen (2001). "PDK1 regulates growth through Akt and S6K in Drosophila." Proc. Natl. Acad. Sci. U.S.A. 98(26): 15020-15025. Rodd, F. H., D. N. Reznick and M. B. Sokolowski (1997). "Phenotypic plasticity in the life history traits of guppies: Responses to social environment." Ecology 78(2): 419-433. Rodd, F. H. and M. B. Sokolowski (1995). "Complex origins of variation in the sexual behavior of male Trinidadian guppies, Poecilia reticulata - interactions between social environment, heredity, body size and age." Anim. Behav. 49(5): 1139-1159. Roff, D. A. and A. M. Simons (1997). "The quantitative genetics of wing dimorphism under laboratory and 'field' conditions in the cricket Gryllus pennsylvanicus." Heredity 78: 235-240. Rohlfs, M. and T. S. Hoffmeister (2004). "Spatial aggregation across ephemeral resource patches in insect communities: an adaptive response to natural enemies?" Oecologia 140(4): 654-61. Rollmann, S. M., M. M. Magwire, T. J. Morgan, E. D. Ozsoy, A. Yamamoto, T. F. Mackay and R. R. Anholt (2006). "Pleiotropic fitness effects of the Tre1-Gr5a region in Drosophila melanogaster." Nat. Genet. 38(7): 824-829. Rouault, J., P. Capy and J. M. Jallon (2000). "Variations of male cuticular hydrocarbons with geoclimatic variables: an adaptative mechanism in Drosophila melanogaster?" Genetica 110(2): 117. Rowe, L. and D. Houle (1996). "The Lek Paradox and the Capture of Genetic Variance by Condition Dependent Traits." Proc. R. Soc. Lond., B, Biol. Sci. 263(1375): 1415-1421. Ruedi, E. A. and K. A. Hughes (2008). "Natural genetic variation in complex mating behaviors of male Drosophila melanogaster." Behav. Genet. 38(4): 424-36. Rulifson, E. J., S. K. Kim and R. Nusse (2002). "Ablation of insulin-producing neurons in flies: growth and diabetic phenotypes." Science 296(5570): 1118. Rundle, H. D., S. F. Chenoweth and M. W. Blows (2006). "The roles of natural and sexual selection during adaptation to a novel environment." Evolution 60(11): 2218-25. Rundle, H. D., S. F. Chenoweth, P. Doughty and M. W. Blows (2005). "Divergent selection and the evolution of signal traits and mating preferences." PLoS Biol. 3(11): e368. Sackton, T. B., B. P. Lazzaro, T. A. Schlenke, J. D. Evans, D. Hultmark and A. G. Clark (2007). "Dynamic evolution of the innate immune system in Drosophila." Nat. Genet. 39(12): 1461-1468. Sarbassov, D. D., S. M. Ali, D.-H. Kim, D. A. Guertin, R. R. Latek, H. Erdjument-Bromage, P. Tempst and D. M. Sabatini (2004). "Rictor, a novel binding partner of mTOR, defines a rapamycin-insensitive and Raptor-independent pathway that regulates the cytoskeleton." Curr. Biol. 14(14): 1296-1302. Sarbassov, D. D., D. A. Guertin, S. M. Ali and D. M. Sabatini (2005). "Phosphorylation and regulation of Akt/PKB by the rictor-mTOR complex." Science 307(5712): 1098-1101. Saucedo, L. J., X. Gao, D. A. Chiarelli, L. Li, D. Pan and B. A. Edgar (2003). "Rheb promotes cell growth as a component of the insulin/TOR signalling network." Nat. Cell Biol. 5(6): 566-571.

75

Scarcelli, N., J. M. Cheverud, B. A. Schaal and P. X. Kover (2007). "Antagonistic pleiotropic effects reduce the potential adaptive value of the FRIGIDA locus." Proc. Natl. Acad. Sci. U.S.A. 104(43): 1698691. Scheiner, R., M. B. Sokolowski and J. Erber (2004). "Activity of cGMP-dependent protein kinase (PKG) affects sucrose responsiveness and habituation in Drosophila melanogaster." Learn. Memory 11(3): 303311. Schieke, S. M., J. P. McCoy, Jr. and T. Finkel (2008). "Coordination of mitochondrial bioenergetics with G1 phase cell cycle progression." Cell Cycle 7(12): 1782-7. Schieke, S. M., D. Phillips, J. P. McCoy, Jr., A. M. Aponte, R.-F. Shen, R. S. Balaban and T. Finkel (2006). "The mammalian target of rapamycin (mTOR) pathway regulates mitochondrial oxygen consumption and oxidative capacity." J. Biol. Chem. 281(37): 27643-27652. Schlossmann, J., A. Ammendola, et al. (2000). "Regulation of intracellular calcium by a signalling complex of IRAG, IP3 receptor and cGMP kinase Ibeta." Nature 404(6774): 197-201. Sergeant, G. P., L. Johnston, N. G. McHale, K. D. Thornbury and M. A. Hollywood (2006). "Activation of the cGMP/PKG pathway inhibits electrical activity in rabbit urethral interstitial cells of Cajal by reducing the spatial spread of Ca2+ waves." J. Physiol. (Lond.) 574(1): 167-181. Sgro, C. M. and A. A. Hoffmann (2004). "Genetic correlations, tradeoffs and environmental variation." Heredity 93(3): 241. Shaulsky, G. and R. H. Kessin (2007). "The cold war of the social amoebae." Curr. Biol. 17(16): R684. Shaver, S. A., C. J. Varnam, A. J. Hilliker and M. B. Sokolowski (1998). "The foraging gene affects adult but not larval olfactory-related behavior in Drosophila melanogaster." Behav. Brain Res. 95(1): 23. Simons, A. M. and D. A. Roff (1996). "The effect of a variable environment on the genetic correlation structure in a field cricket." Evolution 50(1): 267-275. Smolenski, A., W. Poller, U. Walter and S. M. Lohmann (2000). "Regulation of human endothelial cell focal adhesion sites and migration by cGMP-dependent Protein Kinase I." J. Biol. Chem. 275(33): 2572325732. Sohn, J. J. (1977). "Socially induced inhibition of genetically determined maturation in platyfish, Xiphophorus maculatus." Science 195(4274): 199-201. Sokolowski, M. B. (1980). "Foraging strategies of Drosophila melanogaster: a chromosomal analysis." Behav. Genet. 10(3): 291. Sokolowski, M. B. and K. P. Hansell (1992). "The foraging locus - behavioral tests for normal muscle movement in rover and sitter Drosophila melanogaster larvae." Genetica 85(3): 205-209. Sokolowski, M. B. and R. I. Hansell (1983). "Drosophila larval foraging behavior. I. The sibling species, D. melanogaster and D. simulans." Behav. Genet. 13(2): 159. Sokolowski, M. B., C. Kent and J. Wong (1984). "Drosophila larval foraging behaviour: I Developmental Stages." Anim. Behav. 32: 645-651. Sokolowski, M. B., C. Kent and J. L. Wong (1983). "Do rover and sitter larval forager types in Drosophila melanogaster reflect differences in general activity levels?" Behav. Genet. 13(5): 553-553.

76

Sokolowski, M. B. and D. Wahlsten (2001). Gene-environment interaction and complex behavior. Methods in Genomic Neuroscience. H. R. Chin and S. O. Moldin, CRC Press: 3. Sorensen, J. G., M. M. Nielsen and V. Loeschcke (2007). "Gene expression profile analysis of Drosophila melanogaster selected for resistance to environmental stressors." J. Evol. Biol. 20(4): 1624. Spencer, C. C. and D. E. Promislow (2005). "Age-specific changes in epistatic effects on mortality rate in Drosophila melanogaster." J. Hered. 96(5): 513. Sun, S.-Y., L. M. Rosenberg, X. Wang, Z. Zhou, P. Yue, H. Fu and F. R. Khuri (2005). "Activation of Akt and eIF4E survival pathways by rapamycin-mediated mammalian Target of Rapamycin inhibition." Cancer Res. 65(16): 7052-7058. Surks, H. K., N. Mochizuki, Y. Kasai, S. P. Georgescu, K. M. Tang, M. Ito, T. M. Lincoln and M. E. Mendelsohn (1999). "Regulation of myosin phosphatase by a specific interaction with cGMP- Dependent Protein Kinase I." Science 286(5444): 1583. Svetec, N. and J. F. Ferveur (2005). "Social experience and pheromonal perception can change male-male interactions in Drosophila melanogaster." J. Exp. Biol. 208(Pt 5): 891. Tadesse, M. G. and J. G. Ibrahim (2004). "A bayesian hierarchical model for the analysis of Affymetrix arrays." Ann. N. Y. Acad. Sci. 1020: 41-48. Tadesse, M. G., J. G. Ibrahim, R. Gentleman, S. Chiaretti, J. Ritz and R. Foa (2005). "Bayesian error-invariable survival model for the analysis of GeneChip arrays." Biometrics 61(2): 488-497. Tamura, M., J. G. Gu, K. Matsumoto, S. Aota, R. Parsons and K. M. Yamada (1998). "Inhibition of cell migration, spreading, and focal adhesions by tumor suppressor PTEN." Science 280(5369): 1614-1617. Tanay, A., I. Steinfeld, M. Kupiec and R. Shamir (2005). "Integrative analysis of genome-wide experiments in the context of a large high-throughput data compendium." Mol. Syst. Biol. Taniguchi, C. M., T. T. Tran, T. Kondo, J. Luo, K. Ueki, L. C. Cantley and C. R. Kahn (2006). "Phosphoinositide 3-kinase regulatory subunit p85{alpha} suppresses insulin action via positive regulation of PTEN." Proc. Natl. Acad. Sci. U.S.A. 103(32): 12093-12097. Tankanitlert, J., T. A. Howard, et al. (2006). "A pharmacokinetic study of paracetamol in Thai betathalassemia/HbE patients." Eur. J. Clin. Pharmacol. 62(9): 743-8. Tatar, M., A. Kopelman, D. Epstein, M.-P. Tu, C.-M. Yin and R. S. Garofalo (2001). "A mutant Drosophila insulin receptor homolog that extends life-span and impairs neuroendocrine function." Science 292(5514): 107-110. Teleman, A. A., Y. W. Chen and S. M. Cohen (2005). "4E-BP functions as a metabolic brake used under stress conditions but not during normal growth." Genes Dev. 19(16): 1844-1848. Teleman, A. A., Y. W. Chen and S. M. Cohen (2005). "Drosophila Melted modulates FOXO and TOR activity." Dev. Cell 9(2): 271-281. Thompson, J. N. (2005). The geographic mosaic of coevolution. Chicago, Univ. Chicago Press. Topisirovic, I., M. Ruiz-Gutierrez and K. L. B. Borden (2004). "Phosphorylation of the eukaryotic translation initiation factor eIF4E contributes to Its transformation and mRNA transport activities." Cancer Res. 64(23): 8639-8642.

77

Tsokas, P., T. Ma, R. Iyengar, E. M. Landau and R. D. Blitzer (2007). "Mitogen-Activated Protein Kinase upregulates the dendritic translation machinery in long-term potentiation by controlling the mammalian Target of Rapamycin pathway." J. Neurosci. 27(22): 5885-5894. Tu, B. P., R. E. Mohler, J. C. Liu, K. M. Dombek, E. T. Young, R. E. Synovec and S. L. McKnight (2007). "Cyclic changes in metabolic state during the life of a yeast cell." Proc. Natl. Acad. Sci. U.S.A. 104(43): 16886. Tukey, J. W. (1977). Exploratory Data Analysis, Addison-Wesley. Turelli, M. and N. H. Barton (2004). "Polygenic variation maintained by balancing selection: pleiotropy, sex-dependent allelic effects and G x E Interactions." Genetics 166(2): 1053-1079. van der Linden, A. M., S. Wiener, Y. J. You, K. Kim, L. Avery and P. Sengupta (2008). "The EGL-4 PKG acts with KIN-29 Salt-Inducible Kinase and Protein Kinase A to regulate chemoreceptor gene expression and sensory behaviors in Caenorhabditis elegans." Genetics 180(3): 1475-91. Van Homrigh, A., M. Higgie, K. McGuigan and M. W. Blows (2007). "The depletion of genetic variance by sexual selection." Curr. Biol. 17(6): 528-32. van Swinderen, B. and R. J. Greenspan (2005). "Flexibility in a gene network affecting a simple behavior in Drosophila melanogaster." Genetics 169(4): 2151-2163. Vereshchagina, N., M.-C. Ramel, E. Bitoun and C. Wilson (2008). "The protein phosphatase PP2A-B' subunit Widerborst is a negative regulator of cytoplasmic activated Akt and lipid metabolism in Drosophila." J. Cell Sci. 121(20): 3383-3392. Vereshchagina, N. and C. Wilson (2006). "Cytoplasmic activated protein kinase Akt regulates lipid-droplet accumulation in Drosophila nurse cells." Development 133(23): 4731-4735. Vieira, C., E. G. Pasyukova, Z.-B. Zeng, J. B. Hackett, R. F. Lyman and T. F. C. Mackay (2000). "Genotype-Environment Interaction for Quantitative Trait Loci affecting life span in Drosophila melanogaster." Genetics 154(1): 213-227. Waddington, C. H. (1942). "Canalization of development and the inheritance of acquired characters." Nature 150: 563-565. Wade, M. J. (2007). "The co-evolutionary genetics of ecological communities." Nat. Rev. Genet. 8(3): 185195. Wahlsten, D., P. Metten, et al. (2003). "Different data from different labs: lessons from studies of geneenvironment interaction." J. Neurobiol. 54(1): 283. Walling, C. A., N. J. Royle, N. B. Metcalfe and J. Lindstrom (2007). "Green swordtails alter their age at maturation in response to the population level of male ornamentation." Biol. Lett. 3(2): 144-146. Wang, H. G., F. M. Lu, et al. (2005). "Presynaptic and postsynaptic roles of NO, cGK, and RhoA in longlasting potentiation and aggregation of synaptic proteins." Neuron 45(3): 389-403. Wang, L., Z. G. Zhang, R. L. Zhang and M. Chopp (2005). "Activation of the Pl3-K/Akt pathway mediates cGMP enhanced-neurogenesis in the adult progenitor cells derived from the subventricular zone." J. Cereb. Blood Flow Metab. 25(9): 1150-1158. Wang, M. C., D. Bohmann and H. Jasper (2005). "JNK extends life span and limits growth by antagonizing cellular and organism-wide responses to insulin signaling." Cell 121(1): 115-125.

78

Wang, S., R. Pogue, D. M. Morre and D. J. Morre (2001). "NADH oxidase activity (NOX) and enlargement of HeLa cells oscillate with two different temperature-compensated period lengths of 22 and 24 minutes corresponding to different NOX forms." Biochim. Biophys. Acta 1539(3): 192. Wang, X., J. L. Pluznick, D. C. Settles and S. C. Sansom (2007). "Association of VASP with TRPC4 in PKG-mediated inhibition of the store-operated calcium response in mesangial cells." Am. J. Physiol. Renal Physiol. 293(6): F1768-1776. Wang, Y., A. Cheng and M. P. Mattson (2006). "The PTEN phosphatase is essential for long-term depression of hippocampal synapses." Neuromolecular Med. 8(3): 329-336. Weider, L. J., W. Makino, K. Acharya, K. L. Glenn, M. Kyle, J. Urabe and J. J. Elser (2005). "Genotype x environment interactions, stoichiometric food quality effects, and clonal coexistence in Daphnia pulex." Oecologia 143(4): 537. Weinkove, D., S. J. Leevers, L. K. MacDougall and M. D. Waterfield (1997). "p60 is an adaptor for the Drosophila phosphoinositide 3-kinase, Dp110." J. Biol. Chem. 272(23): 14606-10. Weinkove, D., T. P. Neufeld, T. Twardzik, M. D. Waterfield and S. J. Leevers (1999). "Regulation of imaginal disc cell size, cell number and organ size by Drosophila class I(A) phosphoinositide 3-kinase and its adaptor." Curr. Biol. 9(18): 1019-1029. Welshhans, K. and V. Rehder (2005). "Local activation of the nitric oxide cyclic guanosine monophosphate pathway in growth cones regulates filopodial length via protein kinase G, cyclic ADP ribose and intracellular Ca2+ release." Eur. J. Neurosci. 22(12): 3006-3016. Wentzell, A. M., H. C. Rowe, B. G. Hansen, C. Ticconi, B. A. Halkier and D. J. Kliebenstein (2007). "Linking metabolic QTLs with network and cis-eQTLs controlling biosynthetic pathways." PLoS Genet. 3(9): e162 OP. Werner, M. E., P. Zvara, A. L. Meredith, R. W. Aldrich and M. T. Nelson (2005). "Erectile dysfunction in mice lacking the large-conductance calcium-activated potassium (BK) channel." J. Physiol. (Lond.) 567(2): 545-556. Wertheim, B. (2005). "Evolutionary ecology of communication signals that induce aggregative behaviour." Oikos 109(1): 117. Wertheim, B., E. J. van Baalen, M. Dicke and L. E. Vet (2005). "Pheromone-mediated aggregation in nonsocial arthropods: an evolutionary ecological perspective." Annu. Rev. Entomol. 50: 321. Westphal, R. S., R. L. Coffee Jr, A. Marotta, S. L. Pelech and B. E. Wadzinski (1999). "Identification of kinase-phosphatase signaling modules composed of p70 S6 kinase-Protein Phosphatase 2A (PP2A) and p21-activated kinase-PP2A." J. Biol. Chem. 274(2): 687-692. White, R. E. (1999). "Cyclic GMP and ion channel regulation." Adv. Second Messenger Phosphoprotein Res. 33: 251-277. White, R. E., A. B. Lee, A. D. Shcherbatko, T. M. Lincoln, A. Schonbrunn and D. L. Armstrong (1993). "Potassium channel stimulation by natriuretic peptides through cGMP-dependent dephosphorylation." Nature 361(6409): 263. Whitham, T. G., J. K. Bailey, et al. (2006). "A framework for community and ecosystem genetics: from genes to ecosystems." Nat. Rev. Genet. 7(7): 510-23.

79

Whitham, T. G., S. P. DiFazio, J. A. Schweitzer, S. M. Shuster, G. J. Allan, J. K. Bailey and S. A. Woolbright (2008). "Extending Genomics to Natural Communities and Ecosystems." Science 320(5875): 492-495. Williams, G. C. (1957). "Pleiotropy, natural selection, and the evolution of senescence." Evolution 11(4): 398-411. Williams, K. D., M. Busto, M. L. Suster, A. K. C. So, Y. Ben-Shahar, S. J. Leevers and M. B. Sokolowski (2006). "Natural variation in Drosophila melanogaster diapause due to the insulin-regulated PI3-kinase." Proc. Natl. Acad. Sci. USA 103(43): 15911-15915. Willy, P. J., R. Kobayashi and J. T. Kadonaga (2000). "A basal transcription factor that activates or represses transcription." Science 290(5493): 982-984. Wingrove, J. A. and P. H. O'Farrell (1999). "Nitric oxide contributes to behavioral, cellular, and developmental responses to low oxygen in Drosophila." Cell 98(1): 105. Wittwer, F., M. Jaquenoud, W. Brogiolo, M. Zarske, P. Wustemann, R. Fernandez, H. Stocker, M. P. Wymann and E. Hafen (2005). "Susi, a negative regulator of Drosophila PI3-kinase." Dev. Cell 8(6): 817827. Wolf, J. B. (2000). Indirect genetic effects and gene interactions. Epistasis and the evolutionary process. J. B. Wolf, E. D. Brodie III and M. J. Wade. New York, Oxford University Press: 158-176. Wolf, J. B. (2003). "From the Cover: Genetic architecture and evolutionary constraint when the environment contains genes." Proc. Natl. Acad. Sci. U.S.A. 100(8): 4655-4660. Wolf, J. B., E. D. Brodie III and M. J. Wade (2004). The genotype-environment interaction and evolution when the environment contains genes. Phenotypic Plasticity: Functional and Conceptual approaches. T. J. deWitt and R. Scheiner. Oxford, Oxford University Press: 173-190. Wolf, J. B., T. T. Vaughn, L. S. Pletscher and J. M. Cheverud (2002). "Contribution of maternal effect QTL to genetic architecture of early growth in mice." Heredity 89(4): 300-310. Wright, S. (1921). "Correlation and Causation." J. Agric. Res. 20. Wright, S. (1932). The roles of mutation, inbreeding, cross-breeding, and selection in evolution. 6th International Congress of Genetics. Wright, S. (1934). "The method of path coefficients." Ann. Math. Stat. 5. Wu, M. Y. W., M. Cully, D. Andersen and S. J. Leevers (2007). "Insulin delays the progression of Drosophila cells through G2/M by activating the dTOR/dRaptor complex." EMBO J. 26(2): 371-379. Wullschleger, S., R. Loewith and M. N. Hall (2006). "TOR signaling in growth and metabolism." Cell 124(3): 471-484. Wyatt, T. D. (2003). Pheromones and Animal Behaviour: Communication by Smell and Taste. New York, NY, Cambridge University Press. Xia, C. Z. (2001). "Phosphorylation and regulation of G-protein-activated phospholipase C-beta 3 by cGMP-dependent protein kinases." J. Biol. Chem. 276(23): 19770-19777. Yamada, Y. and Y. Ohshima (2003). "Distribution and movement of Caenorhabditis elegans on a thermal gradient." J. Exp. Biol. 206(Pt 15): 2581-2593.

80

Yao, X. P., H. Y. Kwan and Y. Huang (2005). "Regulation of TRP channels by phosphorylation." Neurosignals 14(6): 273-280. Yin, Y. and W. H. Shen (2008). "PTEN: a new guardian of the genome." Oncogene 27(41): 5443-5453. Yin, Y., H. Yuan, C. Wang, N. Pattabiraman, M. Rao, R. G. Pestell and R. I. Glazer (2006). "3Phosphoinositide-dependent protein kinase-1 activates the peroxisome proliferator-activated receptor{gamma} and promotes adipocyte differentiation." Mol. Endocrinol. 20(2): 268-278. You, Y. J., J. Kim, D. M. Raizen and L. Avery (2008). "Insulin, cGMP, and TGF-beta signals regulate food intake and quiescence in C. elegans: a model for satiety." Cell Metab. 7(3): 249-257. Young, H. P., J. A. S. Bachmann and C. Schal (1999). "Food intake in Blattella germanica (L.) nymphs affects hydrocarbon synthesis and its allocation in adults between epicuticle and reproduction." Arch. Insect Biochem. Physiol. 41(4): 214-224. Yuasa, K., H. Michibata, K. Omori and N. Yanaka (1999). "A novel interaction of cGMP-dependent protein kinase I with troponin T." J. Biol. Chem. 274(52): 37429-37434. Zaccolo, M. and M. A. Movsesian (2007). "CAMP and cGMP signaling cross-talk - Role of phosphodiesterases and implications for cardiac pathophysiology." Circul. Res. 100(11): 1569-1578. Zhang, H., J. P. Stallock, J. C. Ng, C. Reinhard and T. P. Neufeld (2000). "Regulation of cellular growth by the Drosophila target of rapamycin dTOR." Genes Dev. 14(21): 2712-2724. Zhang, Y., X. Gao, L. J. Saucedo, B. Ru, B. A. Edgar and D. Pan (2003). "Rheb is a direct target of the tuberous sclerosis tumour suppressor proteins." Nat. Cell Biol. 5(6): 578-581. Zheng, X. Z. and A. Sehgal (2008). "Probing the relative importance of molecular oscillations in the circadian clock." Genetics 178(3): 1147-1155. Zhong, Y., V. Budnik and C. F. Wu (1992). "Synaptic plasticity in Drosophila memory and hyperexcitable mutants: role of cAMP cascade." J. Neurosci. 12(2): 644-651. Zhong, Y. and C. F. Wu (1993). "Differential modulation of potassium currents by cAMP and its long-term and short-term effects: dunce and rutabaga mutants of Drosophila." J. Neurogenet. 9(1): 15-27. Zhong, Y. and C. F. Wu (2004). "Neuronal activity and adenylyl cyclase in environment-dependent plasticity of axonal outgrowth in Drosophila." J. Neurosci. 24(6): 1439-1445. Zhou, X. B., C. Arntz, S. Kamm, K. Motejlek, U. Sausbier, G. X. Wang, P. Ruth and M. Korth (2001). "A molecular switch for specific stimulation of the BKCa channel by cGMP and cAMP kinase." J. Biol. Chem. 276(46): 43239-43245. Zhou, X. B., P. Ruth, J. Schlossmann, F. Hofmann and M. Korth (1996). "Protein phosphatase 2A is essential for the activation of Ca2+-activated K+ currents by cGMP-dependent protein kinase in tracheal smooth muscle and Chinese hamster ovary cells." J. Biol. Chem. 271(33): 19760-19767. Zhou, Y., J. Wang, H. Wen, O. Kucherovsky and I. B. Levitan (2002). "Modulation of Drosophila slowpoke calcium-dependent potassium channel activity by bound protein kinase a catalytic subunit." J. Neurosci. 22(10): 3855-3863.

81

Chapter 1. A Model-Based Analysis of Chemical and Temporal Patterns of Cuticular Hydrocarbons in Male Drosophila melanogaster. Note: this chapter was published as (Kent, Azanchi et al. 2007): Kent, Clement F., Reza Azanchi, Ben Smith, Adrienne Chu and Joel D. Levine (2007). "A Model-Based Analysis of Chemical and Temporal Patterns of Cuticular Hydrocarbons in Male Drosophila melanogaster." PLoS ONE 2(9): e962. I measured male total lipids, did all analyses, and wrote the paper in collaboration with JDL.

Abstract Drosophila Cuticular Hydrocarbons (CH) influence courtship behaviour, mating, aggregation, oviposition, and resistance to desiccation. We measure levels of 24 different CH compounds of individual male D. melanogaster hourly under a variety of environmental (LD/DD) conditions. Using a model-based analysis of CH variation, we develop an improved normalization method for CH data, and show that CH compounds have reproducible cyclic within-day temporal patterns of expression which differ between LD and DD conditions. Multivariate clustering of expression patterns identifies 5 clusters of co-expressed compounds with common chemical characteristics. Turnover rate estimates suggest CH production may be a significant metabolic cost. Male cuticular hydrocarbon expression is a dynamic trait influenced by light and time of day; since abundant hydrocarbons affect male sexual behavior, males may present different pheromonal profiles at different times and under different conditions.

Introduction Chemical communication is fundamentally important in the biology of many organisms. Finding mates and food, identification of good oviposition sites, and defense against herbivores are all mediated by chemical signals in many organisms (Howard and Blomquist 2005). Research on the chemical senses is experiencing a surge due to the cloning and identification of olfactory and gustatory receptors (Hallem and Carlson 2006; Jones, Cayirlioglu et al. 2007). The increasing understanding of chemical signaling at the molecular level suggests that production, emission, reception, and processing of chemical signals exchanged between animals can be understood more mechanistically in the context of social interactions.

Chemical signals emitted by Drosophila are made within the fly and are found on the body surface, and include sex pheromones (Jallon 1984; Cobb and Jallon 1990; Ferveur and Jallon 1996). Courtship between flies is modulated by such cuticular hydrocarbons carbons. In addition, other social

82

interactions in Drosophila, such as female egg laying, aggregation, or the social resetting of circadian clocks (Levine, Funes et al. 2002), are also mediated by chemical cues. The analysis of chemical signals has been complicated, in part because of the high between-fly variability associated with their measure.

Studying chemical signaling in Drosophila is advantageous because advances in molecular genetics are making it possible to associate genes with a signaling phenotype or output. Rapid advances are being made in identifying genes responsible for key enzymes in cuticular hydrocarbon (CH) synthesis (Ferveur 1991; Dallerac, Labeur et al. 2000; Chertemps, Duportets et al. 2005; Chertemps, Duportets et al. 2006; Chertemps, Duportets et al. 2007), and mutations or natural variants of these genes have been shown to change CH levels within a species and to thus contribute to reproductive isolation between sibling species such as D.melanogaster , D. simulans, D. santomea, and D. sechellia (Coyne, Crittenden et al. 1994; Coyne 1996; Takahashi, Tsaur et al. 2001; Gleason, Jallon et al. 2005; Mas and Jallon 2005). In addition, the balance between different CH compounds within a single genetically uniform strain of D. melanogaster is changed by environmental conditions such as rearing temperature (Rouault, Marican et al. 2004). Thus, elucidating how the cuticular hydrocarbon phenotype of the model organism D. melanogaster varies in response to environmental variables and to the passage of time contributes to our understanding of the phenotypic plasticity of this important trait.

Here we use wild-type males to determine (a) the natural range of variation of CH expression of a single genotype under controlled environmental conditions , (b) whether there are time of day patterns in such variation, (c) where patterns of expression co-vary between compounds and what this reveals about chemical pathways of CH synthesis, and (d) what natural variation in CH levels implies about the metabolic cost of pheromone signaling.

We ask whether cuticular hydrocarbon abundances change throughout the day in D. melanogaster males of a single genotype. In order to study daily variation in these compounds we have had to review methods for analyzing these compounds. We have characterized these methods and identified an underlying general linear model of cuticular hydrocarbon abundances which explains a higher percent of variance than previous methods. The model can be used to deduce new features of CH variation, but perhaps its most important application is to a normalization method which considerably reduces error variances in CH measurements.

Using our model-based analysis, we show that there are five clusters of compounds whose abundance patterns co-vary, and that membership of these clusters is a function of chemical structure and chain length. This demonstrates that although CH abundance variation is high, it is nonetheless strongly constrained by chemistry. We find significant differences in the patterns of hydrocarbon abundance in these

83

clusters in LD (a 12 hours light-12 hours dark cycle) and in DD (24 hours dark), significant differences for two clusters in day versus night abundance, and highly significant cyclical components in the 24 hour variation whose frequencies vary between clusters and between LD and DD. We estimate hydrocarbon turnover rates on the cuticle, leading to the conclusion that chemical signaling has a significant metabolic cost for male flies.

We also identify a new feature of CH variation we call Abundance Variability (AV) by analogy to the beta volatility term of mathematical portfolio analysis (Markowitz 1970). AV is not a chemical volatility; it represents the magnitude of an individual compound’s response to fluctuations in CH total abundance (TA), and varies independently of mean compound abundances over time and in response to light. AV is strongly influenced by cluster membership, carbon chain length, and double bond position, further demonstrating that the underlying chemistry of CH production constrains the observed patterns of CH abundance.

Results We consider first the distribution of total CH abundance in individual flies and how this affects different classes of compounds. From this we derive an unbiased normalization technique for CH abundances and use this to characterize patterns of CH coexpression. Clustering methods applied to such normalized data reveal clusters of coexpressed compounds which are related to the chemical synthesis pathways of the compounds. We show that cluster membership correlates with phenotypes such as daynight mean compound differences and effect of light on abundance patterns. Finally we show that frequencies of temporal differences in hydrocarbon abundance span a range from 24 to 3 or less hours, and use this to estimate minimum CH turnover rates per day.

Patterns of Total and Relative Abundance Patterns of abundance of CHs might be unique to particular genotypes or environmental conditions, or they might be constrained by the underlying biochemical synthesis pathways to a sex- and species-specific program. D. melanogaster total CH abundance is highly variable even among genetically identical male flies in the same environmental conditions (over 4-fold variation). High endogenous variation within a condition can obscure between-condition changes.

The pattern in total CH abundance (TA) is illustrated in Figure 1.1, showing the very wide spread of individual fly values. The distribution of total abundance is approximately lognormal; it is not uncommon to find flies in the same vial whose total abundance differs by a factor of 2. However, although individual fly TA varies greatly within a vial or condition, the population distribution of TA values

84

observed is reproducible within one condition. Mean TA does not vary with light environment, but the standard deviation of TA is 23% higher in DD (p=.00015) (Figure 1.1).

85

Figure 1.1 Distribution of individual fly total abundance (TA) values Distribution of individual fly total abundance (TA) values for wild-type males. (A) LD and (B) DD. Mean, LD = 2.27 µg/ fly (N=277), DD= 2.30 µg/fly (N=348.); no significant difference. The standard deviation in DD is 23% higher than in LD (F347,276=1.518, p=.00015).

No significant temporal pattern in mean TA was detected in either light environment via several techniques, including ANOVA and Fourier decomposition (see Methods). Thus mean TA is insensitive to time and light in our conditions, but individual fly TA varies over a four-fold range within samples. When analyzing abundances of individual compounds, this high TA range causes large error variances.

Several normalization techniques have been used to minimize effects of this high within-treatment endogenous variation. If endogenous variation causes all compounds to vary as a simple multiple of the TA, then expressing each compound amount as a proportion of the TA for the sample is an unbiased estimator of relative abundances (RA). The relative abundance measure also corrects for internal variability of the measurement system, and has been used extensively in Drosophila CH literature (Jallon, Kunesch et al. 1997; Rouault, Capy et al. 2000; Mas and Jallon 2005). Other authors have used a log-contrast method in which the logarithm of the ratio of a compound of interest to another compound is used (Blows and Allan 1998; Blows, Chenoweth et al. 2004; Skroblin and Blows 2006). Both methods work well if all compounds respond as a strict multiple to total abundance changes, that is, as a linear relationship passing through the origin. We tested this assumption by fitting a general linear model not constrained to pass through the origin, relating the abundance yi,j of compound j in fly i to a “latent variable” xi:

yi , j (t ) = α j (t ) + β j (t ) xi + ε i , j

(1.1)

In equation 1.1 the latent or hidden variable xi represents an overall compound abundance, α and β the intercept and slope, which depend on time t and compound j, and ε the error or noise term. Latent

86

variable models (factor analysis) are commonly used to test multivariate data for the presence of hidden variables which explain much of the observed variation (Bartholomew and Knott 1999; Loehlin 2004). Our model in equation 1.1 is a generalization of the model implicit in the use of the relative abundance measure RA, for if the relative abundance (proportion) of compounds is independent of total abundance, then equation 1.1 must hold with αj(t)=0 for all compounds j and times t. We call this case the “RA model”. A change of scale helps make clear a prediction of the RA model. If we divide or scale compound and total abundances by their means to get variables y′ and T ′ then the RA model predicts (see Methods for details):

yi′, j = Ti′+ ε i′′, j

(1.2)

In other words, if we regress scaled compound abundances against scaled total abundance, the intercept α ′j of the regression should be not significantly different from 0 and the slope β ′j should be 1. This assumption fails in our data for most compounds (for 15 of 24 compounds in LD and for 19 of 24 in DD, slope differs from 1 with p0) in all cases.

Figure 4.1. Behavioral foraging gene by food interaction. The interaction between for and food [fed vs. food deprived (FD)] is significant (p=0.021) and positive in sign (rovers increase more than sitters; I>0) (for medium composition and ANOVA results see Supplementary Table 4.1a,b and Methods). The food-leaving score (arcsine-transformed proportion of flies leaving a known food source and traversing a maze; see Methods for assay details) is plotted ± 1 standard error of the mean.

211

Metabolites We examined fed and FD forR and fors2 heads to determine compounds most strongly associated with the for response to feeding state, using Fourier Transform Ion Cyclotron Resonance Mass Spectroscopy (FTICR MS, Methods) to detect 750 putative metabolites. We found a significant influence of for (forR or fors2) and feeding state on compounds with the molecular weight (MW) and chemical properties of triacylglycerols (TAG) and polysaccharides (PS) (Figure 4.2a,b; Supplementary Table 4.2a,b). There was a significant main effect of for in carbohydrates, and significant GEI in both carbohydrates and lipids, but in opposite directions (I > 0 for lipids, I < 0 for carbohydrates), with the largest differences found in smaller MW compounds. [Incorporating MW into ANOVA of TAG compounds gave p(for x food)= 1.3·10-13 and p(for x MW)= 1.9·10-5; PS compounds had p(for x food x MW)=0.03 (Supplementary Table 4.2b)]. (Note that the term food in the ANOVA describes the feeding state of Fed vs. FD and the term for refers to the strains forR or fors2). Thus GEI interactions are found for metabolites but their direction I depends on metabolite type. Rovers have a larger drop in lipids when FD than sitters, while sitters have a greater drop in carbohydrates than rovers.

Figure 4.2. Rovers and sitters use energy stores differently. Change in the heads of flies in (a) triacylglycerols (TAG) and (b) polysaccharides (PS) between fed and food deprived (FD) states is higher in rovers than in mutant sitters (positive GEI interaction I > 0) for TAG but lower for PS (negative GxE interaction I < 0; both interactions significant, Supplementary Table 4.2a has ANOVA details). Total signal/noise ratio levels determined using FTICR shown on vertical axis (Methods). FTICR measurements were done on rovers (forR) and mutant sitters (fors2).

212

Whole-fly spectrophotometric measures of total carbohydrates, lipids, and proteins (Methods) showed that adult rovers had almost twice as much energy stored in whole-body lipid and about half the energy stored in carbohydrates compared to adult sitters, whereas protein levels normalized to dry weight were not significantly different between genotypes (Figure 3 gives full statistics). Thus, for genotype strongly affects energy storage strategies. A main effect of genotype in fed flies is consistent with an allocation shift between storage of energy as lipids and as carbohydrates.

Figure 4.3. Fed rovers and sitters store energy differently. The proportion of total calories due to lipids (horizontal axis) and carbohydrates are shown in whole-body measurements of fed 5 day-old males and females for the two sitter (red) and one rover (blue) strains. Data are standardized for fly dry weight (Methods). Rovers store significantly more energy as lipids and significantly less as carbohydrates. For males and for females, mutant and natural sitters didn’t differ (Welch’s t-test, p>.17 all tests) and hence were pooled for a rovers versus sitters comparison. Lipids: males, t=3.26, df=3.41, p=0.039; females t=5.08, df=10.2, p=0.0004. Carbohydrates: males, t=-3.98, df=9.20, p=0.003; females t=-5.64, df=7.84, p=0.0005. Data for n=5 except n=4 for male lipids. Error bars represent ±1 s.e.m.

213

GEI in gene expression depends on metabolic role If for affects both behavioral and metabolic GEI and plasticity in a food dependent manner, how is this reflected at the level of gene expression? To examine for’s effect on transcript levels we performed whole-genome microarray analysis on heads of rovers and sitters and sitter mutants under fed and food deprived (FD) conditions (Methods). Array results were verified using qRTPCR on two genes with strong rover-sitter differences and involved in carbohydrate metabolism (Treh, trehalase, and CG10924, human homolog is PCK1 phosphoenolpyruvate carboxykinase 1 (soluble) (Supplementary Figure 4.3).

Overall, the expression of genes involved in the breakdown of food to provide energy (catabolism) was significantly altered (had strong GEI), with rovers decreasing and sitters increasing their expression when food is present (I for > BG x food > for x food when measured by variance explained.

Our second method uses the Storey-Tibshirani false discovery rate analysis to estimate the rate at which genes are in fact significantly expressed (Storey and Tibshirani 2003). This method uses a mixturemodel approach to estimate the proportion π0 of genes matching the null hypothesis of no effect. Then πalt = 1- π0 estimates the proportion of genes matching the alternative hypothesis of an effect. π0 and hence πalt depend not only on the true rate of differential expression of genes (DEG) but also on the signal to noise ratio of the array technology. Thus for low expression genes with poorer signal to noise ratio π0 will be higher even if the true proportion of DEG is unchanged. We calculated π0 both for the top-1000 highest expressing genes and for increasingly large groups of genes based on minimum expression levels. If signal to noise is the only factor affecting π0 then the intercept of the curve of π0 values based on expression is an estimator of 1-DEG. In practice we found good agreement between the latter method and π0 for the top 1000 genes, so we report the top-1000 figure.

This analysis was done for each ANOVA p-value (for all main and interaction effects). The πalt = 1- π0 for a given set of p-values (e.g. for main effect p values) estimates the true DEG rate for that effect. Our top-1000 πalt values were BG=0.84, food=0.73, BG x food = 0.63, for = 0.59, and for x food = 0.57.

216

That is, among the top-1000 genes by mean expression level, 84% had a main effect of BG, 73% of food, 63% showed BG x food, , 59% had a main effect of for and 57% had for x food GEI. Thus, although for x food GEI affects the smallest proportion of top-expressing genes, it still has an effect on 57% of these genes.

A quantitative measure of plasticity is higher in rovers The interaction between for and feeding state could be due either to the genotypes responding in equal amounts but opposite directions to feeding (same plasticity of each genotype), or to one genotype responding more than the other (differences in magnitude of plasticity). To quantify differences in plasticity between rovers and sitters, we calculated an index of plasticity called Relative Nutrient Sensitivity (RNS) for any given trait as the difference between the size of the trait’s rover response to food and the size of the sitter response: RNS=(|rover change| - |sitter change|)/C (where C=1 for log2 transformed data, otherwise C=mean level of trait). In other words, for each trait compared (e.g., behavior, metabolite, gene expression), RNS compares the absolute magnitude of for-dependent changes in response to food rather than their direction (see Methods).

When RNS>0, rovers show a larger response; when RNS0 (rovers change more) in 8 of 9 behaviour cases (89%, Figure 4.5a, p= 0.03). For metabolites, gene expression, and functional gene categories, a significant majority of traits had RNS>0 (Figure 4.5b-d, p0. Of the traits with RNS0 for 8 of 9 (89%) and RNS=-0.004 for the ninth. Student t for RNS ≠ 0, t=2.99, df=8, p=0.009. (b) Metabolite plasticity: RNS for compounds with a significant response to food. 84% of these had RNS>0. Chi-square contingency test χ2= 65.3,df=1, p= 6.3 ·10-16. (c) Gene expression plasticity: RNS for 1000 genes with significant food response. Of these, 77% had RNS>0 (χ2= 305.3,df=1, p< 2.2 ·10-16 ). (d) Functional group plasticity. RNS for 300 Gene Ontology groups with significant food response. In 77% of these RNS>0 (χ2= 88.6,df=1, p< 2.2 ·10-16 ). Average mutant sitter change on food deprivation is about ½ of rover. For simplicity, only rover versus mutant sitter RNS values are shown. This is conservative; rover versus natural sitter gene and gene group RNS distributions were more biased in favour of rovers than the rover vs. mutant sitter (Kolmogorov-Smirnov test for genes, D=0.236, p< 2.2 ·10-16; for gene groups D=0.233, p=0.000022).

218

Hypothesis: plasticity differences may be due to insulin pathway We hypothesized that the for-dependent metabolic plasticity might be mediated by the insulin signaling pathway. This is because rovers exhibit a higher plasticity in response to feeding state and the insulin signaling pathway is a key regulator of the response to food (Grewal 2009).

In the cell, binding of DILPS (Drosophila insulin-like peptides) to the insulin receptor (InR) triggers a signaling cascade with major effects on gene expression (Wu and Brown 2006) (see Figures I.1I.3). Protein translation is increased via phosphorylation of key members of the TOR pathway by the kinases Akt1 (dPKB) and Pk61C (dPDK) (Rintelen, Stocker et al. 2001; Radimerski, Montagne et al. 2002) (Figure I.3). Negative homeostatic control of insulin signaling occurs on the transcriptional level – when signaling is high, foxo is phosphorylated by Akt1 and sequestered in the cytoplasm, but when signaling drops foxo translocates to the nucleus where it stimulates transcription of genes such as InR and the negative regulator of translation Thor (d4EBP) (Junger, Rintelen et al. 2003; Puig, Marr et al. 2003; Puig and Tjian 2005) (Figure I.2). At the transcript level, then, many insulin pathway genes have an inverse relationship to the level of insulin signaling. Our results show that transcription of positive regulators decreased more in fed rovers than sitters, resulting in a negative GEI interaction coefficient I for the group of positive regulators as a whole (Supplementary Figure 4.2). This normal inverse relationship between transcription and insulin signaling is more evident in rovers than sitters. As with RNS, this is true whether mutant or natural sitters are compared, but again as for RNS, the difference between rovers and natural sitters is larger than the difference between rovers and mutant sitters, suggesting that the genetic background of natural sitters may intensify this difference.

Genetic test of interaction of foraging and insulin genes The finding that rovers show larger responses to food suggests that they might also show a larger impact of changes to insulin signaling. We therefore tested whether for interacts with the insulin signaling pathway by means of quantitative complementation crosses for epistasis between mutant insulin pathway genes and alleles of for (Mackay 2001; Gibson and Dworkin 2004; Williams, Busto et al. 2006). We crossed each of the three for genotypes to loss of function mutants of the fly insulin receptor InR, phosphatidylinositol-3-kinase Pi3K92E (or Dp110), and foxo (Methods). InR and Dp110 are positive regulators of insulin signaling; foxo is a negative regulator. Based on our gene expression data, we hypothesized that rovers had higher insulin signaling than sitters, so we expected crosses of rovers with loss of function insulin mutants to be more sitter-like than their controls. We tested food-deprived adults of the resulting 18 trans heterozygote genotypes and compared scores of for;mutant to the for;Balancer which controlled for genetic background effects (see Methods). Recall that food-deprived homozygous rovers show low levels of food leaving behavior (Figure 4.1) while sitters have higher levels. As expected, the control for;Balancer flies show the previously found lower

219

level of behavioural response for rovers compared to sitters (Figure 4.6a-b, solid lines), indicating no direct effects of, or interactions with, the balancer chromosomes. There is, however, a significant interaction in the for;InR and for;dp110 flies (Figure 6a-b dashed lines; Supplementary Table 4.5). Rovers crossed to these insulin pathway mutants become more sitter-like. In contrast, the interaction with negative regulator foxo is not significant (Supplementary Table 4.5).

Figure 4.6. Insulin pathway genes interact with foraging alleles in expression and in food-leaving assay. Flies crossed to mutants of the positive regulators (a) InR and (b) Dp110 (dashed curves) show almost none of the normal food-deprived rover-sitter food-leaving difference (compare to Figure 4.1 and solid balancer curves in this figure). In these quantitative complementation crosses, the food leaving behavior of unfed for;mutant heterozygotes is compared to the unfed for;Balancer heterozygote controls. The difference in food leaving between the control balancer and mutant cross depends significantly on foraging allele, demonstrating interaction between the mutant gene and for. p(Interaction) = 0.012 (InR), p=0.046 (Dp110). Data is arcsine transformed means ± 1 standard error for trials on n days (n=11 for InR and Dp110). Behaviour assays were performed on food-deprived flies as described in Figure 4.1 and Methods. Full ANOVA statistics are in Supplementary Table 4.5.

220

Mutants of positive regulators of insulin signaling make rover food-leaving behavior more like sitters (reduces RNS), while a mutant of negative regulator foxo trended towards making rovers less like sitters (increases RNS). This suggests that there is a significant (epistatic) interaction between for and the two positive regulators of the insulin signaling pathway tested here. This is consistent with rovers experiencing greater shifts in insulin signaling effects between fed and FD states than sitters. There are also differences between natural and mutant sitters in the interaction with InR (Supplementary Table 4.5) suggesting that the difference in genetic backgrounds between these strains may also affect this interaction. For this behavioural measure, natural sitters are intermediate between rovers and mutant sitters, a difference from the trend found in gene expression overall (via RNS, Figure 4.5) or in regulators of insulin signaling (Supplementary Figure 4.2). Thus the effect of the background difference between natural sitters and the other strains varies between gene expression and behavioural measures.

Meta-analysis of insulin and foraging effects on gene expression We performed a bioinformatic meta-analysis comparing our array results to those from three published microarray studies which manipulated insulin/Tor signaling (Guertin, Guntur et al. 2006; Gershman, Puig et al. 2007; Buch, Melcher et al. 2008). This provides additional evidence for transcriptional parallels between for and insulin. We use these studies to identify sets of genes which were up- or down-regulated by the manipulation of insulin/Tor signaling, and which had high enough expression levels in our data for reliable comparison. To ensure independence of the three analyses we used sets of genes which did not overlap between studies (see Supplementary Table 4.6 for gene selection criteria).

For each up- or down-regulated set of genes identified from a study we calculated the mean log2 fold change between rovers and mutant sitters when Fed or FD. This gave four comparisons per study (up/down regulated in study x Fed/FD in our data). In other words, we used the 3 independent studies to tell us which genes may be transcriptionally regulated by insulin signaling. We then used our data to ask, for the same genes, what the rover-sitter difference in expression is under the two food conditions. Our hypothesis is that rovers have higher insulin signaling when fed than sitters, but not necessarily when food deprived. Hence we predict that genes requiring insulin signaling for their expression should have higher expression levels in fed rovers than in fed sitters, but that this difference may not exist in food deprived rovers and sitters. Similarly, if genes are shown in the independent study to have lower expression when insulin signaling is high (or equivalently, higher expression when insulin expression is reduced), then we predict those genes should have lower expression in fed rovers than in fed sitters.

221

In the first study, Buch et al. (Buch, Melcher et al. 2008) ablated dilp3 secreting cells in adults and used microarrays to compare ablation lines which had reduced insulin signaling to that of controls. Figure 4.7a shows a summary of the four rover-sitter comparisons for this study. Bars on the right labelled “expression down” are for genes whose expression was reduced by dilp3 ablation (i.e. insulin signaling increases expression of these) and bars on the left (“expression up”) are for genes whose expression was increased by dilp3 ablation (genes repressed by insulin signaling). Genes with expression reduced by dilp3 ablation (Fig 4.7a, left) show a negative GEI interaction sign I (rovers higher when food deprived, mutant sitters higher when fed). Conversely genes increased by dilp3 ablation (Fig 4.7a, right) show a positive GEI interaction sign I (no difference when food deprived, rovers higher when fed), in accordance with our predictions. Figure 4.7b gives the four comparisons for genes whose expression was changed by foxo overexpression (Gershman, Puig et al. 2007); Figure 4.7c is for genes changed by rapamycin treatment (Guertin, Guntur et al. 2006). In each case the pattern is similar to dilp3 ablation: genes with expression increased by a manipulation equivalent to lowering insulin/Tor signaling (genes reduced by insulin) show the negative I GEI interaction, while genes whose expression is reduced by the manipulation (genes increased by insulin) show positive I. Full statistics are given in Supplementary Table 4.6. This table also shows that the pattern in I is more significant when natural sitters are used in the analysis than when only mutant sitters are used, so the trends shown in Figure 4.7 apply to both mutant and natural sitters.

In summary, the patterns of GEI interaction strength I in rover-sitter gene expression of genes affected by three different manipulations of insulin/Tor signaling in three independent studies (Guertin, Guntur et al. 2006; Gershman, Puig et al. 2007; Buch, Melcher et al. 2008) are consistent with our hypothesis in each study. Genes requiring insulin signaling for expression show positive rover-sitter I and genes inhibited by insulin signaling show negative rover-sitter I.

222

Figure 4.7. Meta-analysis of 3 manipulations of the insulin/Tor signaling identifies rover-biased genes. Three published studies decreased insulin/Tor effects via (a) ablation of dilp3 expressing cells (Buch, Melcher et al. 2008), (b) overexpression of constitutively active foxo (Gershman, Puig et al. 2007), or (c) rapamycin (Guertin, Guntur et al. 2006). We used data from these papers to identify sets of genes in each study whose expression went up or down in response to the particular insulin/Tor manipulation (Supplementary Table 4.6 gives full statistics and methods). For each gene set (expression up or down), we plot average log2 fold change between rovers and mutant sitters in our study on the vertical axis, one bar for FD flies and one for Fed flies. When gene expression is reduced by insulin signaling (e.g. increases due to dilp3/foxo/rapamycin ablation), food deprived rovers have significantly higher mean expression than sitters (far left in each panel). When gene expression is increased by insulin signaling (e.g. decreases due to dilp3/foxo/rapamycin), fed rover expression is higher than sitters (far right each panel). This was true for gene sets used from the ablation of dilp3 expressing cells publication (Buch, Melcher et al. 2008) (panel A), the overexpression of foxo publication (Gershman, Puig et al. 2007) (panel B), and the rapamycin treatment paper (Guertin, Guntur et al. 2006) (panel C). This results in significant negative interactions (Supp. Table 4.6) for genes repressed by insulin signaling, and significant positive interactions for genes increased by insulin. Error bars = 1 s.e.m. Blue bars, mean rover expression is higher than sitter, red bars, mean sitter expression is higher than rover.

223

Discussion The foraging gene in Drosophila which encodes PKG is known for its importance as a natural variant affecting behavioral and neural plasticity (Engel, Xie et al. 2000; Scheiner, Sokolowski et al. 2004; Mery, Belay et al. 2006; Kaun, Hendel et al. 2007; Kaun, Riedl et al. 2007; Mery, Belay et al. 2007; Kaun, Riedl et al. 2008). We now demonstrate that it also affects metabolic, gene expression, and behavioral plasticity in adult flies. Specifically, rovers show a greater response to changes in their food environment than either mutant or natural sitters for the majority of behavioural, metabolite, and gene expression traits studied here. The pattern of such changes is matched by the pattern of expression of positive regulators of insulin signaling. Combining for alleles with mutants of positive insulin signaling regulators makes rover responses sitter-like, but does not change sitter responses. Collectively these findings suggest that the effect on metabolic, genomic, and behavioural plasticity of foraging works in part through the insulin signaling pathway.

The for product PKG and the insulin pathway are conserved across many animals, from worms to flies and mammals (Fitzpatrick and Sokolowski 2004). PKG has been found to produce behavioral responses to food in flies (Scheiner, Sokolowski et al. 2004; Kaun, Hendel et al. 2007), nematodes (Gray 2004; You, Kim et al. 2008), honeybees (Ben-Shahar, Leung et al. 2003), and ants (Ingram, Oefner et al. 2005; Lucas and Sokolowski 2009). In particular, PKG interacts with insulin and TGF-beta signaling in worms to regulate quiescence, a state possibly related to satiety (You, Kim et al. 2008). In this study we have focused on the insulin pathway, but potential interactions between for and TGF-beta may be a fruitful area for future study in flies.

GEI, foraging, insulin and energy store allocation Expression of insulin pathway genes such as InR and Pi3k92E (Dp110) is inversely related to the strength of insulin signaling via the foxo transcription factor: foxo is retained in the cytoplasm when signaling is high, but translocates to the nucleus and stimulates transcription of pathway genes when signaling is low (Puig, Marr et al. 2003; Puig and Tjian 2005). Since insulin gene expression is opposite to insulin signaling strength, our finding of greater negative transcriptional effects on insulin pathway genes in rovers (Supp. Figure 4.2) suggests the presence of positive insulin signaling in rovers. Since insulin signaling upregulates anabolism and reduces catabolism (Wu and Brown 2006; Lasko and Sonenberg 2007), this is consistent with the patterns we find in genes involved in anabolic (fed rover-biased) and catabolic (fed sitter-biased) processes. And genes which require insulin for expression are overexpressed in rovers, while genes repressed by insulin tend to be overexpressed in sitters (Figure 4.7).

224

The finding of genetic interactions between for and genes in the insulin signaling pathway raises questions for future investigation. Insulin signaling in flies can reduce flow through tricarboxylic acid cycle and oxidative phosphorylation and increase flow through the pentose-phosphate shunt, freeing pyruvate and acetyl CoA for lipogenesis and increasing NADPH and precursors for biosynthesis (Ceddia, Bikopoulos et al. 2003). This is consistent with patterns in fed adult rover catabolic groups (Figure 4.4c, Supplementary table 4.3). Instead of accumulating energy as fat, fed adult sitters accumulate carbohydrates. Because of the lower density of fat and its higher caloric content, rovers store more energy per unit mass than sitters, a difference which should have implications for life history characteristics such as starvation resistance (see below). Several studies note changes in fat stores in flies with mutations in insulin signaling genes. These include, loss of fat in melted mutants (Teleman, Chen et al. 2005), gain of fat in InR, chico and Pi3K (also called Dp110) mutants (Bohni, Riesgo-Escovar et al. 1999; Brogiolo, Stocker et al. 2001) and Pi3Koverexpression in larvae increases accumulation of nutrients in fat (Britton, Lockwood et al. 2002). Nuclear foxo reduces fat, phenocopying starvation (Kramer, Davidge et al. 2003; Teleman, Chen et al. 2005), and it reduces head fat body insulin signaling (Hwangbo, Gershman et al. 2004). Thus, there may be multiple different effects of insulin-related genes on fat. Could foxo mediate lower sitter fat levels? Rapamycin treatment (which acts downstream of foxo specifically on the Tor signaling pathway) also produces similar patterns of effects (Fig 4.6c). Hence indirect effects of insulin signaling on the Tor pathway could also be involved. In support of this, PDK/Pk61C is the gene in the insulin pathway showing strongest transcriptional regulation in rovers versus both natural and mutant sitters (Supplementary Figure 4.2b). PDK phosphorylates ribosomal S6 kinase (S6k), part of Tor regulation of translation (Cohen, Alessi et al. 1997; Rintelen, Stocker et al. 2001; Radimerski, Montagne et al. 2002). Indeed, overexpression of Tor has been shown to increase triglycerides in adult male flies (Teleman, Chen et al. 2005).

Genes repressed by foxo, rapamycin, or dilp3 ablation are rover-biased in fed flies, while genes increased by insulin/Tor knockdown are rover-biased in food-deprived flies (Figure 4.7 a-c). This is an example of the more general trend illustrated in Figure 4.5, where the change between fed and food deprived flies is larger in rovers than in sitters for behaviours (89%), metabolites (84%), genes (77%), and gene groups (77%). In mammals, a reduced physiological response to food is a sign of insulin resistance; whether this is true of sitters awaits further testing.

Our data provides direct evidence through genetic crosses and considerable correlational information through patterns of insulin gene expression and meta-analysis that the GEI effects of foraging in fed and FD flies are mediated at least in part through interaction with the insulin/Tor signaling pathways.

225

Consequences of Lipid-Carbohydrate Allocation differences How might fatty rovers and starchy sitters differ in life-history? A number of life-history and ecological parameters have been shown to be related to lipid or carbohydrate reserves in flies, including flight capacity, starvation and desiccation resistance.

Diptera in general and fruit flies in particular are dependent on glycogen reserves and hemolymph sugars to fuel flight muscles (Wigglesworth 1949; Rowley 1970; Nayar and Vanhande.E 1971). Glycogen phosphorylase, which has strong GEI in rovers and sitters, is a rate limiting enzyme for glycogen mobilization to support flight (Gade and Auerswald 2003) (Figure 4a). Flies selected for postponed ageing show increased flight duration, glycogen reserves and resistance to desiccation, (Graves, Toolson et al. 1992), while desiccation-selected flies show higher glycogen levels (Gibbs, Chippindale et al. 1997). Glycogen, desiccation resistance, longevity and stress resistance may form a cluster of correlated traits in flies (Hoffmann and Parsons 1989).

Lipid content of adult flies correlates with starvation resistance (Hoffmann and Harshman 1999); among lifespan-selected and other lines, starvation resistance was correlated with lipid content and not glycogen (Graves, Toolson et al. 1992). This correlation extends to sibling species D. simulans (Ballard, Melvin et al. 2008). In a cricket species where lipids can be used to support flight, a tradeoff between lipid reserves for flight and for egg production has been reported (Zhao and Zera 2002; Zera and Zhao 2003).

The rover-sitter system, with its dichotomous Y-allocation (Houle 1991) of energy stores to lipids and carbohydrates, may therefore be useful for studying single-gene influences on traits with costs and benefits associated with energy use and storage including flight capacity, desiccation and starvation resistance.

226

GEI due to foraging and neuromuscular function The patterns of GEI and I between foraging and food (fed vs. FD) is well explained for genes whose primary function is in anabolic or catabolic pathways. However, there are many more genes with significant for GEI. An important set of such genes is involved in nerve and/or muscle function (Supplementary Table 4.3). PKG affects synaptic plasticity in mammals (Wang and Robinson 1997; Liu, Rao et al. 2003; Feil, Hofmann et al. 2005; Ota, Pierre et al. 2008) and learning and memory in flies (Kaun, Hendel et al. 2007; Mery, Belay et al. 2007).

The possibility that PKG may cause GEI through its role in regulating ion homeostasis in nerves and muscles deserves further examination. PRKG1, the mammalian homolog of for, regulates calcium and potassium fluxes in smooth muscle relaxation where it is associated with the myosin phosphatase complex, Ca++ATPases, and potassium ion channels (Dostmann, Tegge et al. 2002). We find a cluster of gene groups with I < 0 associated with muscle and actin cytoskeleton, including genes such as wupA (troponin-I) and Prm (paramyosin). These are some of the genes whose expression is most correlated with for in a coexpression analysis across humans, flies, worms, and yeast (Stuart, Segal et al. 2003).

Calcium/potassium levels are important in synaptic function and plasticity, and mutations in potassium channel genes affect habituation in the giant-fiber axon escape reflex in flies (Engel and Wu 1996; Engel and Wu 1998). Habituation of the giant-fiber escape reflex differs in adult rovers and sitters (Engel, Xie et al. 2000). Rover-sitter differences in PKG are also associated with different voltagedependent K+ currents in larval neuromuscular junctions, along with differences in neuronal excitability, neurotransmitter release, and synaptic transmission (Renger, Yao et al. 1999). Rover-sitter differences in neural thermotolerance arise from differences in the regulation of K+ channel activity via a circuit involving PKG, PP2A, and ion channels (Dawson-Scully, Armstrong et al. 2007). Thus our demonstration of rover-sitter differences in gene expression of genes involved in neurotransmitter release, postsynaptic membranes, and calcium- and potassium-channels supports previous studies. it will be important to determine whether foraging interacts epistatically with other genes influencing K+ currents in neurons and muscles. It is also of interest to investigate whether the metabolic effects of allelic variation in for are independent of, or are tied to PKG’s effects on ion homeostasis and neural function.

227

Magnitude of foraging GEI compared to effects of other genes Our study used only a few strains of flies and thus does not speak to the importance of formediated effects on the genome in natural populations. However, we are able to consider allelic affects at the for locus relative to the genetic background affects in a principal components analysis which identifies genetic background and food (fed vs. FD) as the most important factors, followed by the interaction of background and food, for genotype main effects, and for interaction effects. Using the Storey-Tibshirani method to estimate the true proportion of differentially expressed genes πalt shows that for GEI affects 57% of the highest expression genes. We also found that the effect of the natural sitter background was to intensify gene expression contrasts with rovers but to reduce behavioural contrasts. Thus, an important future step is to quantify the relative importance and roles of for and other genes in a wider variety of natural genetic backgrounds.

Our results also speak to evolutionary questions about pleiotropy, epistasis, and plasticity. Pleiotropic genes may affect few traits when redundancy or compensations in gene networks buffer the effects of mutations (Featherstone and Broadie 2002), while mutations in other genes produce large changes (van Swinderen and Greenspan 2005). The number of traits influenced by a gene follows a power law, with a few genes having widespread affects (Featherstone and Broadie 2002; Promislow 2004). It has been proposed that the use of naturally occurring alleles or mild mutations is more relevant to studies of epistasis and network stability than the more common use of knockouts or severe loss of function mutations (Greenspan 1997; Greenspan 2001; Benfey and Mitchell-Olds 2008). The question of whether some genes can increase phenotypic plasticity and thus whether selection can act to increase or decrease plasticity has been the subject of much debate (Sarkar 2004). In for we have an example of a gene with naturally occurring alleles maintained in a stable polymorphism in the wild (Fitzpatrick, Feder et al. 2007). We demonstrate that in adult flies for interacts pervasively with food, producing pleiotropic GEI in behaviours, lipids and carbohydrates, and gene expression. Our quantitative plasticity measure reveals that for x food GEI is often due to rovers having significantly higher food-related plasticity than sitters. We show that for interacts with genes of the insulin signaling pathway to produce some of these effects. The foraging gene may thus provide a suitable context for resolving some of these questions relating to phenotypic plasticity and selection.

228

Methods Strains and rearing. The following allelic variants of the chromosome-2 foraging (for gene were used in this study: the R

for , (natural rover), fors (natural sitter) variants and and fors2 (a sitter mutant strain made on a forR genetic background) (de Belle, Hilliker et al. 1989; Pereira and Sokolowski 1993). All strains share common isogenic third chromosomes from forR and common X chromosomes. Genetic variation on the small fourth and Y chromosomes was not controlled. For tests of epistasis we used three null mutants of genes in the insulin signalling pathway (see below).

Flies were reared on medium b for behavioural assays, gene array and metabolite experiments(Supp. Table 4.1a) (Toma, White et al. 2002). For additional behavioural assays reported in Supplementary Table 4.1b-c flies were raised in media described in Supplementary Table 4.1a and 4.1c. Flies were raised in 40 mL plastic vials containing 10 mL food medium in 12/12 h light/dark cycle (lights on 08:00), 25±1°C, 70±5% relative humidity (standard conditions).

Fly rearing and behavior testing Flies were collected 0-2 days post eclosion, separated under light CO2 anaesthesia, then reared in groups of 25-30 for 4 days. 12-13 males and 12-13 females were used in each rearing group. Adult rearing was done under standard conditions as described above. Flies were transferred to test media the night before behavior tests. Test media consisted of 10 mL of food medium (Fed), or 10 mL 1% agar for food deprivation (FD) tests in vials. Flies were tested in the morning (9-12 a.m.) after 16-18 hours under Fed or FD conditions.

A plexiglass maze was used for the food-leaving assay; the maze is as described (Britton, Lockwood et al. 2002; Toma, White et al. 2002) and is shown in Supplementary Figure 4.1. Each morning mazes were conditioned with one sample of 25 natural sitter flies and placed horizontally on a light table with a uniform light intensity of 1000 lumens. Flies were placed in a 10x75 mm borosilicate glass tube (the “sugar entry tube”) containing 0.5 mL 0.25M sucrose in 1% agar for 15 min prior to test. At start of testing, the sugar tube with flies is placed in the entry port of the maze. Empty glass collection tubes are placed in the 9 exit ports of the maze. After 3 min, flies in collection tubes at exit ports are counted, as are flies remaining in the sugar tube. Food-leaving score is (flies in collection tubes)/(total flies). All treatment conditions were tested on at least 3 different test days.

229

Quantitative complementation crosses To test whether the rover and sitter for alleles interact epistatically with null alleles of the three genes involved in the insulin signaling pathway we used a form of quantitative complementation, a method of complementation developed for testing quantitative effects in this case gene interactions (Mackay 2001). We asked if one copy of a mutant allele of a gene involved in insulin signaling, in the presence of each of the for alleles, changes food-leaving behaviour. Crosses were made between rover (forR), natural sitter (fors), and mutant sitter (fors2) strains and balanced loss of function mutants in three insulin signaling pathway genes: (a) InR, the insulin receptor (mutant allele: InR93dj-4 (Oldham, Stocker et al. 2002)); (b) Pi3K92E/Dp110, the phosphatidylinositol 3-kinase catalytic subunit (Dp110B (Williams, Busto et al. 2006)); and (c) foxo, (foxo21 (Junger, Rintelen et al. 2003)). Strains carrying mutations in the insulin signaling pathway were: (a) In(3R)GC25, InR93Dj-4/TM3, Sb1 (Bloomington stock 9554) (Clancy, Gems et al. 2001); (b) yw;P[ry+,gH], Dp110B/TM3,Ser,y+(C) (Weinkove, Neufeld et al. 1999) ; (c) foxo21/TM6C (Junger, Rintelen et al. 2003) All mutants of the insulin signaling pathway were maintained heterozygous with balancer chromosomes which did not carry mutations in these insulin signaling genes and were sitter. Epistasis is identified as a two-way statistical interaction between the variant (rover or sitter) and test genotype (mutant or control) (Gibson and Dworkin 2004) (see below). The balancer heterozygotes control for effects of natural variation in the genetic background that can result in increases or decreases in food leaving.

Metabolite Analysis Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FTICR MS) was used to analyze homogenized fly heads (equal numbers of males and females) from 5-7 day post eclosion forR and fors2 strains harvested in the morning. Food Deprived (FD) flies had been restricted to water in agar 12 hours before collection. Samples were taken in triplicate. Values shown are Signal to Noise (S/N) ratios. Each sample was analysed as described (Zulak, Cornish et al. 2007) using the DiscovaMetrics (Gray and Heath 2005) package producing parent ion molecular weights accurate to within 0.0005 daltons; compound identifications were cross-checked against Kegg Ligand (Kanehisa, Goto et al. 2006) and Metlin (Smith, O'Maille et al. 2005) databases. FTICR MS has maximum sensitivity to metabolites in the 100-1000 Dalton range, accuracy of 0.0001 Dalton, and uses six different buffer/ionization modes (Supplementary Table 4.2) each detecting different classes of metabolites. For instance, compounds such as sugars and phosphates are detected in mode 1102 using a polar buffer and negative ion electrospray, while mode 1203 uses a nonpolar buffer and positive atmospheric pressure chemical ionization mode to detect compounds such as triacylglycerols.

230

Whole body lipid and carbohydrate analysis was performed separately on males and females of 57 day post-eclosion forR, fors, and fors2 strains. For both lipid and carbohydrate analyses, results were standardized against dry weight. For lipids, the ether extraction method was used as described (Clark 1989). In brief, flies were frozen in liquid nitrogen, then weighed to the nearest 0.01 mg in groups of 5-10 flies on a Mettler Toledo XS205 balance.. Flies were then dried at 60°C for 24 hours and reweighed. Lipids were then extracted in 1 mL of ether for 24 hours, after which ether was decanted and flies were dried at 60°C for 24 hours and weighed. Total carbohydrate levels were determined using amyloglucosidase digestion followed by spectrophotometric determination of total glucose using NAD to NADH reduction (Kunst and et al. 1984). Sigma kit GAHK20. Briefly, hexokinase catalyzes phosphorylation of glucose in the presence of ATP to Glucose-6-phosphate (G6P), which is then oxidized to 6-phospho-gluconate in the presence of oxidized nicotinamide adenine dinucleotide (NAD) in a reaction catalyzed by glucose-6phosphate dehydrogenase (G6PDH). During this oxidation, an equimolar amount of NAD is reduced to NADH. The consequent increase in absorbance at 340 nm is directly proportional to glucose concentration. Protein was measured using the bicinchonic acid (BCA) method. Total energy content was calculated based on ratios of 9:4:4 Kcal/gm for fats:carbs:protein.

Microarray analysis Affymetrix Drosophila Genome 1.0 cDNA microarrays were used to evaluate effect of foraging genotype and feeding state on transcript levels in adult heads. Flies homozygous for each of the 3 alleles forR, fors, and fors2 were raised to 5-7 days post-eclosion, and given Food or FD treatments as described above. Samples of flies (equal numbers of males and females) were frozen in liquid nitrogen, and heads were separated by sieving. RNA was extracted as described (Dierick and Greenspan 2006). Triplicate RNA samples for each treatment were hybridized to Drosophila Genome 1.0 microarrays, for a total of 18 arrays. N=3 within each treatment. Expression levels produced by MAS 5.0 were normalized by quantile normalization (Bolstad, Irizarry et al. 2003) of log2 transformed data. Full MIAME information and expression set data is filed as GEO accession GSE14371. Pathway analysis and gene ANOVA were performed as described in Statistical methods. Analysis of variance was used to detect significant GEI for individual genes after False Discovery Rate correction for multiple testing (Storey and Tibshirani 2003).

qRTPCR analysis Levels of expression of 2 genes (Treh, CG10924) were confirmed by quantitative reverse transcriptase polymerase chain reaction analysis (Supplementary Figure 3). Heads of male and female flies forR and fors2 strains, raised as described for microarray analysis, were frozen in liquid nitrogen. RNA was extracted using the Trizol method (15596-018, Life Technologies) and further purified using the Qiagen RNeasy kit (74106, Qiagen). The amount of RNA in each sample was determined using a Nanodrop spectrophotometer (ND-1000) and sample quality verified using 260/280 micron absorbance ratios.

231

Statistical Methods Scores of each trait are analysed with two-way Analysis of Variance (ANOVA) to determine whether significant GEI exists (see detailed procedures below). The Storey-Tibshirani False Discovery Rate (FDR) (Storey and Tibshirani 2003) is used for multiple testing correction and estimation of π0 using the qvalue package as implemented by Storey, with default parameters. Thus ANOVA p values have been replaced by FDR q values, and q < 0.05 is deemed significant.

RNS measures which strain has higher plasticity and is defined as RNS=(|rover change| - |sitter change|)/C (for log2 transformed data C=1, else C=mean of all treatments). That is, RNS compares the absolute magnitude of changes in response to food and is positive when rovers change more than sitters.

For ANOVA of gene expression data with two sitter strains and one rover, a modified general linear model design matrix was used (Supplementary Methods) to ensure unbiased estimation. Briefly, factors RS (rover or sitter), FD (fed or food deprived), and BG (genetic background, 2 levels, 1 for rovers and mutant sitters, another for natural sitters) were analysed including main effects and the interactions RS x FD and BG x FD. A reduced model omitting the interaction BG x FD was also fitted. For each gene, the first and second models were compared using Schwartz’s Bayesian Information Criterion (BIC) (Schwarz 1978) to determine whether to report full or reduced model results. Thus if interaction of background and food was significant (as determined by BIC) we reported statistics from the full model, else from the reduced model. FDR correction was then applied to p-values from the selected model.

For group-wise ANOVA analysis of groups of metabolites or genes, a linear model as above, with the addition of a factor G with one level for each gene or compound was used; this is similar to adjusting each gene or compound to have a mean of zero, but accounts more conservatively for lost degrees of freedom due to the adjustment. Only transformed variables with approximately equal variances are used in group-wise ANOVAs. Microarray data is log2 transformed, then subjected to quantile normalization and a variance-equalizing monotonic transform. After these steps variances for the top-expressing 60% of genes were approximately equal and data was normally distributed. For group-wise ANOVA of metabolites, log2 data was used. Data was tested with a covariate of molecular weight (MW). If MW or its interactions with food and genotype were significant, the ANCOVA with MW is reported; otherwise group-wise ANOVA results are reported.

For ANOVA of behavioural experiments where behaviours may vary from day to day (Day effect), Day was added as a random factor to the ANOVA described above for single genes, and significance of this mixed-model was determined by F-tests of fixed factor terms to their interactions with Day. See Supplementary Tables 4.1 and 4.5 for examples.

232

In the quantitative complementation crosses in Supplementary Table 4.5, the interaction of for with a factor representing the presence or absence of the mutant insulin gene is tested. That is, we test for epistasis rather than GEI. The ANOVA analysis is identical in format to that just described, with the factor representing presence/absence of insulin mutant replacing the food factor.

233

Supplementary information

Supplementary Figure S4.1. Behaviour testing apparatus. A plastic maze originally used for geotaxis experiments (Britton, Lockwood et al. 2002; Toma, White et al. 2002) is placed horizontal on a light table adjusted to produce 1000 lumens illumination. Darker areas on photograph edges are due to camera contrast adjustment; actual illumination is even over maze surface. The entry tube contains agar with 0.25M sucrose (Methods). 24-26 flies are placed in this tube 15 minutes before entry to the maze. 9 empty (no agar or sugar) collection tubes block exit points from maze. At time 0 the entry tube is placed in the maze entry. Numbers of flies in collection tubes is counted every minute until termination of run at 3 minutes. Experiments were conducted in a darkened room maintained at 25C and humidified to 60%RH.

234

Supplementary Table T4.1. Analysis of variance of behaviour and four food media (a) Food medium compositions A standard water, agar, and minerals base was used in all foods (1,000 ml H2O, 10 g agar, 8 g C4H4KNaO6, 1g KH2PO4, 0.5g NaCl,0.5 g MgCl2, 0.5g CaCl2, 0.5g Fe2(SO4)3). The manufacturer’s reported analysis of yeast per 100 g dry yeast was 50 g protein, 5.3 g fat, 33 g carbohydrate. Media b,c,e contained a commercial dark corn syrup for which the manufacturer reported 125 g total carbohydrates/ 100 ml from an unspecified mixture of glucose, fructose, and blackstrap molasses. Medium b was used for flies in all main figures, including behaviour, metabolite, gene array, and quantitative complementation cross tests. Other media were used to investigate sensitivity of behaviour to food (see below).

Medium a b c d e f

sucrose/L 50 g 15 g 87.5 g 100 g 116.3 g 50g

syrup/L 0 30 ml 30 ml 0 15 ml 0

yeast/L 25 g 35 g 35 g 50 g 17.5 g 0

235

(b) Mixed model ANOVA for food leaving behaviours. See table for food (b) below for analysis of Figure 1 behaviour data. Data for behaviour tests on 3 other media (a,c,d) are also given here. Data for each food was first tested with a full mixed model ANOVA including terms for genetic background differences and their interaction with food, as an approximate test of whether the genetic background difference between the strains contributes to GEI (See Statistical Methods). Day of testing was the random factor, genotype and food presence/absence were fixed factors. For none of the four foods was there significant GEI between background and food (p>0.5 for all four tests; data not shown). Hence (see Statistical Methods) the results of a reduced mixed ANOVA model with fixed factors rover/sitter type (labeled for), food presence/absence (label: food), genetic background (label: BG) and random factor Date of testing are presented (see Statistical Methods for detailed model). Significance of fixed terms is tested against their interaction with the random factor Date.

Food (a) for food BG for x food Date Residual

df 1 1 1 1 4 22

MS 0.179 0.377 0.024 0.168 0.024 0.0148

df x Date 4 4 4 4

MS x Date 0.0042 0.0276 0.0079 0.0131

F 42.37 13.64 3.03 12.82 1.62

Food medium b is used for main figures, arrays, metabolites, and crosses. Food (b) df MS df x Date MS x Date F for 1 0.019 6 0.0028 6.63 food 1 0.258 6 0.0081 31.69 BG 1 0.000 6 0.0074 0.002 1 0.060 6 0.0063 9.65 for x food Date 6 0.020 2.44 Residual 21 0.0080

p 0.0029 0.021 0.157 0.023 0.203

p 0.042 0.0013 0.968 0.021 0.060

Food (c) for food BG for x food Date Residual

df 1 1 1 1 5 10

MS 0.001 0.102 0.027 0.134 0.039 0.0050

df x Date 5 5 5 5

MS x Date 0.0042 0.0276 0.0079 0.0131

F 0.16 12.99 5.43 18.80 7.81

p 0.709 0.023 0.145 0.023 0.003

Food (d) for food BG for x food Date Residual

df 1 1 1 1 4 22

MS 0.207 0.012 0.001 0.092 0.011 0.0109

df x Date 4 4 4 4

MS x Date 0.0172 0.0154 0.0058 0.0100

F 12.06 0.78 0.25 9.14 1.03

p 0.025 0.427 0.647 0.039 0.409

236

(c) Measured I values in 9 nutritional environments We performed behaviour tests on flies raised in 9 different nutritional environments (NE). NE is defined as a specific combination of larval food medium and adult food medium (food media in 1a, above; NE combinations below). We held the effects of larval nutritional environment constant by raising from egg to eclosion on medium (d), and then from eclosion to time of testing on one of medium a-d to produce NE 1-4, below. Behaviour ANOVA for these NE are given above (1b) Additional experiments for NE 5-9 were performed with smaller numbers of replicates so we do not report statistical analyses for these. However we did calculate the value I from arc-sine transformed scores for each environment for rovers versus mutant sitters (R-s2) and rovers versus natural sitters (R-s). The null hypothesis that I=0 over all 9 nutritional environments is rejected by a signs test, since all I values were positive.

Nutr. env. Larval food 1 d 2 d 3 d 4 d 5 d 6 d 7 a 8 b 9 c

Adult food a b c d e d,f† a b c

I (R-s2)

I (R-s)

0.155 0.288 0.298 0.171 0.009 0.242 0.268 0.058 0.029

0.138 0.199 0.205 0.210 0.274 0.181 0.197 0.030 0.029

† NE 6 kept flies on medium d until the night before testing. Flies in the FD treatment had access to water only as usual while Fed flies had access to medium f (sucrose agar).

237

Supplementary Table T4.2. FTICR MS metabolite data (a) Total compound levels per group (b) Group-level analysis. Data for heads of rovers and mutant sitters, fed and food deprived. (a) ANOVA of total compound levels

Group PS TAG

for F 405.0 5.20

for p food F -8 164.1 3.88·10 0.052 73.1

food p 1.31·10-6 2.70·10-5

Int F 39.2 5.59

Int p RNS -4 2.43·10 -0.34 0.046 0.36

Note: in all cases degrees of freedom (d.f.) for F statistics are 1,8. (b) Group-level ANOVA Polysaccharides (n=5 compounds). d.f. for all F values is 1,48

factor for food for x food for x MW food x MW for x food x MW

F 19.07 7.33 0.25 4.41 4.24 4.77

p 0.00005 0.0088 0.615 0.040 0.044 0.033

Triaclyglycerols (n=13 compounds). d.f. for all F values is 1,136

factor for food for x food for x MW food x MW for x food x MW

F 44.99 408.47 66.58 19.49 3.45 0.01

p 3.91·10-10