THE DEVELOPMENTAL AND MOLECULAR-GENETIC MECHANISMS OF SEXUAL SIZE AND SHAPE DIMORPHISM IN DROSOPHILA MELANOGASTER  By  Nicholas D. Testa                A DISSERTATION  Submitted to Michigan State University in partial fulfillment of the requirements  for the degree of  Integrative Biology Ð Doctor of Philosophy Ecology, Evolutionary Biology and Behavior Ð Dual Major  2016  ABSTRACT  THE DEVELOPMENTAL AND MOLECULAR-GENETIC MECHANISMS OF SEXUAL SIZE AND SHAPE DIMORPHISM IN DROSOPHILA MELANOGASTER  By  Nicholas D. Testa  Much of the morphological diversity in nature is due to the sexual dimorphism in both size and shape. While sexual size dimorphism (SSD) is exceedingly common and can be rather extreme, size differences between the sexes are often quite mild. The fact that males and females typically share most of a genome raises important questions about the ubiquity of SSD, namely about how males and females achieve such different phenotypes given a mostly shared genome. Historically, explanations to explain this trend have focused more heavily on evolutionary mechanisms than proximate and/or genetic mechanisms.   Using the fruit fly, Drosophila melanogaster, as a model organism I explore the developmental, physiological, and genetic mechanisms that regulate both sexual size and shape dimorphism. Like most insects, Drosophila females are larger than males and differ in many key morphological features (such as shape). I will describe a series of experiments aimed to successively uncover the mechanistic underpinnings of sexually dimorphic growth that regulate sex-specific size and shape.  To investigate the mechanisms of SSD, I first compared the relative contributions of proximate, developmental mechanisms to size in both males and females. Previous research has also implicated critical size, a hormone-mediated physiological checkpoint, in helping to regulate overall body size in holometabolous insects. Here I demonstrated that, while males and females had equal larval growth durations, females reached their critical size at a larger weight and also grew faster in their final larval instar than males. The resulting SSD was further attenuated by an increase in female weight loss in the period intervening the achievement of peak larval mass and pupariation. Next, I demonstrated that mutants of the IIS pathway (specifically InR) completely eliminated SSD. Since condition-dependence is a common explanation for the evolution of SSD, the next step was to investigate the effects of nutrition on SSD. After demonstrating a negative relationship between SSD and nutritional quality, I investigated the effects of several candidate genetic pathways, including nutrient-sensitive growth pathways, on whole body SSD. I then investigated the sex-limited effects of mutants in similar pathways on sexual size and shape dimorphism where effects are unknown. Finally, I examined the ability of my own candidate pathways to influence both sexual size dimorphism (SSD) and sexual shape dimorphism (SShD).   Collectively these studies bring our understanding of the proximate mechanisms that regulate SSD and SShD to a new and more profound level. While much is known of the selective pressures that generate sexual dimorphism, we know very little about the specific genetic targets of these pressures, and how these targets may facilitate or hamper the evolution of sexual dimorphism. My study goes some way to filling this conspicuous gap.      !"#!ACKNOWLEDGEMENTS   First, I would like to thank my committee members, Dr. David Arnosti, Dr. Jenny Boughman, and Dr. Zach Huang, for their continued support and suggestions in shaping my research to tackle big biological questions. I would like to thank the Graduate school, the Department of Zoology, the Program in Ecology, Evolutionary Biology and Behavior (EEBB) and the BEACON center for providing financial support and diverse intellectual stimulation through seminars, workshops, and informal gatherings. In particular I would like to thank the staff members who have done an extraordinary job of helping me to navigate the requirements and paperwork including Lisa Craft, Debbie Mills, Sue Platte, and Mike Presocki, (Zoology); Pat Resler (EEBB); Connie James and Danielle Whittaker (BEACON). I would also like to thank those who have helped me to develop my teaching through the years: Sonya Lawrence, Dr. Sue Hill, Dr. Will Kopachik, Dr. Austin Dreyer, Dr. Diana Bello-Deocampo, and all the countless graduate and undergraduate TAs IÕve had the honor to teach with. Lastly, I would like to extend a special thanks to Sue Hill, for her demonstrating how to be an outstanding educator in every sense of the word and how to balance work with life outside of the lab; thank you so much.  I would like to thanks Tom Getty for being such a helpful and supportive department chair. Without your help, I would likely have Òfallen through the cracksÓ during my Ògraduate student orphanÓ days. I really appreciate everything youÕve done.  !#!I would like to write a special Òthank youÓ to all of the friends IÕve made throughout my many years in graduate school at MSU, including: Austin Dreyer, Eli Swanson, Carlos Anderson, Aaron Florn, Emily Weigel, Sara Garnett, Michael DeNieu, William Pitchers, Amanda Charbonneau, Mauricio Losilla, Anne Sonnenschein, Abhijna Parigi, Sarah Marzec, Sudarshan Chari, and anyone else that IÕve forgotten to mention. WeÕve had a lot of good times playing video games, board games, eating, drinking (and making) beers, complaining about research at WoodyÕs, etc! Without all of you I would surely have gone insane! and finished on time. Thanks for that.  I would also like to extend a special thank you to those friends/postdocs, without whom I would not have made it this far. Thank you Shampa Ghosh, R. Craig Stillwell, Annat Haber, and Will Pitchers for all of your guidance and assistance throughout graduate school. Your collective guidance helped me to be more organized and thoughtful about my experiments and analysis.   Next, I would like to thank my advisors, Dr. Alexander Shingleton and Dr. Ian Dworkin, for giving me the opportunity to work and develop an exciting research topic. Without either of your labs, I would not have been able to develop the computation, statistical, and experimental skills that I have. Through the trials and tribulations of both of your successive absences, I feel like I gained a level of maturity and independence I may not have otherwise achieved. In my earliest days at MSU I can remember the joint lab meetings with Alex and IanÕs labs and I would never have guessed that I would be a part of both (especially during the !#"!legendary arguments that typically closed out our meetings). All kidding aside, my experience in both labs has helped to mold me into the well-rounded and independent scientist that I am today. Thank you Alex for being taking a chance on me and bringing me into your lab. And thank you Ian for adopting this orphaned graduate student and treating me like one of your own. If I am ever half the researcher that either of you are, I will consider myself lucky.   Finally, I offer my deepest gratitude to my family. Completing a PhD takes a tremendous amount of time, and often, time that is unplanned for. So thank you to my family for understanding the years of 80+ hour workweeks (often including evenings and weekends), the missed time over holidays and vacations, the constant complaining, the constant conversation hijacking, and the phone calls cut short. I would also like to thank my siblings Steve, Dave, and AnnaMarie for their continued support, as well as the support from my brother- and sister-in-law Denny (and Ellie) and Rachel. And to my parents Annette and Dan (and mother-in-laws Donna and Lisa), none of this would have been possible without your continued support. I gave you every reason to give up on me early on, but you always stood by me. Thank you all.   Last, but certainly not least, I would like thank my kind and understanding wife, Darby, and our beautiful daughter Freya. I could always count on Freya to bring a big warm smile to my face, even at my most stressed. And Darby, without your unwavering support I most certainly would not have made it half as far as I have. Every time IÕve fallen (metaphorically), you were there to pick me up, brush !#""!me off, and encourage me to keep going. Every time I got so stressed I couldnÕt think, you were there to talk me off the ledge (also metaphorically). If not for your encouragement and inspiration, I may not have even made it through my undergraduate degree, let alone apply to (and finish) graduate school.   !To you, Darby, I owe everything    !#"""!TABLE OF CONTENTS  !LIST OF TABLES .................................................................................................. xi  LIST OF FIGURES .............................................................................................. xii  CHAPTER 1: How to evolve SSD on the fly .......................................................... 1 Introduction ........................................................................................................ 2 Introduction to Sexual Dimorphism ................................................................. 2 Differences in genome composition between sex .......................................... 3 The genome as an impediment to Sexual Dimorphism .................................. 4 Natural variation in SSD and SShD (despite shared genome) ....................... 7 Ultimate mechanisms of Sexual (Size) Dimorphism .......................................... 9 Sexual selection and sexual conflict ............................................................. 10 Fecundity selection ....................................................................................... 13 Countervailing Selection ............................................................................... 13 Proximate mechanisms of Sexual Dimorphism ................................................ 14 Dearth of knowledge on proximate mechanisms of SSD ................................. 17 Proximate mechanisms of overall body size and SSD ..................................... 18 Genetic mechanisms of size control and SSD ................................................. 20 How are size and shape related? A Primer on Shape Analysis ....................... 27  CHAPTER 2: Testa, N. D., Ghosh, S. M. & Shingleton, A. W. Sex-Specific Weight Loss Mediates Sexual Size Dimorphism in Drosophila melanogaster. PLoS One 8, e58936 (2013) ................................................................................ 33 Introduction ...................................................................................................... 34 Materials and Methods ..................................................................................... 36 Fly Strains and Maintenance ........................................................................ 36 Critical Size ................................................................................................... 37 Growth rate ................................................................................................... 37 Statistical Analysis ........................................................................................ 38 Results ............................................................................................................. 39 Discussion ........................................................................................................ 42  CHAPTER 3: The Genetic Regulation of Whole Body Sexual Size Dimorphism in Drosophila melanogaster ..................................................................................... 47 Introduction ...................................................................................................... 48 Materials and Methods ..................................................................................... 52 Fly strains and Maintenance ......................................................................... 52 Nutritional SSD Assay .................................................................................. 55 Determining candidate genes by pathway .................................................... 56 Titrating gene expression ............................................................................. 58 Experimental Food Preparation .................................................................... 60 Experimental Fly Crosses ............................................................................. 60 !"$!Analysis of SSD ............................................................................................ 61 Results ............................................................................................................. 63 Nutrient Sensitivity of SSD ........................................................................... 63 Titrating gene expression: IIS pathway ........................................................ 63 Titrating gene expression: TOR pathway ..................................................... 68 Titrating gene expression: Ras pathway ...................................................... 68 Titrating gene expression: Hh pathway ........................................................ 72 Titrating gene expression: Sex determination pathway (SDP) ..................... 72 Discussion ........................................................................................................ 75 Effect of Nutrition and Nutrient-Dependent Pathways on SSD ..................... 75 Overexpression of nutrient-sensing growth pathways reduces body size .... 78 Alternative interpretations of data ................................................................. 81 Hh signaling and its link to the SDP ............................................................. 83 At what level in the SDP are sex-specific body sizes determined? .............. 84  CHAPTER 4: Testa, N. D. & Dworkin, I. The sex-limited effects of mutations in the EGFR and TGF-" signaling pathways on shape and size sexual dimorphism and allometry in the Drosophila wing. Dev. Genes Evol. (2016). doi:10.1007/s00427-016-0534-7 .......................................................................... 86 Introduction ...................................................................................................... 87 Materials and Methods ..................................................................................... 90 Provenance of Samples ............................................................................... 90 Analysis of Sexual Size Dimorphism (SSD) ................................................. 92 Analysis of Sexual Shape Dimorphism (SShD) ............................................ 93 Vector Correlations ....................................................................................... 95 Statistical Analysis ........................................................................................ 96 Results ............................................................................................................. 96 Despite many mutations having substantial effects on overall shape, a relatively small number influence SSD and SShD ........................................ 99 Mutations do not substantially alter directions of SShD, nor patterns of allometry. .................................................................................................... 105 Discussion ...................................................................................................... 106 The importance of multiple independent control lineages .......................... 108 Rare sex-limited effects on wing form among mutations in EGFR and TGF-" signaling ..................................................................................................... 109 Disentangling mutant-phenotype relationships ........................................... 112 Reassessing the assumption of common allometry ................................... 114  CHAPTER 5: The Contribution of Candidate Genes to sexual size and shape dimorphism ........................................................................................................ 118 Introduction .................................................................................................... 119 Materials and Methods ................................................................................... 125 Fly strains and Maintenance ....................................................................... 125 Modulating gene expression using the Gal4 sensitivity to temperature ..... 126 Titrating gene expression with GeneSwitch ............................................... 127 Experimental Fly Crosses ........................................................................... 127 !$!Imaging and data acquisition ...................................................................... 130 Morphometric Analysis ............................................................................... 131 Analysis of SSD and SShD ........................................................................ 131 Results ........................................................................................................... 132 Discussion ...................................................................................................... 140 The case for removing the assumption of common allometry .................... 143 Reasons for cautious interpretation ............................................................ 145 Consensus Among Crosses in Size/Shape Space: .................................... 147 When are sex-specific shape and size determined? .................................. 149 Wing SSD vs Whole Body SSD: Insights into Regulation of Allometry ...... 151  CHAPTER 6: Conclusion and future directions ................................................. 153 Summary of Chapters .................................................................................... 154 Chapter 1 .................................................................................................... 154 Chapter 2 .................................................................................................... 155 Chapter 3 .................................................................................................... 155 Chapter 4 .................................................................................................... 156 Chapter 5 .................................................................................................... 156 Future directions to strengthen our arguments .............................................. 157  APPENDIX ......................................................................................................... 159  REFERENCES .................................................................................................. 169 !   !$"!LIST OF TABLES   Table 3.1. Fly Stocks used in Chapters 3 and 5. .......................................... 53  Table 3.2. Estimates for SSD and corresponding confidence intervals for all treatment crosses ........................................................................................... 66   Table 4.1. Summary table of significant effects by background among mutants ............................................................................................................ 101  Table 5.1. List of wings displaying aberrant phenotype and their relative frequencies ...................................................................................................... 128  Table A1a. Significance Table (raw p values) for effects on wing shape where G= genotype, S= sex, B= background and Cs= (centroid) size ...... 160  Table A1b. Sample size of treatments, where M= male, F= female, w= wild-type, m= mutant, O= Oregon-R background, S= Samarkand background   .......................................................................................................................... 164  Table A2. Significance Table (raw p values) for effects on wing shape after the effects of allometry are removed where G= genotype, S= sex, B= background and Cs= (centroid) size ............................................................. 166    !$""!LIST OF FIGURES   Figure 1.1. The Sex Determination Pathway in Drosophila melanogaster  .......................................................................................................................... 16  Figure 1.2. Nutrient-Sensitive Growth Pathways in Drosophila melanogaster .......................................................................................................................... 25  Figure 1.3. Hh signaling pathway in Drosophila melanogaster .................. 26  Figure 2.1. Complete growth profile by sex for Drosophila melanogaster  .......................................................................................................................... 40  Figure 2.2. SSD is lost in insulin-signaling mutants ................................... 46  Figure 3.1. Nutritional Quality vs Weight for both male (blue) and female (red) Sam flies ................................................................................................. 64  Figure 3.2.  Reaction norm for SSD (above) and body size in males and females (below) in sex determination (SD) pathway manipulations .......... 65  Figure 3.3. Reaction norm for SSD (above) and body size in males and females (below) in Insulin and Insulin-like Signaling (IIS) pathway ........... 69  Figure 3.4. Reaction norm for SSD (above) and body size in males and females (below) in TOR signaling pathway manipulations ......................... 70  Figure 3.5. Reaction norm for SSD (above) and body size in males and females (below) in Ras signaling pathway manipulations.......................... 71  Figure 3.6. Reaction norm for SSD (above) and body size in males and females (below) in Hedgehog (Hh) signaling pathway manipulations ...... 73  Figure 4.1. Natural variation in SSD and SShD for two wild type strains .. 98  Figure 4.2. Magnitude of SSD and SShD for 42 mutants in Oregon-R (left) and Samarkand (right) wild type backgrounds ............................................ 103  Figure 4.3. Vector correlations to assess similarity of direction for sexual shape dimorphism (mutant VS. wild type) by background ......................... 104  Figure 4.4. Variation in the magnitude of association between shape and size allometric coefficients among mutations in the Oregon-R (top) and Samarkand (bottom) wild type backgrounds ............................................... 107   !$"""!Figure 5.1. Sexual Shape Dimorphism in knockdowns that caused an increase in SShD ............................................................................................ 133   Figure 5.2. Pairwise comparision between SSD (first row/column), unmodified SShD (second row/column), allometry corrected SShD (third row/column), and Mean Size for males and females (fourth row/column)  .......................................................................................................................... 135  Figure 5.3. The allometry of SSD (above) and SShD (below) for feminized (left) and masculinized (right) crosses ......................................................... 136  Figure 5.4. Magnitude of SSD and SShD for knockdown and overexpression crosses ............................................................................................................ 138  Figure 5.5. Vector correlations to assess similarity of direction for sexual shape dimorphism (treatment vs. control .................................................... 141  Figure A1. Magnitude of SSD and SShD for 42 mutants in Oregon-R (left) and Samarkand (right) wild type backgrounds, after correcting for the influence of allometry (shape on size) .......................................................... 168    !%!CHAPTER 1:  How to evolve SSD on the fly  !&!Introduction Introduction to Sexual Dimorphism One of the quintessential themes of modern biology has centered on describing natural variation and answering the question of how such variation came to be. Indeed, this theme has remained prevalent since the early days of modern biology and has made a profound impact on how we study biology today. It was precisely the study of natural variation that lead both Charles Darwin and Alfred Russel Wallace to conclude that heritable variation associated with increased fitness causes some populations to produce more offspring than others.  Alongside his theory of descent with modification due to natural selection, Darwin also proposed the mechanism of sexual selection. Sexual selection, he reasoned, could be used to explain the variation between the sexes seen in many sexually reproducing species. Trait variation between sexes is known as sexual dimorphism and can affect virtually any trait.   Perhaps the most conspicuous trait an animal can possess is body size. It is often the first characteristic we notice when looking at unfamiliar species. After all, the difference between a housecat and a tiger is largely its size; mistaking one for the other could be catastrophic. While the difference in size between males and females is not commonly so stark, it is often still quite obvious. In the fruit fly, Drosophila melanogaster, females not only differ in key morphological features, but are also ~15% larger than their male counterparts (Testa et al. 2013). Total body size can be one of the simplest and most !'!intuitive ways that a species can vary by sex. Sex-specific variation in body size is called Sexual Size Dimorphism (SSD).   Differences in genome composition between sex While scientists Thomas Hunt Morgan and Theodor Boveri were debating which part of the cell contributed inheritance to their offspring (cytoplasmic elements versus the nucleus, respectively), it was Nettie Stevens and Edmund Beecher Wilson that first demonstrated that males and females differed by the existence of a sex-specific chromosome (Stevens 1905; Wilson 1905). By independently demonstrating in numerous species that females had paired (XX) chromosomal contributions to their gametes and males had unpaired (XY or X0), Stevens and Wilson revealed that sex was ÒinheritedÓ by the existence of a sex-specific chromosomal pattern (Gilbert 1987). The discovery of sex chromosomes in 1905 revolutionized the way biologists view the genetics of sex and has allowed us to further elaborate on the mechanisms that generate sexual dimorphisms.   Overall, males and females largely share the same genome. In other words, it can be said that there is typically a low genomic sexual dimorphism within a given species (Mank 2009). When two sexes differ by a unique chromosome, they are referred to as heterogametic. Regardless of how different these sex chromosome may be, however, they still only occupy a relatively small portion of the entire genome (Mank 2009). Other forms of sex determination exist, nonetheless, that make genomic differences between heterogametic sexes seem comparatively small or large. In polygenic sex determining !(!systems, sex determination can be caused by epistatic interactions between multiple genes and alleles (Moore & Roberts 2013), whereas few, if any, genetic differences exist between species whose sex is controlled environmentally (Devlin & Nagahama 2002). Any differences in gene expression therefore, often have more to do with the differential regulation of genes rather than possession of genes.   Heterogamety is an important source of genetic dimorphism between males and females. The familiar form is one in which females possess two X chromosomes and males possess an X and a Y. The XX/XY system, while common, is not the only genetic sex determination system. Most birds and lizards use ZW/ZZ as their sex chromosomes, but in this system the females are the heterogametic sex and males are homogametic (Mank 2009). Haplo-diploid species are those in which all fertilized eggs develop as females and unfertilized haploid eggs develop as males (Heimpel & de Boer 2008). Recent models suggest that rapid transitions between types of heterogamety are possible under certain conditions (Van Doorn & Kirkpatrick 2010), which may explain the staggering variation in the variety of sex chromosome types. Truly, there is a surprising amount of variation in sex chromosomes and sex-determination itself (Mank 2009; Adkins-Regan & Reeve 2014).  The genome as an impediment to Sexual Dimorphism That males and females can be so divergent in phenotype and still essentially share most of a genome is astounding. Any individual speciesÕ ability to evolve sex-specific differences in size should be hindered when both sexes share the same set of genetic !)!instructions for generating body size. Despite the fact that males and females of a given species have different sex chromosomes, most genes reside on autosomal chromosomes and the alleles are therefore not segregating preferentially to either sex. Recent studies have even demonstrated the incomplete sex-specific inheritance of certain developmental traits responsible for generating SSD; traits like increased growth rate that define female-specific size are inherited by both males and females (Stillwell & Davidowitz 2010a).   If sexual dimorphism is expected to evolve given specific selective pressures (see below), it is constrained to work within the bounds of a speciesÕ genetic architecture. A given genome must be able to evolve sex-specific phenotypes. Males and females therefore have a reduced ability to pass on sex-specific beneficial alleles to offspring of the same sex; i.e. mothers and fathers transmit alleles to both daughters and sons. Selection therefore cannot always respond to sexual conflict, whether it occurs over different traits (interlocus sexual conflict) or over different alleles of the same gene influencing the same trait (intralocus sexual conflict; see discussion below).   This has been clearly demonstrated from studies of both artificial selection and experimental evolution. For example, simply selecting for larger female body size will not necessarily increase SSD, because males also share the same genes that caused females to grow larger in the first place (specific examples discussed below). The resulting selection experiment would be expected to only increase absolute size, rather than dimorphism in size (Teuschl et al. 2007; Lande 1980), though in some cases SSD !*!may actually evolve as a correlated response (Reeve & Fairbairn 1996). Even simultaneously selecting for smaller male size and larger female size at the same time should theoretically not be enough to overcome the genetic impediments, but it often is. Indeed, unless males and females have identical genomes, their shared genome will only slow their evolution towards sex-specific trait optima.   Because males and females typically share most of their genome, they cannot always respond individually to selection pressures, especially when males and females differ in their optima for a given trait (e.g. size), both conditions cannot be satisfied; this is called intra-locus sexual conflict. Sexually antagonistic selection may cause gene frequencies within populations to fluctuate until the conflict is resolved (Cox & Calsbeek 2009), the degree to which this occurs depends on the magnitude of the inter-sex genetic correlation. Furthermore, when one sex is experimentally allowed to ÒwinÓ this genetic tug-of-war, the consequences are usually detrimental to the other (Prasad et al. 2007; Holland & Rice 1999; Rice et al. 2006). By using females that carried a pair of X-chromosomes fused at the centromere and a Y-chromosome, the authors were able to reduce the sex-specific information passed on in the female lineage; females would pass on one pair of fused X-chromosomes to their daughters and a Y-chromosome to their sons, while male Drosophila would be the ones to pass on their X-chromosome to their sons. This process effectively allowed genes on the X-chromosome to evolve based on selection pressures on the male and only the male, resulting in the evolution of male-specific fitness optima at the femaleÕs expense.   !+!Attempts to increase/decrease SSD by artificial selection are often accompanied by a correlated response in the opposite sex (Teuschl et al. 2007). These results are particularly important, because they illustrate a key concept in understanding the evolution of SSD: the genome cannot respond to selection on one sex completely independently of the other. This phenomenon has been observed in natural systems as well (Karubian & Swaddle 2001).The authors demonstrated that selection on one sex for increased or decreased body size tended to have a fully correlated response in female body size with very little countervailing selection. In order to circumvent the issues with sexual conflict predicted in previous models (Lande 1980), Reeve and Fairbairn (1996) used a more complicated selection regime. Instead of just selecting males or females with the most desirable trait value, Reeve and Fairbairn selected half-sib groups. This was achieved by allowing one male to mate with multiple females, whose offspring would all be half-siblings. The unit of selection was therefore a group of half-siblings, rather than by individual. Only by selecting on SSD within these families, rather than of unrelated individuals, the authors finally able to artificially select on SSD. Their response was strong and consistent with their respective selection regimes.   Natural variation in SSD (despite shared genome) What is most shocking is that despite the potential constraint of both sexes sharing the majority of their genome (and transmitting alleles to both sons and daughters), SSD is exceedingly common within the animal kingdom. In fact, we see this pattern across all animal groups (Fairbairn 1997a). The magnitude and direction of dimorphism, however, is often consistent among closely related groups (Fairbairn 1997a). For example, !,!among mammals and some groups of birds, males are the larger of the two sexes. Among mammals, up to 40% of species display a moderate to extreme male biased SSD of 20% or more (Lindenfors et al. 2002). In extreme cases, such as with the elephant seal, males can be up to 8 times larger than females (Fairbairn et al. 2007). The levels of SSD reached by male-biased species pales in comparison to the level achieved by species with female-biased dimorphism. When females are the larger sex, this size dimorphism can be so extreme that females may be thousands of times larger than their male counterparts (Fairbairn et al. 2007). Some species, like the blanket octopus, have large (~2m long) females that are between 10,000-20,000 times larger than their nearly-microscopic male counterparts (Norman et al. 2002). Others still, like the anglerfish, are only hundreds of times larger than males, but upon mating, the male becomes absorbed into the female until it becomes merely a resource for gametes (Pietsch 2005). Extreme examples like these demonstrate that whatever the genetic correlation between sexes for these species, for such sexual dimorphism to evolve either selection must win, or the correlation be broken down.  Such examples of extreme dimorphism are rare; SSD is typically of smaller magnitude, with most species attaining between 10% and 20% dimorphism. It is this milder, more typical variety of SSD that is most broadly applicable, because it is the norm for sexually reproducing (invertebrate) animals (Stillwell et al. 2010). Ostensibly, these are the examples that showcase the difficulty of overcoming the genetic correlation between sexes.  !-!Males and females donÕt diverge in size randomly from one another. A pattern known as RenschÕs rule has attempted to explain much of the observed variation in SSD. In his 1950 paper, Rensch posited that SSD tends to increase with increasing body size when the male is the larger sex, and decrease with increasing body size when the female is the larger sex (Fairbairn 1997a). While the mechanisms for RenschÕs rule are as of yet unknown, it would appear that it is much more difficult to increase female size relative to male size than it is the other way around. Unfortunately, there are many examples citing the failures of this rule (Blanckenhorn et al. 2007; Blanckenhorn et al. 2006; Webb & Freckleton 2007; De Lisle & Rowe 2013), thus it should only remain a curious trend.   Ultimate mechanisms of Sexual (Size) Dimorphism  Ultimately, sexual dimorphism arises due to the fact that males and females usually differ genetically (Dawkins & Krebs 1979; Mank 2009) and also contribute unequally to their gametes (Trivers 1972; Bateman 1948; De Lisle & Rowe 2015; Fairbairn 2005; Mank 2009; Bell 1978). While males and female investment in reproductive efforts is generally equal, females spend more energy on a small number of gametes, whereas males spend less energy on more gametes (plus the energy spent on finding and attracting mates). This phenomenon is called anisogamy. As each sex becomes adapted to its role as a large (female) or small (male) gamete contributor, specializations may arise that lead to the separation and definition of each sex, both physiologically and ecologically. Anisogamy, therefore, may create sexual conflict where males and females have differing fitness optima at the cost of one another.  !%.!Sexual selection and sexual conflict Sexual conflict and sexual selection are commonly misinterpreted phenomena and often painted as opposing forces. The original definition of sexual selection dates back to Darwin (1871), where he described the advantages one sex may have over another with respect to competition over mates and fertilization. More recent definitions expanded the definition to include the differences in reproduction due to individual trait variation that affects success in competition over mates and fertilization (Andersson 1994; Kokko & Jennions 2014). While the definition of sexual conflict is more nuanced, it differs from sexual selection in that it only exists when sex A might gain a (cost-free) advantage over sex B at a selective cost to B (Kokko & Jennions 2014). Sexual conflict can therefore exist in any sexually reproducing species where selection may unequally favor one sex (Matsuda & Abrams 1999) and differs from sexual selection, precisely because it considers not only the effect of selection on one sex, but also the costs on the other. Regardless, both phenomena are expected to influence the evolution of SSD (Blanckenhorn 2005; Abbott et al. 2010; Abouheif & Fairbairn 1997).   Since females may invest heavily in offspring production, they often cannot afford to mate with low quality males. However, males of many species produce many ÒcheapÓ gametes, thus lowering some of the reproductive costs (although finding mates may remain costly). Intersexual selection occurs when one sex chooses its mate with a nonrandom bias or set of criteria. Typically, this translates to females choosing between males of varying phenotype to mate with, which increases the fitness of some trait values and decreases others. Trait values that are more desirable to females are !%%!usually associated with greater fitness (Moller et al. 1999), although this is not necessary (Prokop et al. 2012). There are many competing hypotheses to explain how females choose traitsÑsexy sons (Weatherhead & Robertson 1979), good genes (Moller et al. 1999), Fischerian ÔrunawayÕ (Lande 1981), sensory bias (Fuller et al. 2005), etc!Ñbut the common theme is that intersexual selection involves one mateÕs preference for traits in the other (Andersson & Simmons 2006; Andersson 1994).   Alternatively, limited access to females may cause another form of sexual selection, forcing males to compete against one another. In some cases, such as the horned Onthophagus beetles (Emlen 1997 and sources within), amphipods (Birkhead & Clarkson 1980), elephant seals (Mesnick & Le Boeuf 1991), baboons (Alberts et al. 1996), access to females is often guarded by stronger males. Intrasexual selection is frequently cited as the cause of male-biased SSD, particularly within male-biased bird and mammal species. Larger males are more likely to outcompete smaller males for access to females. This then imposes further selection for increased male size (Bininda-Edmonds & Gittleman 2000; Serrano-Meneses et al. 2007; Sz”kely et al. 2004). It is not always, however, in the femaleÕs best interest to increase body size when male size increases (Lindenfors 2002). In cases of male-biased SSD, larger female size can potentially slow down and increase the expenditure of reproduction (Lindenfors et al. 2007).   As stated previously, a fitness advantage for one sex may negatively affect the fitness of the opposite sex, creating sexual conflict. Stark examples of sexual conflict (that are !%&!also under sexual selection) include examples such as: traumatic insemination (Stutt & Siva-Jothy 2001), corkscrew and reverse-corkscrew duck genitals (Brennan et al. 2010), sexual cannibalism  (Arnqvist & Henriksson 1997; Birkhead et al. 1988) but sexual conflict need not always be so dramatic.   One good example of sexual conflict over body size involves the sperm plug used by Drosophila males. Once mated with a female, a male leaves behind a sperm plug whose function is twofold. First, it is a mechanical barrier to prevent competitor sperm from entering the female (Chapman 2001). Second, the plug releases hormones into the female, reducing her propensity to mate with other males (Chapman 2001; Chapman et al. 1995). Unfortunately, the hormones released by the sperm plug are also toxic to the female (Chapman et al. 1995; Pitnick & Garc™a-Gonz⁄lez 2002). All of these effects scale with body size, making larger males more effective at preventing smaller males from mating with the same female, but also reducing the lifespan of their mates (Pitnick & Garc™a-Gonz⁄lez 2002).   However, if sexual conflict were to be experimentally reduced, selection will bias both sexes toward the same optimum, regardless of selection applied to either sex (Abbott et al. 2010; Holland & Rice 1999; Pitnick et al. 2001; Simmons & Garc™a-Gonz⁄lez 2008). In their study, Abbott et al. (2010) removed conflict by crossing an evolving population of males to a static population of females that was unable to respond to selection (based on Rice 1996). Since recombination does not happen in males and all chromosomes were tracked with phenotypic markers, the authors were able to !%'!effectively tie female fitness to the male genome. In the face of several artificial selection regimes, both male and female wings (that are typically dimorphic; detailed below) responded by shifting toward the maleÕs optimum wing shape for each regime.  Fecundity selection It has long been accepted that fecundity selection, at least in invertebrates, can be a major driver of SSD (Andersson 1994; Darwin 1874). The idea is simple: larger females can generate more eggs and/or supply more nutrition to them, which is certainly true for some species (Preziosi et al. 1996). In such cases, if large male size is an impediment to mating, small male size may also be selected for (Moya-LaraŒo et al. 2002; Serrano-Meneses et al. 2007). In an important study, Reeve and Fairbairn (1999) demonstrated that selection for increased fecundity could, in fact, increase SSD; selecting for decreased fecundity, however, had no significant effect. Some sources dispute the fact that fecundity selection alone is sufficient to alter SSD, given that female body size does not always correlate with increased fecundity and increased body size frequently comes at a cost of increased energy requirements (Shine 1988). Still, most sources agree that fecundity selection often plays a large role (Reeve & Fairbairn 1999; Preziosi et al. 1996; Stillwell et al. 2010; Head 1995).   Countervailing Selection Not all selection acts to drive population means in one specific direction. Instead, countervailing selection is a type of selection may prevent phenotypes from becoming too extreme. While other selective forces may be pulling sexual phenotypes apart, !%(!opposing selection may be keeping it from going too far (Schluter et al. 1991). Physiological and environmental niche constraints may bind sexes to differing optima, limiting their ability to respond to other forms of selection (Cox & Calsbeek 2010; Blanckenhorn 2011; Schluter et al. 1991). In some cases, males and females can occupy separate ecological niches, incurring unique selective pressures on one another (Pearson et al. 2006; Herrel et al. 1999). Furthermore, viability selection may select against large body sizes in one or both sexes due to the time and energy constraints it takes for an organism to grow larger in the first place (Blanckenhorn, 2011 and sources within). Given that countervailing selection is by definition an opposition to selection (e.g. fecundity selection), other forms of selection are still more likely to cause increases in SSD, rather than simply preventing it from becoming too extreme.     Proximate mechanisms of Sexual Dimorphism !Fortunately, biologists have made great strides in understanding the genetic mechanisms that govern variation in sex-specific phenotypic differences in phenotype can be explained with genetic mechanisms. Perhaps the most obvious regulator of genetic differences between males and females is the sex chromosome, their primary source of genetic divergence (for many species). While it is the only place that different genes can be found between the sexes, sex-specific gene expression (and its potential influence on SSD) often has less to do with the genes of the sex chromosomes, and more to do with the instructions they carry for regulating shared autosomal genes (Fagegaltier et al. 2014; Barmina et al. 2005; Fear et al. 2015). For instance, males and !%)!females differ in in the expression of many genes, yet the Y chromosome of male Drosophila largely only carries instructions for sperm motility (Brosseau 1960; Carvalho 2002).   The initial steps in sex determination in Drosophila is regulated by the ratio of sex:autosomal chromosomes. Both numerator (number of X chromosomes) and denominator (number of autosomes) elements work together to determine the expression of the first component of the pathway: sex-lethal (Sxl) (Cline 1993; Salz & Erickson 2010; Fear et al. 2015; Cline & Meyer 1996). Alternative splicing then plays a large role in determining how genes are expressed in males and females. Starting with Sxl in the sex determination pathway, nearly every subsequent gene has male and female isoforms, including: transformer (tra), transformer2 (tra-2), double-sex (dsx), intersex (ix), and fruitless (fru) (see Figure 1.1). Even though Sxl is on the X-chromosome (Maine et al. 1985), downstream genes such as tra (McKeown et al. 1987) and dsx (Baker & Wolfner 1988) are found on the autosomes. Genes like tra follow the pattern of Sxl, where in males, an inactive splicing variant is generated. Progress down the pathway proceeds, where the presence of Sxl promotes expression of female active tra, which promotes female-specific isoforms of further downstream elements (Cline 1984; Fear et al. 2015; Salz & Erickson 2010). Absence of a functional Sxl leads to production of the inactive male form of the tra transcript, further activating male-specific splicing of downstream elements. Evidence for the effect of tra can be seen in flies with mutant tra genes. Mutant adult females are nearly indistinguishable from males: the transformation is near complete (Sturtevant 1945).   !%*! Figure 1.1. The Sex Determination Pathway in Drosophila melanogaster. Unlike most gene expression pathway diagrams, expression within this pathway is mediated primarily by sex-specific splicing patterns. The resulting male or female isoforms, displayed in the pathway above, ultimately depend on the expression of Sxl. In females, Sxl activity causes the female isoform of tra to be expressed, leading to female development. In males, the absence of Sxl leads to the production of the (non-functional) male isoform of tra, leading to male development.     !"#!"#$%&'($%&')*"+,&-,&-!%+!After sex determination, the mechanisms of sex-specific regulation of genes and pathways become somewhat less clear. One of the best examples we have involves sex-specific control of pigmentation in Drosophila, where dsx is known to activate sex-specific coloration through cis-regulatory element binding. For example, males have a characteristically dark posterior abdomen, while females are typically much lighter. This sexual pigmentation dimorphism is mediated by alternatively spliced dsx products (Kopp et al. 2000). The male variant, dsxM, represses the pigment repressor bric-a-brac (bab) in the posterior abdomen, allowing homeotic gene Abdominal-B (Abd-B) to control pigmentation. Conversely, the female variant, dsxF, promotes bab expression and ultimately represses pigmentation. Examples like this paint a clear picture of the capacity of the sex determination pathway to regulate sex-specific phenotypes.  Dearth of knowledge on proximate mechanisms of SSD  While our understanding of the evolution of SSD and the development of many sexually dimorphic characteristics is relatively well studied, there is a dearth of knowledge on the proximate mechanisms that generate SSD.   Almost 15 years ago, Badyaev (2002) described the types of developmental mechanisms required to generate SSD. His developmental model was simple. Badyaev posited that SSD could be achieved if either sex deviates in initial body size, growth rate, and/or growth duration. Models like BadyaevÕs may sound intuitive, but the mechanisms governing each developmental characteristic are often complex. A more !%,!complete understanding of the individual developmental mechanisms is therefore required to better understand how SSD is regulated.   Proximate mechanisms of overall body size and SSD  While initial size was originally interpreted as size at birth/hatching, this measure is somewhat more complicated for organisms with multiple discrete life stages. In Drosophila, initial egg size is controlled by the mother (Azevedo et al. 1997) and is assumed invariant with respect to sex. Because Drosophila is holometabolous, though, each stage of life has its own initial body size. Offspring are hatched as eggs, experience three larval molts, pupariate, then metamorphose into adults. Several of these stages are discrete enough to argue for their consideration as Òinitial body sizeÓ proxies. Given this view, it is critical to investigate the initial size, growth rate, and growth duration dimorphisms within each stage of life. If we can pinpoint at which stage and by which mechanism dimorphism occurs, we can more precisely describe the ontogeny of SSD.   Many discrete life stages within Drosophila are governed by hormone-mediated developmental events. One very important developmental-timing event is the point at which larvae commit to pupariation, referred to as the critical size (or critical weight) (Davidowitz et al. 2004; Chown & Gaston 2010; Nijhout 2003; Davidowitz et al. 2003). During development, once a larva achieves a specific size, it triggers an irreversible hormone cascade that terminates in pupariation. This cascade is putatively controlled !%-!primarily by the timing of juvenile hormone (JH), prothoracitropic hormone (PTTH), and ecdysone release within the prothoracic gland (Orme & Leevers 2005; Yamanaka et al. 2013). Since larvae that never reach this critical size (due to starvation) will delay pupariationÑoften indefinitelyÑthis is an excellent proxy for adult body size. Critical size is therefore a reasonable proxy for the size at which adult body development begins.   Like Drosophila, the hawkmoth Manduca sexta is a holometabolous insect with female-biased SSD, but unlike Drosophila, we know a great deal more about the development of its SSD (Stillwell & Davidowitz 2010a; Stillwell et al. 2012; Davidowitz et al. 2003). The rate of growth is not sexually dimorphic, regardless of environmental conditions. Males and females differ initially by their critical size, with females reaching a larger value. Females grow for a longer duration between critical size and peak larval mass than males do (Stillwell & Davidowitz 2010b). Sex-specific differences are attenuated in both low nutrition and low temperature environments. In these high stress environments, it is usually the male whose trait values decrease sharply relative to females. (Stillwell & Davidowitz 2010a; Stillwell & Davidowitz 2010b)  Knowing the development of SSD in a species like Manduca is certainly helpful, but in order to truly understand how it is regulated, we must be able to describe the development of SSD in a model species with excellent genetic tools. Drosophila melanogaster is a premier system for studying genetics and arguably has more available genetic tools than any other metazoan (Duffy 2002; Nicholson et al. 2008; !&.!Blair 2003; Ashburner 1989). Furthermore, its relatively short generation time, high fecundity, low maintenance rearing, and visibly complex phenotypes make for an extraordinary model organism. Importantly it displays a great deal of sexual dimorphism including sex-specific structures like the sex comb and anal plate as well as females being on average 15% larger in overall size (Testa et al. 2013; Huey et al. 2006; David et al. 2003). The final differences in size are in large part due to differences in the overall cell size of the epidermal cells (Azevedo et al. 2002; Partridge et al. 1994).   In chapter 2, I will elucidate the proximate developmental mechanisms of SSD to provide insight into the potential underlying genetic mechanisms. Theoretically, sex-specific differences in the above mechanisms may be responsible for generating SSD, which may arise at any or all of the above physiological checkpoints. In chapter 2, I investigated SSD at each of these checkpoints and found that females diverge from males initially upon reaching their critical size. Females wait until larger to commit to pupariation. Females were also found to exhibit a faster growth rate, until reaching a peak size, after which females also exhibited a faster rate of weight loss. This pre-pupal weight loss was shown to account for a significant reduction in SSD between peak larval mass and pupariation (which did not significantly alter SSD).   Genetic mechanisms of size control and SSD  Originally borrowed from yeast, the Gal4-UAS system has been called the ÒSwiss army knifeÓ of fly biology. It allows for the precise spatial and temporal control of the !&%!manipulation of gene expression (Duffy 2002). Features such as increased transposition rate, maternal germline expression, precise inducibility of expression, refined mosaic analysis, and targeted RNAi knockdowns to mention just a few (Duffy 2002). Basically, the Gal4-UAS system works by crossing flies bearing a UAS-GeneX with those bearing GeneY-Gal4. Once combined, expression of GeneX is turned on in the expression pattern of GeneY (to a first approximation). If a UAS-GeneX-RNAi construct is used, this then knocks down the transcript abundance of GeneX in the expression domain of GeneY. This Gal4-UAS system can be used to efficiently tease apart differences in growth rate, growth duration, and initial size by controlling expression in candidate gene pathways (that are known to modulate these growth parameters). RNAi knockdowns can test whether candidates are necessary for the development of SSD, while over-expressions can test whether the increased expression of candidate genes is sufficient to alter SSD.    Knowing the proximate, physiological and developmental mechanisms of SSD is a great start to understanding the development of SSD; however, there still remains the mystery of how these differences are being regulated genetically to modulate development in sex-specific ways. Very few studies have investigated the direct genetic contribution of specific genes and pathways to SSD. While many studies have investigated heritability and genetic architecture of SSD (Stillwell & Davidowitz 2010a; Bonduriansky & Rowe 2016), few if any directly address the specific genetic contributors to SSD. Moreover, none (except Rideout et al, 2015) have arguably been able to establish a causal relationship between a certain gene/pathway and SSD.  !&&! One way to search for candidate genetic pathways is to observe the conditions associated with the loss of SSD. We know from certain studies that environmental factors play a role in regulating SSD (Stillwell & Davidowitz 2010a; Blanckenhorn et al. 2007; Stillwell et al. 2010). This link is strong enough that it has lead researchers to posit that perhaps the evolution of SSD itself is linked to condition dependence (Bonduriansky 2007b; Johnstone et al. 2009; Stillwell et al. 2010). Mechanisms responsible for generating a plastic response to an environmental factor like temperature or nutrition are thought to acquire sex-specific expression patterns to resolve conflicting selective pressures on body size. If this were the case, mechanisms that cause SSD to fluctuate with the environment could be the very same genes that cause differences in male and female body size. We have many examples that suggest sexual dimorphism is condition dependent with respect to temperature (Azevedo et al. 2002; Stillwell et al. 2010) and nutrition (Stillwell et al. 2010; Emlen et al. 2012; Gotoh et al. 2014; Koyama et al. 2013; Moczek & Kijimoto 2014). Given these examples, the best place to start searching or the genes that generate SSD are those that also control nutrient-sensitive growth.   One well-studied growth pathway, the Insulin/Insulin-like Signaling (IIS) pathway has already been implicated in regulating overall growth, and potentially modulating SSD. Mutants of the IIS pathway, such as Insulin-Receptor, InR (Testa et al. 2013), and Foxo (Carreira et al. 2011) eliminate or reduce SSD. The resulting reduction in body size for IIS mutants is typically said to phenocopy starvation conditions by forcing cells to act as !&'!though there are fewer nutrients available than are truly available. Unfortunately, it is unknown whether there are sex-specific differences in InR knockdowns and overexpressions. That the IIS pathway is necessary for SSD is evidence that IIS and possibly other nutrient sensitive growth pathways may be involved in positively and negatively regulating SSD.   To determine causation within a given pathway, one must be able to determine three important factors: necessity, sufficiency, and proximity. Necessity refers to the dependence of a trait on a particular gene. If a trait is affected by the absence of a gene, it is said to be necessary for that traitÕs expression. In this way, InR has been shown to be necessary for SSD. What is unknown is whether overexpression of InR is able to increase the value of SSD (i.e. whether it is sufficient). Even if genes are shown to be necessary and sufficient for a trait, however, expression must be somehow linked to the spatial or temporal region of that trait to truly be a candidate.   In Drosophila, nutrient-sensitive growth can be regulated through several different signaling pathways. In response to circulating nutrients, Drosophila Insulin-like Peptides (dILPs) and insulin-like growth factors are produced primarily in the brain by the medial neurosecretory cells (but also in the fat body and imaginal discs) (Edgar 2006). Insulin signaling then begins at the transmembrane insulin-receptor, InR. Upon binding to a ligand, e.g. dILPs, InR begins a signaling cascade that ends with the repression of a growth repressor, Foxo (see Figure 1.2). Many other pathways, such as TOR and Ras, are more straightforward (Figure 1.2). For example, TOR signaling responds to !&(!circulating levels of amino acids, which derepresses RHEB by repressing the negative growth regulator TSC1/TSC2 (Edgar 2006). TOR signaling then proceeds by activating a positive growth regulator, S-6Kinase (S6K). Since Ras was the only component of the Ras pathway that we considered, it will not be discussed here.   Another candidate growth pathway we tested was the Hh signaling pathway (Figure 1.3). In 2005, Horabin demonstrated the ability of Sxl to affect levels of full length Ci and also wing imaginal disc size. Their hypothesis was that female wings grow larger because of endogenous levels of Sxl increasing activation of the Hh pathway. Normally Hh ligands bind to the transmembrane receptor patched (ptc), which activates growth by derepressing the positive growth regulator smoothened (smo). In the presence of Sxl, which exists in females and not males, nuclear entry of full length Ci is increased, speeding up growth (Horabin 2005).     In Chapter 3, I examined the genetic mechanisms responsible for generating SSD. Here, I tested the necessity and sufficiency of genes in several candidate pathways to affect whole body SSD (see figures 1.1, 1.2, and 1.3). The resulting data add support to recent findings (Rideout et al. 2015) with regard to the control of SSD in the sex determination pathway (Figure 1.1), and also contribute a novel candidate gene capable of affecting SSD. Furthermore, we test the necessity and sufficiency of other growth pathways to generate SSD, including the following: nutrient sensitive growth pathways (Figure 1.2) and the Hh pathway (Figure 1.3). Our findings suggest caution when attempting to uncover the genetic basis of traits. While preliminary data suggested that   !&)! Figure 1.2. Nutrient-Sensitive Growth Pathways in Drosophila melanogaster. Represented are three candidate pathways for controlling SSD in Drosophila, including: (black) Insulin and Insulin-like Signaling (IIS) pathway, (medium gray) TOR signaling pathway, and (light gray) the Ras signaling pathway. Arrows represent positive interactions between two components and bars represent negative interactions (repression).    !"#$%&'()*&+,#-./0/123435'6-7-'8/17,9:17,;)<=>13)?@/)A7=)/9:=)/;!&*! Figure 1.3. Hh signaling pathway in Drosophila melanogaster. This pathway was put forth by Horabin (2005) as a hypothesis for the regulation of SSD in Drosophila. In the absence of the Hh ligand, a short and repressor version of the transcription factor, Ci, is produced since patched (ptc) is inhibiting smoothened (smo). In the presence of Hh, ptc de-represses smo, allowing for the expression of the full length version of Ci, which upregulates growth. Horabin (2005) demonstrated that, in the presence of endogenous Sxl protein, this reaction is catalyzed, increasing growth in females.   !"#$%&'()"*+,-#*(."(&,+/0*1.*1--&*234!"#$%&5'(678..1,4(741-"/&61*9*&:;&(-'&%./15/&61-"34!"#$%&5'(678..1,4(741-"/&61*9*&:;&(-'&%./15/&61-"$5/!&+!IIS pathway was our best candidate pathway, our experiments yielded conflicting results. Finally, we reveal an unanticipated effect of ubiquitous over-expression of nutrient-sensitive growth   How are size and shape related? A Primer on Shape Analysis   Since most of the natural variation visible in organisms is related to size and/or shape, it is unthinkable that we know so little about how it is regulated with respect to sex, especially on the genetic level. Much of the literature investigates the size or shape control in sexually dimorphic structuresÑe.g. beetle horns, (Kijimoto & Moczek 2016; D.J. Emlen et al. 2005), Drosophilid sex combs (Barmina & Kopp 2007) and anal plate (Glassford et al. 2015), C. elegans tail phenotype (Suzuki et al. 1999), queen bee phenotype (Kamakura 2011)Ñbut many more have explored size and shape dimorphism in homologous structures exhibited by both sexes (Gidaszewski et al. 2009; Abbott et al. 2010; Cheng & Kuntner 2015; David et al. 2003; Stillwell & Davidowitz 2010a).  While the regulation of either size or shape is relatively well-studied (Nijhout 2008; Oldham et al. 2000; Edgar 2006; Weber 2005; Gidaszewski et al. 2009; Day & Lawrence 2000), much less is known about what individual genes/pathways contribute to both size and shape.  Furthermore, the extent to which size and shape controlling genes overlap in function is poorly understood. Some studies, such as Carreira et al. 2011, are beginning to uncover sex-specific differences in how size and shape are !&,!regulated. In their study, the authors used an unbiased random screen to identify mutations that differed in sexual size and shape dimorphism. They showed that, of the mutations used, males were more likely than females to respond with changes in wing size only. Conversely, roughly two thirds of mutations caused an effect on wing shape in only one sex, suggesting that most mutations have sex-specific effects. While this is certainly an important contribution to the field, its scope is too broad to really elucidate how certain pathways contribute to sexual size dimorphism (SSD) or sexual shape dimorphism (SShD).   Because sexually dimorphic traits are often multi-dimensional, studying characteristics that are more quantifiable than size would allow for a more complete understanding of the underlying sexual dimorphism. One such sexually dimorphic, multi-dimensional trait is shape.   One important question in shape analysis is ÒWhat makes size so inherently easy to quantify while shape is so complicated?Ó Shape change can be thought of as a relative change in size of a given trait. If you change the size of the anterior portion of a wing, for example, you change the entire shape of the organ. Similarly, by changing the size of the wing, relative to the body size, the shape of the whole organism changes as well.   The study of scaling relationships (allometry) have been central to the study of body size variation. DÕArcy Thompson was one of the first scientists to outline a thorough approach to the study of allometry in his 1917 book, On Growth and Form (Thompson !&-!1917). Thompson illustrated the differences in relative size of organs to overall body size by using visual deformation grids. The most famous example demonstrates the transformation from the porcupine fish, Diodon, to the sunfish, Mola. The deformation grid allows us to see that major differences occur only in the relative size of structures between these organisms. This is an example of an early method of describing allometric ÒshapeÓ changes. Unfortunately, ThompsonÕs illustrations are merely qualitative in nature. How then do we quantify differences in shape between organisms?   Early crude methods focused on using linear measurements between predetermined points. Usually measurements were taken for variables such as length and width. In his 1932 book, Problems of Relative Growth, Huxley described how one part of a structure might change with respect to another, e.g. wing vs body size (Huxley 1932). The allometric scaling can certainly change the overall shape of an organism, but unfortunately measures like these are still not good estimates of shape. Other crude methods use ratios of univariate measurements to quantify Òshape.Ó This method, however, is often severely flawed, because two entirely different starting shapes may yield the same results (e.g. a diamond and an oval). Obviously these methods are not sufficient to capture shape variation in any real sense.   Outline analysis was yet another promising method to describe changes in shape. Elliptical Fourier Analysis, a common example, works by estimating the minimum number of ellipses required to mimic the shape. Again, this process is flawed, because it only uses the outline of the specimen and does not take any biologically meaningful !'.!traits into account. Because of this, outline-based approaches can be used to directly compare the shape of a human scapula, a potato chip, and a chocolate chip cookie without regard for actual homology (Zelditch et al. 1995).   To date, landmark-based approaches are our best method for presenting a clear picture of shape differences between two specimens. Landmarks are placed on a 2- or 3-dimensional representation of the specimen with X, Y (and sometimes Z) Cartesian coordinates. Where the landmarks are placed is somewhat arbitrary, but there are several rules that must be followed. All landmarks must be homologous between specimens and be as unambiguous as possible. Early methods for analysis focused on quantifying the variation of one point to adjacent points, such that a ÒtrussÓ was measured (Bookstein 1985). These methods were soon replaced by more sophisticated methods that described shape as the residual variation left over after location, rotation, and scale effects were removed. Calculating shape variation now allows for all specimens to be standardized to a common baseline. BooksteinÕs shape coordinates allowed for a common baseline, but since these points are arbitrary, they can certainly influence shape inferences made on their behalf.    Least-squares superimposition, i.e. Ordinary Procrustes Analysis (OPA) resolves most of the issues of early landmark-based techniques. This technique uses the centroid as the seat of translation, rotation based on optimal fit of landmarks between configurations, and centroid size for ensuring consistent scale. Generalized procrustes !'%!analysis (GPA) goes a step further and repeats this superimposition of OPA in an iterative fashion to minimize variation due to translation, rotation, and scale.    In Chapter 4, I begin to dissect the contributions of specific genes/pathways to the generation of shape and size dimorphism. I use a previously published data set to investigate the sex-limited effects of mutants on size and shape dimorphism. The primary use of this Chapter was to describe and develop the methods used in the following chapter. Both chapters utilize landmark-based geometric morphometrics to quantify the size and shape differences between wild-type fly wings and those of treatment groups. Here I demonstrate the sex-limited effects of genes within two growth pathways (EGFR and TGF-§), where effects on SSD and SShD are previously unknown.   Growth pathways can conceivably influence both size and shape of developing organisms, yet little attention has been paid to their combined efforts. In Chapter 5, I test explicit hypotheses about the ability of candidate pathways with known effects on whole body SSD to influence SSD and SShD in the wing. Similar genes were found to be necessary for whole body and wing-specific SSD, and many more found to be sufficient to increase SSD. Surprisingly, data from this chapter conflict with the literature in a few key ways, but nonetheless suggest important ways that our candidate pathways influence differences in shape and size.    !'&!Collectively these studies bring our understanding of the proximate mechanisms that regulate SSD and SShD to a new and more profound level. While much is known of the selective pressures that generate sexual dimorphism, we know very little about the specific genetic targets of these pressures, and how these targets may facilitate or hamper the evolution of sexual dimorphism. My study goes some way to filling this conspicuous gap. Only by quantifying the effects of genes on all aspects of trait variation can we fully appreciate a geneÕs effect on the sexual dimorphism of a trait. Understanding the proximate developmental and genetic mechanisms of sexual dimorphism can allow us to truly appreciate the ability of a shared genome to generate Òendless forms most beautiful and most wonderful [that] have been, and are being, evolved.Ó    !''! CHAPTER 2:  Testa, N. D., Ghosh, S. M. & Shingleton, A. W. Sex-Specific Weight Loss Mediates Sexual Size Dimorphism in Drosophila melanogaster. PLoS One 8, e58936 (2013).    !'(!Introduction  Sexual Size Dimorphism (SSD), the difference in body size between males and females, is an extraordinarily widespread and conspicuous phenomenon in the animal kingdom (Fairbairn 1997a). This condition is extremely variable and evolutionarily labile. For example, male southern elephant seals can weigh seven times that of a female (Bininda-Edmonds & Gittleman 2000), while female blanket octopuses can weight 10,000-20,000 times their male counterparts (Norman et al. 2002). The degree of SSD in insects is generally less extreme, however. Consistent with most invertebrates, the female is often the larger sex among insects, a pattern seen in approximately 88% of insect species (Stillwell & Davidowitz 2010b). Despite the ubiquity of SSD, however, very little is known of the underlying developmental mechanisms that generate it or how these mechanisms evolve.   In general, final body size is regulated by a combination of three developmental factors: initial body size (size at hatching/birth), growth rate, and growth duration (Badyaev 2002; Blanckenhorn et al. 2007). Changing any of these individually or in combination results in an alteration of adult body size and may underlie size differences between males and females. Nevertheless, the molecular-genetic and physiological regulators of initial size, growth rate, and growth duration are poorly understood except in a very few organisms. One such organism for which these mechanisms are known, is the fruit fly Drosophila melanogaster, which like most insects, shows marked SSD between female and male body size (Stillwell & Davidowitz 2010b; Blanckenhorn et al. 2007). The !')!extensive research on growth regulation in Drosophila and its readily apparent SSD make these animals an ideal model to more deeply elucidate the proximate mechanisms that regulate SSD.    Drosophila are typical holometabolous insects: they begin life as worm-like larvae, molting through three larval instars before undergoing complete metamorphosis as a pupa and eventually eclosing into their adult form (Nijhout 1981). Adult flies, like all arthropods, have a stiff exoskeleton, meaning they cannot grow. Larval body size upon termination of growth, therefore, ostensibly determines adult body size.    In Drosophila, the timing of metamorphosis is regulated by a larva reaching a size checkpoint called critical size (or critical weight) early in its final larval instar. Attainment of critical size is associated with initiation of a hormonal cascade that ends in metamorphosis. There is, however, temporal separation between the attainment of critical size and the subsequent rise in the ecdysteroid titer that causes the larva to stop feeding and ends body growth. This delay provides a final period of growth for the larvae, called the Terminal Growth Period (TGP), during which Drosophila larvae can more than triple their mass (Shingleton et al. 2008; Stieper et al. 2008). Body size in Drosophila is therefore regulated by the critical size, plus the amount of growth achieved during the TGP (Shingleton et al. 2008; Shingleton et al. 2007; Davidowitz et al. 2003; Nijhout et al. 2006; DÕAmico et al. 2001), or more formally:    Final Body Size = Critical Size + (Growth Rate * TGP) !'*!where Critical Size is the weight at which larvae commit to pupariation, TGP is the time between critical size and cessation of growth, and Growth Rate refers to the rate of growth within the TGP. SSD in Drosophila is therefore a consequence of sex-specific differences in one or all of these parameters.   Research over the last twenty years has begun to establish the developmental mechanisms that regulate critical size, growth rate and the duration of the TGP (Mirth et al. 2005; Caldwell et al. 2005; Nijhout & Williams 1974; Shingleton 2005). The goal of this study is to determine the proximate mechanisms responsible for SSD in Drosophila melanogaster. To test the hypothesis that sex-specific differences in a combination of developmental events underlies SSD, we measured critical size, growth rate, and growth duration for the developing larvae and pupae. Identifying how these developmental parameters differ between male and female flies therefore allows us to begin to link the observed SSD to the endocrine, and ultimately molecular-genetic, mechanisms that regulate growth and development.   Materials and Methods  Fly Strains and Maintenance All flies were derived from an isogenic stock of Samarkand (SAM) Drosophila melanogaster. Ubi-GFP (y1w67c23P{Ubi-GFP.D}ID-1) flies were obtained from Bloomington Stock Center and back crossed into a SAM background for five !'+!generations to eliminate background effects. Flies were raised on standard cornmeal-molasses medium at 25¡C  on a 24 hour light cycle.   Critical Size Mid-third instar larvae that weighed between 0.3 to 2.3mg were placed into individual tubes and starved. Time to pupariation (TTP) was recorded on an individual basis and critical size was calculated as the weight at which starvation no longer delayed pupariation [see Stieper et al., 2008 for additional details]. Flies that survived to the late pupal period were sexed using the presence or absence of sex combs. For those pupae that died before the presence or absence of sex-combs could be scored, sex was determined by presence or absence of the Y-chromosome-specific Ppr-y gene using PCR and gel electrophoresis. DNA from un-sexable pupae extracted using a Qiagen DNA extraction kit and PCR was conducted under standard conditions. Optimal annealing temperature for PCR primers (below) was found to be 58¼C.   Forward:  5'   TGT GTT GAT GAC CGT GAC GCC A   3'  Reverse:  5'   CGA GTC GCA ATT GTG TCT TCT CGC   3'  Growth rate Eggs were laid in six-hour cohorts from which larvae were sampled every six hours and developmental stage and mass were recorded. Larval sex was determined by using presence or absence of a paternally inherited X-chromosome marked with a constituently active GFP. Sex was recorded based on presence or absence of GFP, to detect females and males respectively. Pupae were staged into four-hour cohorts at !',!pupation and massed every 12 hours. Timing of pupariation was determined by using SAM flies laid in six-hour cohorts. Starting at 94 hours, we recorded pupariation state for individual larvae. Pupal sex was determined retrospectively by presence or absence of sex combs.   Statistical Analysis All statistical analyses were conducted using R statistical software (version 2.14.1). Critical size was calculated using the methods described in Stieper et al. (2008). To assess the probability of observed sex-specific differences in critical size, we used a permutation test with one thousand replicates to generate a null distribution of the difference in critical size between males and females. The same test was also applied to determine differences in time to pupariation from the critical size data. Growth rate was calculated using a linear regression of log-transformed weight against time, while interactions with sex were tested using an Analysis of Covariance (ANCOVA). We calculated TGP by subtracting the time at which critical size is attained from the time at which larval weight no longer significantly increases, for each sex. We applied the values for critical size to the growth curve to determine the timing of critical size and used multiple comparisons analysis (Hsu's MCB) to determine the age at which there is no longer any significant increase in mass for each sex. Since this approach does not allow us to calculate 95% confidence intervals for the timing of growth cessation, confidence intervals for the duration of the TGP were predicted using those for critical size alone. In all larval cultures, we noticed that some larvae stopped growing prematurely and subsequently failed to pupariate. In order to avoid including these !'-!abnormal larvae in our growth calculations, any larvae whose weight was below critical size by the time the rest of the population had stopped growing were excluded from the analysis. Values for the timing of developmental stages were determined by logistic regression of developmental stage against age. Finally, all SSD indices were calculated as per the 1992 Lovich and Gibbons paper (Lovich & Gibbons 1992; Smith 1999), such that SSD = (F/M)-1, where F is female weight and M is male weight.   Results  There are three potential mechanisms by which SSD can be generated in Drosophila, namely, sex-specific differences in critical size, TGP, and growth rate. We found that male larvae have a significantly smaller critical size than females (permutation test, P = 0.008) (Figure 2.1a). This does not, however, wholly explain the adult SSD. Females also grow more rapidly than males during their TGP (ANOVA, P = 0.0084) (Figure 2.1b), although their TGPs are approximately the same duration (17.5 and 16 hours, respectively). The nature of our data does not allow us to test this statistically, however, males have a significantly longer time to pupariation from critical size than females (permutation test, P = 0.01), which is a proxy for the TGP. Finally, the timing of both larval and pupal molts as well as eclosion timing do not differ significantly between sexes (logistic regression; molt to 2nd instar, P = 0.5330; molt to 3rd instar, P =  0.8282; pupal molt, P = 0.7432; eclosion, P = 0.9628). (Figure 2.1d)     !(.!  Figure 2.1. Complete growth profile by sex for Drosophila melanogaster. Factors shown to contribute to SSD include (a) critical size, (b) growth rate, (c) and pre-pupal weight loss and are reflected in the sex-specific growth curve (d). The SSD at specific   !(%!Figure 2.1. (contÕd)  developmental events (hatching, critical size, peak larval mass, pupariation and eclosion) illustrates the changes in SSD throughout development (e). Error bars show 95% confidence intervals.   !(&!Surprisingly, SSD at peak larval mass is twice that of the uneclosed adult fly: females were 30% larger than males at peak larval mass and 12% larger than males by end of pupal development (Figure 2.1e). To determine why this difference in SSD exists, we measured pre-pupal weight loss, weight lost in the period intervening peak larval mass and pupariation, and pupal weight loss, weight lost during the pupal stage. The female rate of weight loss during the larval stage is significantly greater than the male rate (ANCOVA, P = 0.0116) (Figure 2.1c), whereas there was no significant difference in pupal weight loss (ANCOVA, P = 0.6078).   Discussion  Consistent with the female biased dimorphism in insects, female Drosophila adults are significantly larger than their male counterparts. Our data indicate that this sexual size dimorphism arises because females 1) initiate metamorphosis at a larger size than males, that is they have a larger critical size, and 2) grow faster than males in the terminal growth period between critical size and the cessation of larval growth. Surprisingly, however, the resulting SSD at the peak of larval mass is subsequently reduced before metamorphosis because females lose more mass during the pre-pupal period. Additionally, our data show that the timing of larval molts and pupation are nearly identical in males and females, and that the duration of growth is not different between the sexes. Males do, however, eclose slightly earlier than females.  !('!There is a paucity of data concerning the patterns of growth that generate SSD in other insects (Blanckenhorn et al. 2007; Stillwell & Davidowitz 2010a; Molleman et al. 2011; Fischer & Fiedler 2001). Perhaps the best study has been in the tobacco hornworm, Manduca sexta, where critical size and the duration of the TGP (called the ICG in M. sexta) are important mechanisms contributing to SSD at the cessation of larval growth (Stillwell & Davidowitz 2010a). Additional studies indicate that SSD in other Lepidopterans accumulates during development primarily due to females adding more instars than males (Esperk et al. 2007). This is consistent with females having a longer TGP/ICG.  In contrast, a previous study indicates that SSD among Drosophilidae is a consequence of sex-specific differences in growth rate and this is supported by our study (Blanckenhorn et al. 2007). Different insect species therefore appear to generate SSD using different developmental mechanisms. It is possible, however, that differences in SSD for both Drosophila and Manduca are a consequence of the differing environments in which each was reared.   The observation that SSD is influenced strongly by the loss of mass between the cessation of growth and pupation is a novel one, although post-eclosion weight loss has been implicated in regulating SSD in Lepidopterans (Molleman et al. 2011; Fischer & Fiedler 2001). To a certain extent, mass loss after a larva has stopped feeding is an inevitable consequence of ongoing metabolic and developmental activity. What is not clear is why females lose more mass than males; it seems counterintuitive for females to accrue mass only to lose it. One hypothesis is that selection for larger female size targets a systemic increase in growth rate, both of the body as a whole, but also of the !((!imaginal discs, the precursors of adult organs. Importantly, growth and development of the imaginal discs continues after the cessation of feeding (Freeman 1997; Garcia-Bellido & Merriam 1971), relying on stored nutrients to proceed (Slaidina et al. 2009; Okamoto et al. 2009). Thus, we might expect that larger females with larger organs will utilize more of these stored nutrients during post-feeding imaginal disc growth. Consequently, both the increase in the body's growth rate before cessation of feeding and the increase in weight loss after the cessation reflect the same mechanisms of elevated growth rate for increased body size in females.   There are a number of pathways that control growth rate, which include IIS, TOR, MAPK, and HIF-signaling pathways (Britton et al. 2002; Wullschleger et al. 2006; Liu et al. 2006; Seger & Krebs 1995; Shingleton et al. 2005). Of these, the insulin signaling pathway has been demonstrated to have an important role in regulating final body size (Ikeya et al. 2002; Chen et al. 1996). This pathway regulates the rate of cell growth and proliferation in response to insulin-like peptides that are released in a nutrient dependent manner by the brain and other tissue around the body (Shingleton et al. 2005; Edgar 2006; Emlen et al. 2012). Ostensibly, therefore, insulin signaling regulates growth and final body size with respect to developmental nutrition. However, data from Drosophila and other animals suggest that differences in insulin signaling may account for body size variation among different populations (Fabian et al. 2012; De Jong & Bochdanovits 2003; Sutter & Bustamante 2007), suggesting that it may be a more general regulator of size. An intriguing hypothesis therefore, is that female Drosophila are larger than males because elevated levels of insulin signaling increases growth rate. !()!The insulin signaling hypothesis was first proposed to explain SSD in Manduca sexta (Stillwell et al. 2010), however, evidence suggests that it may be important in regulating SSD in Drosophila as well. Support for this hypothesis comes from the observation that SSD is eliminated in flies mutant for the insulin receptor (InR) (Figure 2.2), indicating that insulin signaling is necessary to generate size differences between males and females. However, the fact that there is no SSD in InR mutants suggests that insulin signaling is also regulating other mechanisms that generate SSD, specifically difference in critical size.  Regardless of the function of pre-pupal weight loss, our understanding of how body size is regulated in Drosophila melanogaster needs to be extended. Pre-pupal weight loss should now be viewed as an additional variable for calculating final body size, such that:  Final Body Size = Critical Size + (Growth Rate * TGP) - Weight Loss.   In conclusion, our data suggest that the mechanisms regulating critical size and growth rate are responsible for generating SSD in Drosophila melanogaster. Our understanding of the underlying molecular-genetic mechanisms that regulate these processes indicate that these studies can be extended to generate a deeper understanding of the development of SSD.    !(*! Figure 2.2. SSD is lost in insulin-signaling mutants. The dry mass of male and female adult InrE19/InrGC25 and wild-type (InrE19/TM3) control flies reared at low  density at 24ûC. Columns with different letters are significantly different (Tukey HSD at  P<0.05). Error bars are standard errors.      !(+! CHAPTER 3:   The Genetic Regulation of Whole Body Sexual Size Dimorphism in Drosophila melanogaster   !(,!Introduction  One of the most striking differences between male and female fruit flies is that of size. Sexual size dimorphism (SSD) is nearly omnipresent among sexually reproducing organisms, only ever completely vanishing in truly monogamous species (Sz”kely et al. 2007). Drosophila melanogaster is no exception to the ubiquity of SSD. While some species may display flashy and extreme SSD, DrosophilaÕs is of the much more common variety: moderate (Huey et al. 2006). Males and females differ in body size by only about 15-20%, depending on the measure of SSD (Testa et al. 2013). Given how typical their SSD is, studies on Drosophila are well suited to making more general statements about SSD than those with more dramatic dimorphisms.   Drosophila are holometabolous insects, which means they experience several discrete life stages. Once their eggs hatch, Drosophila larvae cycle through three instars (stages) before pupariating, metamorphosing, and eclosing as adults. Recent evidence demonstrates that SSD arises in the final larval instar because females have a greater critical size and growth rate than males (Testa et al. 2013). The increased growth rate in females causes them to gain more mass than males until peak larval weight, then lose more before pupariation.   The reason for this pre-pupal weight loss is still speculative; however, current hypotheses suggest that differences in energy used during the pre-pupal wandering stage, where larvae leave the food in search of suitable pupation locations, is to blame. !(-!If this is the case, it appears that, for an equivalent time spent eating and fasting, females gain more mass while feeding and lose more when not. SSD should therefore be dependent on nutritional state.   We know that body size is condition-dependent (Ghosh et al. 2013), specifically with respect to environmental nutrition (Britton et al. 2002), but whether or not this phenomenon is sex-specific is unknown. Determining to what extent SSD is nutrient-dependent, will allow us to confirm a suite of candidate pathways responsible for generating SSD.   One nutrient-sensitive pathway, the Insulin/Insulin-like Signaling (IIS) pathway has already been implicated in regulating SSD. One study demonstrates that mutants of the Insulin-Receptor (InR) completely lack SSD (Testa et al. 2013). The resulting reduction in body size for IIS mutants phenocopies starvation by forcing cells to act as though there are fewer nutrients available than there really are. Evidence shows that this appears to be the case (Britton et al. 2002; Marshall et al. 2012), but sex-specific differences in other growth-related genes are still unknown. Ultimately, InR controls growth by repressing the protein product of the negative growth regulator, the forkhead transcription factor Foxo. Further research has demonstrated the ability of Foxo overexpression in the wing to decrease size in a sex-specific manner (Carreira et al. 2009). The dependence on Foxo and the requirement for an active IIS pathway is necessary for SSD is evidence that IIS and possibly other nutrient sensitive growth pathways may be necessary for regulating SSD.  !).! Unfortunately it is unknown whether IIS activity is sufficient to positively regulate SSD.  Research in Drosophila demonstrates the ability of IIS overexpression to increase organ size (Brogiolo et al. 2001; Aron et al. 2010). It is conceivable that an increase in IIS expression may be met with a sex-specific response in growth, especially if females are more sensitive to levels of IIS. If increase in IIS expression is sufficient to increase SSD, then a causative link may be established between IIS and SSD. One link between the sex determination and IIS pathways was proposed by Rideout, Narsaiya, & Grewal (2015), where mutation of a negative regulator of the IIS pathway, PTEN, was found to rescue female body size in artificially masculinized crosses (via tra knockdown). We hypothesize that it is growth-related pathways such as IIS that are regulating sex-specific differences in nutrient-sensitive growth.  The IIS pathway is not the only nutrient-sensitive growth pathway, however (Gokhale & Shingleton 2015). Neither is it the only pathway previously implicated in sex-specific growth patterns. We know very little about the effect of many of these other pathways responsible for controlling growth in Drosophila. We suspect that perturbations of other nutrient-sensitive growth pathways will also result in changes to SSD. Alternate candidate pathways include the following:  TOR signaling pathway- This pathway interacts with the IIS pathway in regulating growth. One major difference is that, rather than responding indirectly to levels of nutrition (through insulin-like peptides, dILPs), TOR signaling responds to cellular levels !)%!of amino acids. According to a recent paper, the downstream element S6-Kinase (S6K) is responsible for the drastic increase in SSD found in honeybee queens (Kamakura 2011). He further demonstrated that it has similar body size effects in Drosophila, though they did not specifically investigate its effect on SSD.   Ras pathway- While this pathway was originally described for its role in cancer biology, it has long since been known to control bodily growth through regulating the rate of cell division (Shingleton et al. 2007). In Drosophila, Ras has also been shown to control body size through the release of the molting hormone, ecdysone, via the prothoracic gland (Caldwell et al. 2005).   Hh signaling pathway- Recent evidence suggests that this pathway is involved in the regulation of sexually dimorphic growth rate of wing imaginal discs (Horabin 2005). The proposed mechanism is that circulating proteins used in the sex-determination pathway catalyze growth regulation in this pathway. Further research has implicated the Hh signaling pathway in generating sexually dimorphic appendage size in Onthacophagus beetles, through regulating IIS activity (Kijimoto & Moczek 2016).   Sex determination pathway (SDP)- This pathway is ultimately responsible for the regulation of SSD. What is unclear though, is where along this pathway SSD is actuated. Recent evidence suggests the gene transformer (tra) is involved (Rideout et al. 2015), but it remains unclear whether additional elements of the pathway are as well (Cline 1984). !)&! We propose that the above pathways are involved in regulating SSD. We already know that some (IIS and SDP) are necessary for SSD, but which pathways are sufficient to increase SSD are as of yet unknown. Specifically, our hypothesis for nutrient-sensitive growth pathways (IIS, TOR, Ras) is that females are larger than males due to an increased sensitivity to pathway activity. If any candidate genes within these pathways are, in fact, involved in SSD regulation, then genetic manipulations should reveal their effects. Specifically, if a pathway is a positive regulator of SSD, then increasing signaling through the pathway by increasing expression of its constituent genes, should increase SSD. Conversely, if a pathway is a negative regulator of SSD, the reverse should be true. Other candidate pathways would be expected to yield similar, positive correlations with SSD if they are responsible. Fortunately, the genetic tools available in Drosophila allow for the very precise control of gene expression. By using the GeneSwitch Gal4-UAS system, we were able to titrate gene expression to explore the effects of increasing or decreasing signaling through different pathways on SSD. Our results suggest that suppressing signaling through several pathways can suppress SSD, but that it is difficult to increase SSD.  Materials and Methods Fly strains and Maintenance All fly lines were chosen based on their ability to overexpress or knockdown specific predetermined candidate genes. With few exceptions, all flies were obtained from the Bloomington Stock Center. See table 3.1 for complete stock list. Flies were maintained   !)'!Table 3.1. Fly Stocks used in Chapters 3 and 5. Table includes response element (gene), direction of manipulation, genetic background into which each gene is introgressed, Bloomington Stock center ID (where applicable), and full stock genotype.                                    gene direction genetic background Stock # Stock Akt overexpression Sam 8191 y[1] w[1118]; P{w[+mC]=UAS-Akt1.Exel}2 Akt knockdown y,v 31701 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.HM04007}attP2 Ci overexpession Sam 32570 w[*]; P{w[+mC]=UAS-ci.HA.wt}3 Ci knockdown y,v 28984 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF01715}attP2 dsx F overexpression Sam 44223 y[1] w[*]; P{w[+mC]=UAS-dsx.F}24-3 dsx M overexpression Sam 44224 y[1] w[*]; P{w[+mC]=UAS-dsx.M}2 dsx knockdown y,v 26716 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF02256}attP2 FOXO overexpression Sam Shingleton Lab y, w: +:UAS-dFOXO-wt (f19-5 II) FOXO knockdown y,v 27656 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF02734}attP2 fru overexpression Sam 17551 y[1] w[67c23]; P{w[+mC] y[+mDint2]=EPgy2}fru[EY09280] fru knockdown y,v 31593 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF01182}attP2 !)(!Table 3.1. (contÕd) !!InR overexpression Sam 8262 y[1] w[1118]; P{w[+mC]=UAS-InR.Exel}2 InR knockdown y,v 35251 y[1] sc[*] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.GL00139}attP2 Ptc overexpression Sam 5817 w[*]; P{w[+mC]=UAS-ptc.J}B1 Ptc knockdown y,v 28795 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF03223}attP2 Ras overexpression Sam 4847 w[1118]; P{w[+mC]=UAS-Ras85D.V12}TL1 Ras knockdown y,v 31653 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF01445}attP2 S6K knockdown y,v 41702 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.HMS02267}attP2 S6K overexpression Sam 6910 w[1118]; P{w[+mC]=UAS-S6k.M}2/CyO Sxl overexpression Sam 17354 y1 w67c23 P{EPgy2}SxlEY06108 SxlEY06108 Sxl knockdown y,v 34393 y[1] sc[*] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.HMS00609}attP2 Tor overexpression (dominant negative) Sam 7013 y[1] w[*]; P{w[+mC]=UAS-Tor.TED}II Tor overexpression Sam 7012 y[1] w[*] P{ry[+t7.2]=hsFLP}1; P{w[+mC]=UAS-Tor.WT}III Tor knockdown y,v 35316 y[1] sc[*] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.GL00222}attP2 tra  F overexpression Sam 4590 w[1118]; P{w[+mC]=UAS-tra.F}20J7 tra knockdown y,v 28512 y[1] v[1]; P{y[+t7.7] v[+t1.8]=TRiP.JF03132}attP2 !))!on standard cornmeal-molasses food medium at 24ûC on a 12/12 hours light/dark schedule.   Each overexpression line was introgressed into the common wild-type background, Samarkand (Sam), for seven generations to avoid background effects. We performed introgressions by backcrossing red-eyed females (carrying the insertion element) to males of the Sam background. Since RNAi knockdown lines utilized TRiP insertional elementsÑall of which are already in a common background, Òyellow, vermillionÓ (y,v), and lack readily available identifiersÑwe did not introgress these into Sam.   Nutritional SSD Assay For this assay, all flies used were from the Sam genetic background. Food was prepared ahead of time for four separate treatment groups: 100, 10, 5, and 2% food. Each group consisted of a specific percentage of standard cornmeal-molasses medium mixed with a 2% agar solution. All food treatments yielded food of similar volume and consistency, but varying greatly in nutrition (see Stieper et al., 2008). For each experimental group, we allowed 8-10 flies of each sex to oviposit on standard apple juice-agar plates for 12 hours. After oviposition, eggs were washed and distributed to experimental food vials. All surviving pupae were collected and placed in individual 1.5ml Eppendorf tubes along with a square of ideally-moist Kimwipe. Twelve hours after eclosion, flies were harvested with 70% ethanol, and then allowed to dry overnight. Individual masses were recorded for flies on a Mettler Toledo XPE26 Microbalance.   !)*!Determining candidate genes by pathway Positive vs Negative growth regulators Ð Within our candidate pathways, genes can either positively or negatively affect growth through their interactions. Any individual component can either be a positive regulator (one that increases expression of the pathway and increases growth), or a negative regulator (one that represses activity of the pathway and reduces growth). Most of the genes within our candidate pathways are positive regulators, meaning that gene expression and phenotype are expected to positively covary. For negative regulators, the covariance between gene expression and phenotype is reversed (negative). In the case where genes are negative growth regulators, their expected effects will be explicitly made clear, otherwise it can be assumed that a gene is a positive growth regulator.   IIS pathway Ð To increase IIS pathway activity we chose to overexpress InR and Akt and knock down expression of Foxo with RNAi. Foxo is a negative growth regulator, and so reducing its expression is theoretically equivalent to increasing expression of positive growth regulators. InR is the first component in this pathway, and is a positive growth regulator, so an increase in InR also increases Akt and deactivating Foxo protein activity. Likewise, an increase in Akt will also inactivate Foxo. To decrease IIS activity, we did the opposite by knocking down InR and Akt with RNAi and overexpressing Foxo. Overexpression of Foxo will lead to growth suppression, the same result expected by knocking down InR and Akt. See Figure 1.2 for a summary pathway diagram.  !)+!TOR pathway Ð To increase TOR pathway activity we chose to overexpress the positive growth regulators Tor and S6K. Since Tor is upstream of S6K, an increase in Tor should cause a corresponding increase in S6K protein activity. To decrease TOR activity, we knocked down expression of Tor and S6K with targeted RNAi. See Figure 1.2 for a summary pathway diagram.  Ras pathway Ð To increase Ras pathway activity we chose to overexpress the positive growth regulator Ras. To decrease Ras activity, we knocked down expression of Ras with targeted RNAi. See Figure 1.2 for a summary pathway diagram.  Hh pathway Ð To increase Hh pathway activity we chose to increase expression of both Sxl and full length Ci and knock down expression of the negative growth regulator ptc. Normally, ptc inhibits the transcription of full length Ci. By knocking down ptc with a targeted RNAi, we will increase the positive growth regulator Ci. Increasing Sxl will catalyze this process, further increasing activity. To decrease Hh activity, we knocked down Ci and Sxl and overexpressed ptc. See Figure 1.3 for a summary pathway diagram.  SD pathway Ð To increase female-specific expression of the SD pathway we chose to overexpress the female isoforms of Sxl, tra, and dsx. Each component is expected to feminize males, while leaving females unaffected. To increase male-specific expression of the SD pathway we chose to knock down genes Sxl, tra, and dsx and also overexpress the dsx male isoform. Since male isoforms of each of these genes (except !),!dsx) are essentially non-functional, these crosses were expected to masculinize females while leaving males unaffected. See Figure 1.1 for a summary pathway diagram.  The above experiments would allow us to overexpress or knockdown individual components of each candidate pathway. Unfortunately, this method does not allow us to control the level to which we are actually affecting activity. In some cases, a full knockdown or overexpression is expected to be lethal. Only by explicitly controlling the level of expression in each pathway could we confidently conclude the effect each gene is having (without inducing lethality).   Titrating gene expression Each cross was initiated by crossing virgin Actin5c-GeneSwitch-Gal4 females with males of each genetic treatment. The way the GeneSwitch system works is very similar to the traditional Gal4/UAS system in Drosophila with one important caveat: the GeneSwitch driver is condition-dependent and will only activate transcription once bound to a specific ligand, mifepristone (RU486). Because RU486 is synthetic progesterone, and progesterone is a mammal-specific hormone, the addition of RU486 to the Drosophila diet is believed to be biologically inert. Since RU486 additively activates the GeneSwitch-Gal4, the amount of RU486 ingested by each larva corresponds to activation level of each response gene.  !)-!Overexpression and knockdown lines were carefully chosen to affect only the gene of interest in each cross. The goal of each was to demonstrate the necessity of a gene for SSD (knockdown crosses) and also the sufficiency of a gene to increase SSD (overexpression crosses). Overexpression lines were chosen to drive increased expression of the common, wild-type allele of each gene. Similarly, knockdown lines were chosen to dampen the expression of genes with a targeted gene-specific RNAi. In this way, overexpression crosses would increase the production of target genes, while knockdowns would cause a decrease in final gene product.   Titration of gene expression was achieved through the combined use of the Actin5c-GeneSwitch-Gal4 driver and introduction of a synthetic hormone, RU486, into the food ingested by larvae. Each gene-treatment contained each of the following sub-treatments: high dose RU486 (1.25µM), low dose RU486 (0.75 µM), ethanol solvent without RU486, or no addition to food.   Ultimately, the concentrations of RU486 doses were based on a literature search for appropriate levels (Nicholson et al. 2008; Shen et al. 2009), which were further refined by preliminary experiments. To determine the ideal concentrations of RU486 that would be strong enough to drive expression without being lethal, we crossed flies bearing Actin5c-GeneSwitch-Gal4 with those bearing a copy of vg-RNAi, reared them on varying doses of RU486, and assessed the resulting wing phenotypes. Since the vestigial gene (vg) is necessary for proper wing development, its absence (caused by driving expression of vg-targeted RNA interference) causes a distinct ablation of the !*.!wings; fortunately, this wing ablation is dose-dependent. High doses of RU486 were therefore expected to completely ablate adult wings, while the wings of low dose flies were expected to show only partial ablation. As expected, low doses of RU486 in vg-RNAi crosses typically yielded flies with wings of moderate vg phenotype, while high doses usually caused a much more severe phenotype (though, penetrance was variable).   Experimental Food Preparation All food used in our experiments was based on standard cornmeal-molasses food medium with extra yeast added to the surface to stimulate mating and oviposition. Food was refrigerated in small batches until needed, then melted down and, once cool enough, mixed with the proper dose of RU486 solution. Since RU486 is ethanol-soluble, we prepared solution for each batch based on experimental group, varying the amount of RU486 and ethanol to maintain a constant volume across groups. We then used an immersion blender to thoroughly mix the food to ensure an even distribution of RU486 solution, before pouring a standard amount of food into each vial.   Experimental Fly Crosses All crosses were maintained at 24C with a 12/12 hour light/dark schedule. For all treatments, we crossed four virgin female Actin5c-GeneSwitch-Gal4 flies with four male flies of the appropriate treatment group. Parental flies were then allowed to oviposit on the RU486 food medium, allowing their progeny to be raised completely on the respective experimental food. All vials were monitored for egg density to ensure that no !*%!vials were under-crowded (20 eggs) or over-crowded (50 eggs). Once pupae began to sclerotize (red eyes and black wings visible through pupal case), they were collected and placed into individual 1.5ml Eppendorf tubes along with a damp square of kimwipe; kimwipe squares were sufficiently damp to keep flies hydrated until they were old enough to collect and store in ethanol, but not so damp that flies would drown. Crosses typically yielded between 15-30 adult males and females, but varied by treatment. Those that produced inadequate numbers of males or females were omitted (e.g. weak Foxo knockdown). Once each adult fly had completely sclerotized, based on the presence of a fully formed adult wing, we filled each vial with ~1ml of 70% ethanol to preserve each fly along with its own pupal case. Body size estimates were based on pupal area, which has been shown to be an ideal proxy for overall body size (unpublished data; Stillwell, Dworkin, Shingleton, & Frankino, 2011). Pupal area was calculated by imaging dry pupal cases using the methods in (Stillwell et al. 2011; Tang et al. 2011).   Analysis of SSD SSD was calculated based on pupal size using the same index as in Chapter 2.   The direction of effect of each geneÕs effect was assessed by means of multiple statistical models. In all cases, each treatment was assigned a directional value corresponding to the expected level of gene expression, such that a high dose of RNAi was assigned -2, low dose of RNAi was assigned a -1, control flies were assigned 0, low dose overexpression was assigned +1, and high dose overexpression was assigned !*&!+2. Using these values, we were able to calculate the additive effects of gene expression as it was either added or subtracted from the control. The specific models we used (in R) were as follows: PupalSize ~ Sex + ExpressionLevel + Sex:ExpressionLevel This model was run twice: once as a linear model with ExpressionLevel as a numerical variable (which assumes a linear relationship between expression level and pupal size) and once with it as a categorical (which does not assume linearity or rank order). Significance values were collected from both versions under the following scenarios: standard model, where additive effects of expression from lowest to highest expression levels (using all values), and the perturbation models, where absolute change from control (using the absolute value of all values, such that a -1 was coded as +1) was used to assess the similarity of effect in either direction.   Significance values were calculated for differences between one treatment at a time to and each control group separately. FisherÕs summation of P-values was used to summarize the effect of a given gene manipulation on each line in the control pool. For example, the InR strong knockdown treatment was compared to each control, yielding a significance value for each; the resulting list of significance values were then summarized using FisherÕs summation, yielding the likelihood that InR strong knockdown was different from the control group.  All statistical analyses were conducted using R statistical software (version 3.2.2). Calculation of pupal area was conducted using the custom macro used in Tang et al. !*'!(2011). All error bars are 95% confidence intervals generated by non-parametric bootstraps.  Results Nutrient Sensitivity of SSD As expected, reduced nutrition caused an elimination of SSD (Figure 3.1). Individual t-tests demonstrate highly significant SSD for 100% (P < 0.001) and 10% (P < 0.05), but no significant differences at 5% (P > 0.5) or 2% (P > 0.4) nutrition. Additionally, linear models (P < 0.01), analysis of variance (P < 0.01), and analysis of covariance models (P < 0.01) show that food quality has a major effect on SSD.   Each genetic cross was carefully constructed to either increase or decrease the activity of each candidate pathway. We increased pathway activity by driving expression of positive regulator of each pathway or by knocking down expression of negative regulators of each pathway, using RNAi. Conversely, we decreased pathway activity by knocking down expression of positive regulators using RNAi, or by driving expression of negative regulators (see Figure 3.2 for example and Table 3.2 for confidence intervals).  Titrating gene expression: IIS pathway For the IIS pathway, Inr and Akt are positive regulators of pathway activity, while Foxo is a negative regulator of activity. Previous research indicates that suppression of IIS reduces SSD (Bıhni et al. 1999; Testa et al. 2013; Rideout et al. 2015), and, consistent   !*(! Figure 3.1. Nutritional Quality vs Weight for both male (blue) and female (red) Sam flies. Lines show regression estimates for each sex. As nutritional quality decreases, so does the difference between male and female body size. Note, once nutritional quality is reduced below 10%, SSD is effectively eliminated (P < 0.05).    0204060801000.050.100.150.20Nutritional Quality vs WeightNutritional Quality (% food)Weight (mg)femalemale!*)!  Figure 3.2.  Reaction norm for SSD (above) and body size in males and females (below) in sex determination (SD) pathway manipulations. Genetic manipulations reveal that as expression of Sxl and dsx increase, SSD is reduced. For Sxl, this is by disproportionately affecting female body size, but in dsx this is due to increased dsx expression causing both sexes to develop intersex body sizes. In nearly all treatments, tra eliminates body size.    !!0.10!0.050.000.050.100.150.20Reaction Norms for SD pathwayWhole Body SSDRelative Expression Level!2!1012dsxdsx!!!!frufru!!!!sxlsxl!!!!tratra!!!!!!!!controlsfemalemalep > 0.05p < 0.05dsxfrusxltra1.52.02.53.0Pupal Area (mm2 )!2!1012dsx!!!!!!!!!2!1012fru!!!!!!!!!2!1012sxl!!!!!!!!!2!1012tra!!!!!!!!!**!Table 3.2. Estimates for SSD and corresponding confidence intervals for all treatment crosses. Treatment is given in the form of Ògene to manipulateÓ _ Òdirection and magnitude of manipulation,Ó such that Òakt_-2Ó corresponds to a strong knockdown of akt and Òakt_1Ó corresponds to a weak overexpression of akt. 95% confidence intervals were generated using non-parametric bootstraps.     Treatment SSD low 95% CI high 95% CI control_pool 0.13015 0.02281 0.22959 akt_-2 0.10785 0.05377 0.15818 akt_-1 0.07974 0.02204 0.14277 akt_1 0.10844 0.01089 0.21374 akt_2 0.09142 0.00802 0.17537 ci_-2 0.0865 -0.01663 0.20296 ci_-1 0.02358 -0.04963 0.09193 ci_1 0.16113 0.0862 0.24504 ci_2 0.14584 0.07303 0.22845 dsx_-2 0.12973 0.02077 0.27605 dsx_-1 0.17058 0.08808 0.25583 dsx_1 0.12544 0.00947 0.2596 dsx_2 0.0478 -0.04057 0.14703 foxo_-2 0.13883 -0.00397 0.28284 foxo_1 0.04405 -0.02863 0.11647 foxo_2 0.10219 0.04705 0.1548 fru_-2 0.18884 0.09059 0.29949 fru_-1 0.07781 -0.01746 0.17407 fru_1 0.17414 0.08885 0.2624 fru_2 0.13071 0.03415 0.23956 inr_-2 0.10107 0.03982 0.16093 inr_-1 0.03014 -0.02886 0.0796 inr_1 0.09294 -0.12688 0.34509 inr_2 0.08778 -0.15454 0.35885 ptc_-2 0.11948 0.06427 0.167 ptc_-1 0.09009 0.02287 0.15809 ptc_1 0.06426 0.01348 0.11772 ptc_2 0.13618 0.06047 0.21706 !*+! Table 3.2. (contÕd)                 ras_-2 0.17031 0.1041 0.24263 ras_-1 0.12479 0.05399 0.20805 ras_1 -0.10125 -0.24716 0.01256 ras_2 0.10593 0.02712 0.17878 s6k_-2 0.01559 -0.05578 0.09016 s6k_-1 0.15234 0.01497 0.34868 s6k_1 0.1048 -0.05056 0.28696 s6k_2 0.05893 -0.00787 0.12934 sxl_-2 0.16837 0.09873 0.24015 sxl_-1 0.13627 0.06474 0.19776 sxl_1 0.09259 0.03229 0.15113 sxl_2 0.11579 0.00312 0.23073 tor_-2 0.14399 0.09261 0.20266 tor_-1 0.13997 0.06834 0.20438 tor_1 0.08637 0.01393 0.16531 tor_2 0.05565 -0.00298 0.11705 torted_1 0.1914 0.12913 0.26932 torted_2 0.18598 0.09591 0.28152 tra_-2 -0.02437 -0.10875 0.05601 tra_-1 0.01724 -0.14088 0.15869 tra_1 0.16489 0.10024 0.23537 tra_2 0.00358 -0.08119 0.08748 !*,!with these studies, we found that weak knockdown of InR also reduced SSD (Figure 3.3, weak knockdown; P = 0.05). However, an increase in InR expression did not increase or decrease SSD. Further, changes in the expression of both Akt and Foxo had no effect on SSD or body size (Figure 3.3).   Titrating gene expression: TOR pathway In the TOR signaling pathway, both Tor and S6K are positive growth regulators and therefore were expected to behave as such. Instead, we found that the Tor has a negative effect on SSD with increasing expression (P < 0.001; Figure 3.4). Conversely, the downstream element, S6K, exhibits a significant reduction in SSD in strong knockdowns (P < 0.001). Strong knockdowns reduce SSD primarily through impacting female size relative to male size. Knock down of Tor and overexpression of S6K has no effect on body size or SSD.  Titrating gene expression: Ras pathway While both male and female body size tend to increase with Ras expression, Ras exhibits a negative effect on SSD with increasing expression (P < 0.01; Figure 3.5). Specifically, it is the weak (P < 0.001) overexpression line of Ras that causes the most marked reduction in SSD. Thus, as Ras activity increases body size increases, but female size reacts inconsistently, thus SSD actually decreases.    !*-! Figure 3.3. Reaction norm for SSD (above) and body size in males and females (below) in Insulin and Insulin-like Signaling (IIS) pathway manipulations. Of all possible treatment groups, only the weak knockdown of InR significantly reduced SSD. In general, SSD was relatively consistent across expression levels, despite body size for males and females being more labile.   !!0.10!0.050.000.050.100.150.20Reaction Norms for IIS pathwayWhole Body SSDRelative Expression Level!2!1012aktakt!!!!foxofoxo!!!inrinr!!!!!!!!controlsfemalemalep > 0.05p < 0.05aktfoxoinr1.52.02.53.0Pupal Area (mm2 )!2!1012akt!!!!!!!!!2!1012foxo!!!!!!!2!1012inr!!!!!!!!!+.! Figure 3.4. Reaction norm for SSD (above) and body size in males and females (below) in TOR signaling pathway manipulations. Tor demonstrates a negative relationship with SSD; as Tor expression increases, SSD decreases due to a reduction in female relative to male body size. S6K demonstrates a different pattern, where strong knockdowns eliminate SSD. Since TorTED is a dominant negative allele of Tor, its sign is effectively flipped. Its effects are consistent with, but stronger than, those of Tor manipulations.    !!0.10!0.050.000.050.100.150.20Reaction Norms for Tor pathwayWhole Body SSDRelative Expression Level!2!1012s6ks6k!!!!tortor!!!!torted!!!!!!controlsfemalemalep > 0.05p < 0.05s6ktortorted1.52.02.53.0Pupal Area (mm2 )!2!1012s6k!!!!!!!!!2!1012tor!!!!!!!!!2!1012torted!!!!!+%! Figure 3.5. Reaction norm for SSD (above) and body size in males and females (below) in Ras signaling pathway manipulations. As Ras expression increases, SSD is reduced. Surprisingly, weak overexpression causes a reversal of SSD through an exaggerated female reduction in body size.  !!!!0.100.000.100.20Reaction Norms for Ras pathwayWhole Body SSD!2!1012rasras!!!!!!!!controlsfemalemalep > 0.05p < 0.05ras!1.52.5Pupal Area (mm2 )!2!1012ras!!!!!!!!!+&!Titrating gene expression: Hh pathway For the Hh signaling pathway, we expected similar results to the nutrient-sensitive growth pathways. Normally, ptc inhibits smoothened (smo) activity, which is a positive regulator of full-length Ci. Once Hh ligands bind to ptc, it derepresses smo and full-length Ci is expressed. Since ptc is a negative growth regulator, knockdowns are expected to increase SSD and overexpressions to decrease SSD. We do observe the trend that ptc decreases overall body size, but SSD remains unaffected (Figure 3.6). The other genes, Ci and Sxl, were then expected to function as positive growth regulators. Indeed, with Ci, we see a strong positive relationship with SSD. Ci strong-knockdowns reduce and weak-knockdowns eliminate (P < 0.001) SSD, which then increases with additional overexpression. The results of Sxl are described below.  Titrating gene expression: Sex determination pathway (SDP) Finally, the sex determination pathway crosses were set up in such a way that the resulting offspring should be either feminized or masculinized, depending on the cross. The way that the SDP works is fundamentally different than growth-regulating pathways; rather than components activating and/or repressing downstream genes, each element of the SDP alternatively splices the protein products of downstream elements. Because knockdowns in female-specific isoforms would cause male-specific versions to be spliced, knockdowns were expected to masculinize females. Conversely, overexpressions of female-specific isoforms were expected to feminize males, by driving expression of the female-specific isoforms. In both cases, we expect to see a reduction in SSD.  !+'! Figure 3.6. Reaction norm for SSD (above) and body size in males and females (below) in Hedgehog (Hh) signaling pathway manipulations. Within individual treatment groups, SSD is only significantly affected in weak Ci knockdowns where SSD is eliminated. Regardless, there is a positive significant trend between Ci expression and SSD. This trend is countered by a lack of significance in ptc and a significant negative trend in Sxl expression.     !!0.10!0.050.000.050.100.150.20Reaction Norms for Hh pathwayWhole Body SSDRelative Expression Level!2!1012cici!!!!ptcptc!!!!sxlsxl!!!!!!!!controlsfemalemalep > 0.05p < 0.05ciptcsxl1.52.02.53.0Pupal Area (mm2 )!2!1012ci!!!!!!!!!2!1012ptc!!!!!!!!!2!1012sxl!!!!!!!!!+(!At the top of the SDP is Sxl, whose name properly suggests that loss-of-function and constituently expressing mutants of the gene are lethal for both females and males, respectively. It was expected that females would respond to Sxl knockdowns by masculinizing (decreasing) their size, leaving males unaffected. Ubiquitous overexpression and knockdowns in the larval stage, however, were not lethal, but yielded counterintuitive results. While expression of Sxl is normally only found in females, we found that when its expression was increased, Sxl reduced and masculinized female body size, displaying a fully additive negative relationship with SSD. Starting with strong knockdowns (that increase SSD; P < 0.02), as Sxl expression increases, body size and SSD decrease. Males appear to be relatively resistant to perturbations in Sxl activity; it is fluctuations in the female body size that cause these changes in SSD.   Furthermore, both downstream elements tra and dsx also have reducing effects on SSD (Figure 3.2). Statistical models suggest that as tra increases, so does SSD, but this is likely due to the weak overexpression being the only treatment failing to eliminate SSD. In the perturbation statistical modelsÑthe version of our models where absolute expression, rather than overall expression, is taken into accountÑwe see a highly significant trend of perturbations causing reduction in SSD (P < 0.0001). Strong perturbations (-2 or 2) also eliminated SSD (P < 0.0001). As expected, tra is both necessary and sufficient to alter SSD in every case except for weak overexpression. Essentially, as tra increases or decreases, we expected SSD to be lost by either   !+)!feminizing males or masculinizing female flies, respectively. Certainly, females are masculinized in tra knockdown crosses, but they are also masculinized in strong overexpression crosses as well. Dsx also has a directional effect, where SSD increases with decreasing dsx activity. In spite of this, no individual doses had a significant effect in dsx crosses. Since fru is ostensibly only expressed in the nervous system and only affects male behavior, it was not expected to contribute to SSD. As expected, fru had no significant effects on SSD, though overall body size appears to fluctuate somewhat with dose.   Discussion Effect of Nutrition and Nutrient-Dependent Pathways on SSD Our results support the hypothesis that SSD is nutrition-dependent. As food quality decreases below 10%, SSD is eliminated. These data are consistent with predictions made by Bonduriansky (2007) about the condition-dependence of SSD with respect to nutrition. Clearly the mechanisms of SSD are linked to those of nutrient-sensitive pathways. Because our low quality food treatments are ostensibly phenocopies of nutrient-sensitive growth pathway mutants (Testa et al. 2013), we can confidently conclude that SSD is regulated (at least in part) by nutritional responses.   Our primary hypothesis was that, since IIS was a nutrient-sensitive pathway, IIS would be both necessary and sufficient for SSD. IIS reduction experiments verify that IIS signaling is necessary for SSD, but unfortunately our initial hypothesis that IIS activity is sufficient to increase SSD cannot be supported. Recent evidence suggests that IIS !+*!activity may be responsible for differences in SSD (Rideout et al. 2015). Flies with a mutation of the negative IIS regulator, PTEN, were found to rescue female body size in artificially masculinized crosses (via tra knockdown), placing IIS downstream of tra activity. When individual components of the IIS pathway are overexpressed throughout the body in both sexes, however, a similar trend was certainly not observed. The reason for this discrepancy is as of yet unclear. Notably, overexpression in the IIS pathway causes neither an increase nor decrease in SSD. This is important to note for all of our data. Regardless of the effect of over- or under-expression on body size, it is strictly the difference in male and female body sizes and the direction of effect that is of most relevance to this study. Thus, even when overexpression causes growth reduction, this reduction is not sex-specific.   In the TOR pathway, Tor shows a significant trend of decreasing SSD with increasing expression. The strong knockdown of S6K supports this claim: S6K is necessary for SSD, but when overexpressed ubiquitously, it also appears to decrease SSD (though not significantly so; P = 0.06). Our evidence from S6K knockdowns are consistent with previous data for overall body size, where the sex-specific effects were previously unknown (Montagne et al. 1999). The reason for the discrepancy between Tor and S6K data is unclear. Given that S6K knockdown reduces female size and Tor overexpression increases male size, it is plausible that sex-specific expression of TOR signaling, rather than sex-specific sensitivity to it, is regulating SSD. Further data are needed to assess this hypothesis.   !++!For Ras signaling, our only candidate gene was Ras itself. Ras demonstrates a similar relationship to TOR signaling in that increased expression caused a reduction in SSD. This is strongest in the weak overexpression lines, where SSD is reversed, due to females losing considerably more size than males (Figure 3.5).  Our data support the claim that nutrient-sensitive growth pathways (defined as IIS, TOR, and Ras) tend to be necessary for SSD. Individual components of IIS and TOR signaling both demonstrate their necessity for SSD. If SSD is indeed nutrient-sensitive, these results are to be expected. Surprisingly, our ubiquitous knockdowns do not appear to function quite the same way that traditional mutants do. For example, Testa et al. (2013) demonstrated that IIS mutants eliminated SSD, but only the weak knockdown of InR had a significant effect on SSD in our crosses. Since it is still unknown how mutation of other growth regulation pathways affects SSD, we cannot fully interpret our results until more data are collected, though we suspect the results will be similar to InR.   In general, it appears to be much more difficult to increase SSD than it is to decrease it. An alternative hypothesis to ÒSSD is influenced by sex-specific differences in sensitivity to signaling pathwaysÓ could be that for signaling systems to produce SSD, there must be a very specific balance of inputs; any other general state of signaling (within reasonable limits) leads to the ÒdefaultÓ of sex-invariant sizing. In this case, it is possible that our genetic manipulations have simply disrupted normal developmental processes leading to reduction in the ability to produce sex-specific phenotypes, rather than direct !+,!manipulation of SSD. Perhaps a less directed set of candidate genes might yield more unbiased results. Finally, given how high the environmental and background variation tend to be in populations (as seen in Chapter 4), we may need to rethink how we approach these sorts of questions in future studies.   Our data generally support Takahashi and Blanckenhorn's (2015) claim that deficiencies in most genes tend to reduce rather than increase SSD. Of the 15 candidate genes, no individual treatment was sufficient to increase SSD (while the linear statistical model found that increasing Ci expression increased SSD, no individual treatments were significant. We believe that this effect is a statistical artifact of knockdowns reducing SSD, while overexpessions had no effect). Individual dosage effects revealed that, in the case that a gene caused a significant effect, that effect was always to decrease SSD. It seems likely then, that directional effects on SSD are caused more by the uniform reduction of SSD on one end of the dosage spectrum (knockdown or overexpression) than they are by reduction on one end and increase on the other.  Overexpression of nutrient-sensing growth pathways reduces body size One of the most surprising results of these experiments is the fact that nutrient-sensitive growth pathway (IIS, TOR, Ras) overexpression crosses tended to decrease overall body size. This is surprising, because previous research has shown that overexpression of individual components of the IIS pathway, such as InR, Akt, and/or PI3K, has the capacity to increase organ size when overexpressed in one organ (Aron et al. 2010; Gokhale et al. 2016). Indeed, these findings are somewhat of a novelty in the !+-!Drosophila literature; no study to date has investigated the effects of ubiquitous overexpression of multiple growth control pathways. Our driver, Actin5c-GeneSwitch, causes expression of UAS response elements everywhere in the body that Actin, a critical cytoskeletal element of most cells, is expressed (therefore it is referred to as ubiquitous). One such study did find that a ubiquitous overexpression of InR caused a reduction in growth rate, presumably in both sexes (Wong et al. 2014), but seemingly no further examples exist. Our data suggest that overexpression in nutrient-sensitive growth pathways is more nuanced that simply overexpressing activity in a single organ.   One explanation for this phenomenon could be that ubiquitously driving expression of each gene may be activating negative feedback loops in their respective pathways. Prolonged exposure to insulin is known to cause insulin-resistance in mice, effectively deactivating the IIS pathway (Manning 2004; Um et al. 2004; Rui et al. 2001). According to Rui et al. (2001), serine (Ser307) phosphorylation at the insulin receptor substrate (IRS) is a common mechanism to counter-regulate insulin signaling in scenarios such as overstimulation. These negative feedback loops are also associated with other nutrient-sensitive growth pathways, including TOR (Manning 2004; Shah et al. 2004; Um et al. 2004) and Ras (Dougherty et al. 2005; Macrae et al. 2005). If negative feedback loops are indeed being activated in overexpression lines, then insulin resistance is a likely cause for the anomalous decreased body size in overexpression lines.   !,.!Another explanation could be that tumor suppressor genes are repressing the increased growth caused by our overexpression crosses. Tumor suppressors are genes known to regulate overgrowth by regulating cell cycle arrest and apoptotic cell death within many different nutrient-sensitive growth pathways. Since the goal of our overexpression crosses was to increase growth, and the primary function of tumor-suppressor genes is to stop over-growth, it is possible that tumor suppressor genes may be responsible for the reduction in body size in our overexpression crosses. Relevant examples within the IIS, TOR, and Ras signaling pathways, respectively, include: PTEN (Di Cristofano & Pandolfi 2000), TSC1/TSC2 complex (Garami et al. 2003), and RASSF1 (Armesilla et al. 2004). It is therefore plausible that tumor suppressor activity is suppressing growth ubiquitously in response to sustained activation of nutrient-sensitive growth pathways.   Even if overexpressing genes within nutrient-sensitive growth pathways causes a reduction in size, our data still do not conform to our expectations of how growth pathway mutants function. In InR mutants, SSD is eliminated. We do not see a similar reduction in SSD for overexpression crosses even though body size is often reduced below that of RNAi knockdowns. One would expect that in the above scenarios SSD would decrease as though that pathway were knocked down. Instead, what we see is a reduction in body size with SSD behaving more-or-less independently of body size.   Perhaps the most plausible explanation is that, growth suppression (either by insulin resistance or tumor suppressor) might be physiologically different in males and females rather than acting as a sex-uniform suppression. In other words, males and females !,%!may be responding independently not only to growth regulators (those that facilitate growth), but also to growth mediators (those that moderate growth). Indeed, Testa et al. (2013) demonstrates that there is already precedence for this explanation as males and females both gain and lose mass at differing rates. This explanation would account for both the reduction in body size and maintenance of SSD in growth pathway overexpression crosses.   Understanding the mechanisms that cause our observed body size reduction is both interesting and important, but well beyond the scope of this study. While this discovery is significant, its impact is no less diminished if we do not have a molecular explanation for this effect. Further experiments are certainly required to test these hypotheses. By individually testing the hypotheses above, we may gain some insight into the molecular regulation of this phenomenon.   Alternative interpretations of data However interesting the above results may be, it is important to consider that the effects observed in these experiments may have alternative or non-biological explanations. Since we have no molecular data about how these perturbations are influencing the activity of each signaling pathway, we must be cautious in interpreting our results. Potentially, changes in transcript abundance may be insufficient to substantially alter activity for some of these genes. However, for some of these genes, previous evidence has shown that genetic manipulations like this have been shown to work (Aron et al. !,&!2010; Dworkin et al. 2011; Rideout et al. 2015).  It is possible that strong knockdowns and over-expressions are simply generating conditions that are outside of normal pathway functioning and thus ÒbreakingÓ the established systems. If this is the case, we might expect to see a stronger effect in the weak manipulations than the strong ones, because the weak manipulations are the only ones affecting pathway activity as expected. In some way, we know that these manipulations are definitely physiologically altering the phenotype (see Table 5.1). Additionally, we see that most of the pupal sizes also occupy a similar range to those generated in Tang et al. (2011), where Foxo expression was over-expressed and knocked down.   There are also other, non-biological, explanations for certain trends in our data. First, we know that sample size can have large effects on analysis. Particularly, in the case of Ras weak over-expression, we know that sample sizes are lower than average and that a couple of low female outliers are pulling down the mean, causing female size to be lower than male. In instances like this, the trends we see in Ras overexpression may be more likely due to statistical artifacts rather than true biological effects. Additional data would confirm that this effect is, in fact, due to sampling error and low sample size.   Even though the progesterone mimic, RU486, is not found in insects, it could potentially be having some effect on dimorphism. We did control for these effects, but since our !,'!analyses involved analyzing the effects of each treatment against a pool of controls, these effects may not specifically be explicitly dealt with.   Hh signaling and its link to the SDP One of the stronger competing hypotheses to explain the mechanism of SSD was the Hh-Sxl hypothesis put forward by Horabin (2005). The authors demonstrated that ectopic Sxl protein was sufficient to cause overgrowth in female wing imaginal discs, but not in males. This sex-specific overgrowth is supposedly caused by an increase in full length Ci (Horabin 2005). While our results indicate a statistical trend for increasing Ci expression causing a corresponding increase in SSD, it is unclear if our experimental over-expression of full length Ci is leading to the activator Ci protein. Ogden et al. (2004) suggests that the full length activator version of full length Ci may not be produced in the absence of Hh signaling ligand. In this case, we may not be able to make strong statements about the overexpression of Ci in our experiments. Regardless, there is no evidence for any other factor within the proposed Hh signaling pathway, including the proposed link between the Hh pathway and Sxl of the sex determination pathway. This might be due to the fact that we used a ubiquitous driver, rather than an organ-specific driver. While female wings are likely to be in the presence of greater amounts of Sxl protein, so is the rest of the body. Indeed, it appears that ubiquitous expression of Sxl in females has the exact opposite effect of organ-specific overexpression of Sxl. Therefore, we suspect that the Hh-Sxl pathway is unlikely to be regulating whole body SSD.   !,(!At what level in the SDP are sex-specific body sizes determined? The earliest accounts of tra demonstrated by loss of function mutants for tra had no effect on male development, while females were found to be transformed into males that were wild-type in every respect except for body (and genital) size (Sturtevant 1945). For nearly 70 years the conventional wisdom was that SSD was controlled at some level in the SDP other than tra (Sturtevant 1945; Baker & Ridge 1980; Cline & Meyer 1996; Salz & Erickson 2010; Cline 1984; Horabin 2005) . More recent literature has challenged the notion that tra is unable to control SSD (Rideout et al. 2015). Our data support the recent literature in determining that tra is the component of the sex determination pathway most active in regulating SSD. Indeed, it is tra that shows the greatest amount of support for regulating SSD. While dsx and Sxl also demonstrate that they also decrease SSD with increasing expression, they are not necessary for SSD.   As male and female isoforms of dsx are expressed, SSD is lost. The response of SSD is nearly identical regarding male vs female isoforms; both produce equally intersex flies (with respect to size), thus we reported both together. These results would be expected if dsx were positively regulating SSD. Indeed, while many studies have investigated the phenotypic effects of dsx expression (Shirangi et al. 2006; Erdman & Burtis 1993; Kijimoto et al. 2012; Burtis & Baker 1989; Waterbury et al. 2000; Williams et al. 2008), few have examined its effects on size (Kijimoto et al. 2012). Our data suggest that correct sex-specific expression of dsx is required to produce correct male and female sizes; when either sex-specific isoform is inappropriately expressed, SSD is lost. !,)!Interestingly, a lack of dsx is expected to masculinize female body size. Since it does not, dsx cannot be necessary for SSD.   Upstream of both dsx and tra is Sxl, which, according to some studies (Horabin 2005; Salz & Erickson 2010), should be sufficient to increase SSD. In fact, our data directly contradict these claims. As Sxl increases, not only does SSD decrease (and vice versa), but whole body size decreases as well. This effect appears to be the result of greater female sensitivity to Sxl, where decreased levels produce larger females and increased levels smaller females.   As discussed in Carreira et al. (2009), body size regulation is far too complex to be accounted for by few genes of large effect. Both Carreira et al. (2009) and Takahashi et al. (2015) demonstrate that it is much more common for genetic perturbations to decrease SSD. Our data are less consistent with ÒsensitivityÓ hypothesis proposed by Takahashi et al., where it would be expected that both sexes respond unequally to genetic perturbations. Regardless, our data certainly support the view that body size regulation is immensely complex.      !,*! CHAPTER 4:  Testa, N. D. & Dworkin, I. The sex-limited effects of mutations in the EGFR and TGF-" signaling pathways on shape and size sexual dimorphism and allometry in the Drosophila wing. Dev. Genes Evol. (2016). doi:10.1007/s00427-016-0534-7 !,+!Introduction  In spite of our wealth of knowledge about the natural world, biologists continue to be fascinated by the prevalence of sexual dimorphism. Where sexual dimorphism is often found, it is most often subtle, despite important exceptions of sex-limited characteristics (Bonduriansky & Day 2003), or traits that are highly exaggerated in one sex, but not the other (Lavine et al. 2015). This is particularly evident for morphological traits that demonstrate sexual size (SSD) or sexual shape (SShD) dimorphism (Kijimoto et al. 2012). Within evolutionary biology, explanations for sexual dimorphism have focused on a number of mechanisms that are likely responsible for the origin and maintenance of sexual dimorphism (Reeve & Fairbairn 2001; Allen et al. 2011; Bonduriansky & Chenoweth 2009; Mank 2009; Cox & Calsbeek 2010; Hedrick & Temeles 1989; Shine 1989; Fairbairn & Blanckenhorn 2007) including sexual conflict, differences among the sexes in the variance of reproductive success leading to sexual selection (Fairbairn 2005), and sex specific aspects of natural selection (Preziosi & Fairbairn 2000; Ferguson & Fairbairn 2000). Despite this, our understanding of the genetic mechanisms that contribute to variation in sexual shape and size dimorphism is still lacking (Mank 2009; Blanckenhorn et al. 2007; Fairbairn & Roff 2006; Fairbairn 1990).  There is considerable experimental evidence demonstrating that patterns of SSD and SShD can be altered by influencing the condition of individuals (Bonduriansky & Chenoweth 2009; Bonduriansky 2007b). There has unfortunately been less success on directly experimentally evolving consistent changes SSD or SShD, with some notable !,,!exceptions where dimorphism evolved in response to selection on fecundity (Reeve & Fairbairn 1999) or due to experimental manipulation in the degree of sexual conflict (Prasad et al. 2007). There are even fewer instances where experimental evolution has been able to alter existing size/shape (allometry) relationships (Bolstad et al. 2015).   Despite previous difficulties with directly selecting for SSD or SShD, we still find evidence for genetic variation in SSD within a number of species (David et al. 2003; Merila et al. 2011). Several studies have utilized induced mutations (Carreira et al. 2011) or defined genomic deletions to examine patterns of SSD (Takahashi & Blanckenhorn 2015). They find that, in general, mutations tend to attenuate differences in SSD and sexual developmental timing difference. Interestingly, while ~50% of the random insertion mutations influenced size and shape, only half of those were consistent between males and females, suggesting considerable sex limitation of the mutational effects (Carreira et al. 2011).  With respect to the influence of mutations on sexual dimorphism, one important consideration is whether the mutations themselves are directly influencing aspects of sexual dimorphism. Alternatively, mutations may be influencing size and shape of the organism, but are modulated in a sex-limiting fashion. Arguably, it is difficult to distinguish between these possibilities, although for the purposes of this study, we consider a mutation to be modulated by the influence of sex if it influences size or shape as well as having an additional influence on sex (i.e. a sex-by-genotype interaction). !,-!The extent to which such mutations influence SSD and SShD remains poorly understood.  To address these questions, we examined the influence of characterized induced mutations that influence two signaling pathways important for wing development, Epidermal Growth Factor Receptor (EGFR), and Transforming Growth Factor - ", (TGF-"). The Drosophila wing is an excellent model for the study of SSD and SShD. First, as it is a premiere model system for the study of development, and as such a great deal is known and understood about the mechanisms governing overall growth and patterning (Garc™a-Bellido et al. 1994; Weinkove et al. 1999; Day & Lawrence 2000; Weatherbee et al. 1998). Additionally, Drosophila melanogaster and closely related species have a strong pattern of sexual size dimorphism for many traits (and overall body size), with wing size demonstrating some of the greatest degree of overall dimorphism (Testa et al. 2013; Abbott et al. 2010; Gidaszewski et al. 2009).  There is extensive variation for size and shape within and between Drosophila species, and for the extent of SSD and SShD as well (Gidaszewski et al. 2009). Importantly, the mutational target size for wing shape (Weber 2005) is high (~15% of the genome), thus providing plenty of opportunity for mutations to influencing shape, and potentially those modulated by sex.   In this study we utilize a previously published data set that examine the influence of 42 mutations in the EGFR and TGF-" signaling pathways when examined in a heterozygous state. We re-analyze this data set to examine the extent to which the mutations have sex-limited phenotypic effects that influence SSD or SShD. Furthermore !-.!we examine how patterns of allometric variation between size and shape are altered by both sex and wild type genetic background of the mutations. Despite most mutations having substantial phenotypic effects on either size, shape or both, only a small subset of them appear to have their effects modulated by sex, with respect to both direction and magnitude of effects. Furthermore, we demonstrate that the allometric relationship between size and shape is only subtly influenced by sex and genetic background for these alleles. We discuss these results within the context of sex-limited effects of mutations and their influence on SSD and SShD, and how to interpret allometric relationships between size and shape in Drosophila.  Materials and Methods Provenance of Samples The data used for this study was originally published by (Dworkin & Gibson 2006). We compared wings from flies across several treatment groups, including: sex, wild type genetic background (Oregon-R and Samarkand), progenitor line and genotype (mutant vs. wild type allele). Fifty different p-element insertion lines, each marked with w+, were introgressed into two common wild type backgrounds (Samarkand and Oregon-R), were used along with their respective controls. All wing data, in the form of landmarks, were collected from digital images, as detailed in Dworkin and Gibson (2006). For a more detailed description on the source of these strains and the experimental design, please refer to Dworkin and Gibson (2006).  !-%!Insertional mutations were selected from the Bloomington Stock Center and subsequently introgressed into two wild-type lab strains, Samarkand (Sam) and Oregon-R (Ore). Introgressions were performed by repeated backcrossing of females bearing the insertion to males of Sam and Ore-R. Females from replicate vials within each generation were pooled for the subsequent generation of backcrossing. Since both backgrounds contain a copy of the mini-white transgene, eye color for all flies lacking p-elements was white. Selection was therefore based entirely on the presence of the eye color marker, precluding unwitting selection for wing phenotypes. While the introgression procedure (14 generations of backcrossing) should make the genome of the mutant largely identical to that of the isogenic wild types, some allelic variation in linkage disequilibrium with the insertional element may remain. All experimental comparisons of mutant individuals were therefore made with wild-type siblings from a given cross and should share any remaining segregating alleles unlinked to the p-element. We separated mutants and their wild-type siblings by their corresponding mutant ÒlineÓ number (supplementary Table 1) to avoid these and potential Òvial effectsÓ. All crosses were performed using standard media, in a 25¡C incubator on a 12/12-hr light/dark cycle.  Two vials for each line were set up carefully to result in low to moderate larval density. The temperature of the incubator was monitored cautiously for fluctuations, and vial position was randomized daily to reduce any edge effects. After eclosion and sclerotization, flies from each cross were then separated into mutant and wild type individualsÑthose with and without the p-element-induced mutations, respectivelyÑ!-&!based on eye color and stored in 70% ethanol. A single wing from each fly was dissected and mounted in glycerol (see supplementary table 1B for sample sizes). Images of the wings were captured using a SPOT camera mounted on a Nikon Eclipse microscope. Landmarks (as shown in Figure 1) were digitized using tpsDig (v. 1.39, Rohlf 2003) software.   Our analysis necessitated that there be flies from each representative treatment group; those lines with flies missing (e.g. from one background or sex) were left out of the analysis. Of the original 50, 42 lines were ultimately used.  Analysis of Sexual Size Dimorphism (SSD) Centroid size (i.e., the square root of the sum of the squared distances from each landmark to the centroid of the configuration) was used as the size variable in our analyses. Individual size values for male and female within each line and background were taken from the coefficients of a linear model where centroid size was modeled as a function of genotype, sex and their interaction.   SSD was then calculated based on a common index, wherein the dimorphism is represented as the proportion of female size to male size (Lovich & Gibbons 1992; Smith 1999):    !!"!!!"#$!!"#$!!!  !-'!The resulting index represents the relative size difference between males and females where 0 indicates a complete lack of dimorphism and 1 indicates that females are 100% larger than males. Negative values represent male-biased dimorphism.   Analysis of Sexual Shape Dimorphism (SShD) Generalized Procrustes Analysis (GPA) was used to super-impose landmark configurations after correcting for position and scaling each configuration by its centroid size.  This procedure removes non-shape variation from the dataÑsize, orientation and position. From the nine two dimensional landmarks, we are left with 14 dimensions of variation, and thus applied a Principal Components Analysis (PCA) to the Procrustes coordinates (i.e., the shape coordinates after GPA) and the first 14 PC scores were used as shape variables in subsequent shape analyses.   Two different shape scores were used in this study: one to examine sexual shape dimorphism and one to assess the strength of the allometric relationship of shape on size. First, SShD was estimated using the tangent approximation for Procrustes distance (i.e. Euclidian distance) between the average of male and female wing shape for a given treatment. Additionally, we calculated shape scores from the multivariate regression of shape onto size based on Drake & Klingenberg (2008). Specifically we projected the observed shape data onto the (unit) vector of regression coefficients from the aforementioned multivariate regression. We used these shape scores and regressed them onto centroid size to approximate allometric relationships. Confidence !-(!intervals for SSD and SShD as well as allometric coefficients were generated with random non-parametric bootstraps, using 1000 iterations.   All significance testing for the analyses involving shape data was done with Randomized Residual Permutation Procedure (RRPP) as implemented in the geomorph library in R (Collyer et al. 2015). This method differs from the analyses in the original paper in two important ways. First, the linear model is based upon Procrustes distances, and second the resampling procedure more easily enables inferences within nested models (Collyer et al. 2015) with interaction terms. Specifically, this approach samples (without replacement), the residuals from the ÒsimpleÓ model under comparison, adding these to fitted values, and refitting under the ÒcomplexÓ model. We used the following models to assess the difference in shape dimorphism for each line and wild type background:  Model1: Shape ~ Sex + Genotype Model2: Shape ~ Sex + Genotype + Sex:Genotype  We then performed such analysis for increasing degrees of interactions for the influence of sex, genotype, genetic background and size (for models of shape variables).  SShD was calculated with one of two methods: the advanced.procD.lm() function in the geomorph package (v.2.1.8) in R (v. 3.2.2) and standard Euclidean distances among treatment groups using the lm() function; both approaches yielded equivalent results. To !-)!evaluate the mean shape difference caused by sex, we used linear models based upon Procrustes distance (with RRPP) to compare models where sex is and is not a predictor of shape using the procD.lm and advanced.procD.lm functions in geomorph. These analyses were randomized (by individual) and repeated 1000 times per treatment group to assess whether the magnitude of effect was greater than expected by chance.   Despite having separate and independent ÒcontrolÓ (wild type) lineages for each cross (to control for any potential vial effects or residual segregating variation), we utilized a sequential Bonferroni correction to maintain our Òexperiment-wideÓ nominal alpha of 0.05. Given the large number of comparisons being made, it is likely that this will yield extremely conservative results, and we expect this underestimates the number of mutations that influence sexual dimorphism or mutational effects of allometry of shape on size.   Vector Correlations While the above linear model assesses the magnitude of the effects, for shape it is also important to examine the direction of effects. Specifically, whether the mutations influenced the direction of SShD. To examine this, the vector of SShD was calculated within each genotypic group (wild type VS. mutant). We then estimated the vector correlation between the vectors of SShD for the wild type and mutant as follows: !!"!!!!!!!!"!!!!!!"!!!!!!!!"!!!!!!!!!!!!"!! where the SShD for each genotype is equal to difference between the female and male vectors within each genotype. We used the absolute value of the numerator to avoid !-*!arbitrary sign changes. The denominator consists of the product of the length (norm) of each vector. As with a Pearson correlation coefficient, a value of 0 corresponds to no correlation, while a value of 1 means that each vector is pointing in the same direction (even if they differ in magnitude).  Approximate 95% confidence intervals were generated using a non-parametric bootstrap of the data for each line (The alpha used for the 95% CIs were not adjusted for the number of mutant alleles tested).  Statistical Analysis All statistical analyses were conducted using R statistical software (version 3.2.2). Significance testing (specifically those involving RRPP) was conducted using functions within the geomorph package (v. 2.1.8) and with custom functions. All error bars are 95% Confidence intervals generated by non-parametric bootstraps. All scripts including custom functions are available on github (https://github.com/DworkinLab/TestaDworkin2016DGE).   Results  Different wild type strains vary for Sexual Size Dimorphism (SSD) and Sexual Shape Dimorphism (SShD) in wing morphology, for both direction and magnitude.  As each mutation was repeatedly backcrossed into two distinct wild type strainsÑOregon-R (Ore) and Samarkand (Sam)Ñwe first examined patterns of sexual size and shape dimorphism between these two strains.  !-+! We observed considerable, and highly significant, differences in both SSD and magnitude of SShD between the wild type strains (Figure 4.1A). Further, with respect to the vector of SShD, both wild-type backgrounds were somewhat divergent (Figure 4.1B). The computed vector correlation for SShD between both backgrounds falls within the same range as those calculated for SShD by genotype (0.937, 95% CI 0.92, -0.95), suggesting only subtle changes in direction. Additionally, the allometric relationship between shape and size differs between the two wild type backgrounds. While Ore has a stronger overall slope than Sam, the magnitude of both males and females slopes are reversed by background; for females, shape has stronger association with size relative to males in Ore (F 0.113, 95% CI 0.105, 0.122; M 0.099, 95% CI 0.091, 0.107), whereas the opposite is true for Sam (F 0.105, 95% CI 0.097, 0.113; M 0.120, 95% CI 0.112, 0.129). These differences in size, shape and allometry are all significant based on the randomized resampling permutation procedure (see methods).  Despite tight control of experimental variables (food, temperature) we observed a surprising amount of residual environmental variation for SSD and SShD among each replicate of the two wild type lineages. In the design of the experiment, where for each mutation, within each background, wild-type controls were generated from the cross that shared the environment (vials) with their otherwise co-isogenic mutant sibling. As all of these offspring across the vials are genetically co-isogenic, and only differ in the subtle aspects of rearing environment across vials, this allows us to assess some aspects of how environmental variation influences SSD and SShD. As shown in Figure 4.1A, in   !-,! Figure 4.1. Natural variation in SSD and SShD for two wild type strains. A) SSD and SShD in both wild type background strains are represented three ways: scatterplot (center), SSD histogram with density curve (x-axis, top) and SShD histogram with density curve (y-axis, right). Data points represent mean value for wild-type siblings of each heterozygous mutant cross from a given vial. The Samarkand wild type background has a wider range of SSD, encompassing the low end of the spectrum, whereas Oregon-R tends to be more consistently large in SSD. B) Average direction of SShD in a typical Samarkand (left) and Oregon-R (right) wild type wing. Landmark coordinates are mapped onto a typical wing to demonstrate shape. Arrows represent the vector of shape change (magnified 5x) from female to male wing shapes.   !"#"!--!addition to differences between the two wild type strains for SSD and SShD, there is also variation around the mean estimates for each. Since each data point in Figure 4.1A corresponds to each mutantÕs wild-type siblings from a given cross, these points largely reflect variation among ÒvialÓ effects. Indeed, models based on Procrustes distance suggest that there are significant vial effects (P = 0.009) and vial by sex (P = 0.001) even within the background control populations, which are largely attributable to micro-environmental variation. This is somewhat surprising as external sources of variation such as food (all from a common batch) and rearing temperature (all vials reared in a common incubator, with daily rotation of vials to minimize edge effects) were highly controlled in the experiment. This suggests that the magnitude of SSD and SShD for wing form is influenced by subtle environmental changes, suggesting that high levels of replication to control for these factors is generally necessary.  Despite many mutations having substantial effects on overall shape, a relatively small number influence SSD and SShD.  As demonstrated in the original study (Dworkin & Gibson 2006) and confirmed here, the vast majority of mutations have a significant influence on shape when measured in the heterozygous state (supplementary table 1). Of the subset of 42 mutations used in the current study (from the original 50), all but 10 had a significant effect for genotype (most surviving even a conservative Bonferroni correction) using the Residual permutation (Collyer et al. 2015). Of those 10, most had significant genotype-by-background effects, consistent with the earlier study (despite a different underlying inferential approach). !%..!Despite this, only 18 of the mutations showed evidence for ÒsignificantÓ sex-limited genotypic effects (based on the sex-by-genotype effects), of which 2 survived sequential Bonferonni correction. Additionally, another 12 show evidence for significant effects of sex-by-genotype in combination with other factors in the model (size and/or background). Only one of these 12 survived correction for multiple comparisons. While inferences based on significance alone is quite limited (see below), these results suggest that only a small subset of mutations appear to have sex-limited influences on shape (Table 4.1).   To understand these results more fully, we next focused on the magnitudes of SShD and the SSD index, using non-parametric bootstraps to generate confidence intervals on our estimates.  We performed the analyses separately for each wild type genetic background given that they can differ for both magnitude and direction of SShD. As shown in Figure 4.2, while several mutants show significant effects for either SSD, SShD or both in one or both of the backgrounds, the magnitudes of these effects are small, especially considering the relatively large amount of environmental variation in SSD and SShD observed within strains (Figure 4.1A). Interestingly, while only a few mutations showed evidence for an overall effect on size, these tend to have sex-limited effects (Figure 4.2).   In addition to examining the magnitude of effects, we also examined the direction of effects, and whether the mutations substantially changed the direction of SShD relative to their co-isogenic wild type. As shown in Figure 4.3, the mutations examined in this  !%.%!Table 4.1. Summary table of significant effects by background among mutants. All significant values are taken from Figures 1-4. In the case of vector correlations, 80% was chosen arbitrarily to represent only a small subset of mutants of large effect.    Mutant Allele Pathway # SSD # SShD # Vector Correlation  # Allometry aos W11 Egfr     omb md653 TGF-" Ore Sam, Ore Sam, Ore  cv-2 225-3 TGF-"     GAP1 mip-w[+] Egfr     ksr J5E2 Egfr    Sam dad J1E4 TGF-" Ore Ore   drk k02401 Egfr  Ore   bs/DSRF k07909 Egfr  Ore  Sam s k09530 Egfr    Sam spi s3547 Egfr     mad k00237 TGF-"    Ore ed k01102 Egfr    Sam tsh A3-2-66 TGF-" Sam    cos k16101 Hh     tkv k19713 TGF-"     babo k16912 TGF-"/Hh     trl S2325 TGF-"     rho-AP BG00314 ? Sam  Sam  pka-C1 BG02142 Hh     sbb BG01610 TGF-"     psq kg00811 Egfr Ore   Ore osa kg03117 Chromatin Remodeling   Sam  rasGAP kg02382 Egfr     pnt kg04968 Egfr Ore Ore Ore  !%.&!Table 4.1. (contÕd)   drk k02401 Egfr     cbl kg03080 Egfr     mam kg02641 N/Egfr     rho-6 kg05638 Egfr  Ore Sam, Ore  dpp kg04600 TGF-"   Ore Sam pka-C3 kg00222 Hh   Sam Sam p38b kg01337 TGF-"/Egfr     tkv kg01923 TGF-"     wmd kg07581 Unknown     mad kg00581 TGF-"     ast kg07563 Egfr     dpp kg08191 TGF-"     rho1 kg01774 Egfr?    Sam sax kg07525 TGF-"     sax* sax4 TGF-" Sam Sam Sam  egfr k05115 Egfr  Sam, Ore Ore Sam src42A kg02515 Egfr    Ore rho/stet kg07115 Egfr     !%.'! Figure 4.2. Magnitude of SSD and SShD for 42 mutants in Oregon-R (left) and Samarkand (right) wild type backgrounds. The effect of each mutant is mapped out in a size-and-shape dimorphism space. Genotypic means for each mutant are indicated by point style and connected by a solid line. SSD is plotted on each x-axis for all plots and SShD is displayed on the y-axis. The plots above display the entire range of variation observed, while those below display only the area with the highest density of points. Lines with significant sex-by-genotype effects are highlighted as follows: effect on both size and shape, shape only and size only. Only significant genes (after sequential Bonferroni correction) from the linear models are colored. Few mutations in this study alter sexual dimorphism of size or shape. In addition, the effect of mutations also appears to be highly background dependent, as only two lines, Omb and Egfr, were consistent in both backgrounds. Error bars are 95% confidence intervals (unadjusted alpha). All gene names are displayed lower-case, regardless of dominance.   !%.(! Figure 4.3. Vector correlations to assess similarity of direction for sexual shape dimorphism (mutant VS. wild type) by background. While genetic background appears to have little effect on the direction of SShD for most mutations, several stand out with more divergent directions of SShD. Those mutations with large background effects are also notable for their large effect on size and/or shape. Error bars are 95% confidence intervals (unadjusted alpha).    !%.)!study generally do not substantially influence the direction of SShD, with several notable exceptions such as the mutation in the Omb gene, as well as more subtle effects from mutations such as sax, pnt, drk (among others). Even when the bootstrap confidence intervals do not approach 1, the estimated vector correlation are still generally greater than ~0.9, suggesting only modest changes in the direction of SShD.  Mutations do not substantially alter directions of SShD, nor patterns of allometry. One important aspect of assessing variation in shape, and in particular in situations where there is either (or both) SSD or SShD, is to account for the allometric effects of size on shape when computing the magnitude and direction of SShD. One important approach is to assume a common allometric relationship between size and shape across the sexes (after adjusting for mean differences in size and shape), and regressing out the effects of size. Then using either the residuals or predicted values of shape (after accounting for size) to compute an Òallometry correctedÓ measure of SShD (Gidaszewski et al. 2009). To utilize such an approach requires that the assumption of a common allometric relationship be valid, as has been observed across Drosophila species for the wing shape and size relationship (Gidaszewski et al. 2009).   Prior to computing the allometry-corrected measure we examined this assumption among the mutations used in this study. Of the 42 independent mutations (with their independent controls), 13 had a significant interaction of sex-by-size on the influence of shape (with three surviving the sequential Bonferroni correction). Another eight of them had a sex-by-size interaction imbedded within a higher-order interaction term. Despite !%.*!this the overall magnitudes of effects and directions of allometric relationships appear to be highly similar, with a few important exceptions (Figure 4.4). Thus it is unclear whether using an allometry-free correction is warranted within the context of this study. It is worth noting that making the assumption of a shared allometric relationship, and computing the allometry-corrected measure of SShD did not substantially alter our findings (Supplementary Figure 4.1; Supplementary Table 4.2).      Discussion  While previously underappreciated, it is clear that mutations in genes in several growth factor pathways can act in a sex-specific manner. Of the 42 mutations analyzed, 12 had a significant sex-by-genotype interaction on size, shape or both (Figure 4.2). Only a few mutant alleles had the ability to affect the sexual dimorphism in allometry, the relationship between shape and size (Figure 4.4). Furthermore, nearly all of the mutants appear to act in a background-dependent manner, affecting shape or size in one genotype, but not the other (Figure 4.2).   Previous research has demonstrated the ability of growth pathways to respond to various perturbations, including: individual mutation (Palsson & Gibson 2004; Gao & Pan 2001; Tatar et al. 2001), genetic background (Chandler et al. 2013; Dworkin & Gibson 2006; Paaby & Rockman 2014) and environment (Ghosh et al. 2013; Shingleton et al. 2009; de Moed et al. 1997). Our results are unique in that they allow us to directly assess the effects of these perturbations on relative growth based on sex for both   !%.+!  Figure 4.4. Variation in the magnitude of association between shape and size allometric coefficients among mutations in the Oregon-R (top) and Samarkand (bottom) wild type backgrounds. The ÒslopeÓ of the allometric relationship for shape on size is displayed by sex and genotype. The magnitude of allometric effects appears to be relatively stable across strains, with few mutants substantially altering the wild type pattern of allometric co-variation. Individual lines whose mutants cause a significant sex-by-size interaction are represented dark in contrast to non-significant (faded) lines. Error bars are 95% confidence intervals (unadjusted alpha) for each individual treatment; significance is assessed based solely the interaction terms from the multivariate linear models.   !%.,!direction and magnitude. Relative differences between male and female growth patterns due to these mutations are ultimately responsible for the generation of SSD and/or SShD.   The importance of multiple independent control lineages As expected, different wild type strains vary in magnitude and direction of effects for SSD and SShD (Figure 4.1). The Oregon-R wild type background displays greater dimorphism in both size and shape compared to Sam. Implicit in our results is the understanding that genetic background itself has a profound effect on the underlying wild-type growth pathways and all of the downstream consequences this can have.   Somewhat more surprising is that both SSD and SShD appear quite environmentally sensitive (despite the genotypic effects being relatively insensitive based on our previous work). While great care was taken to reduce the effects of microclimactic variation, edge effects, nutritional variation and even genotypic variation, our results demonstrate that size and shape dimorphism remain highly variable (Figure 4.1).   There always remains the possibility that environmental variation does not entirely account for the wild-type variation observed. For each backcrossed line, a small amount of genetic information surrounding each p-element insertion site is unavoidable, especially during recombination in final cross with mutants and wild-types. This effect is somewhat unlikely, however, due to the fact that these recombination events are rare and affect only single measured individuals. Regardless, such a large amount of !%.-!variation in trait values within ÒisogenicÓ lines is unexpected. Most studies attribute any such variation within genetically (and environmentally) identical lines to stochastic variation in gene expression (Rea et al. 2005; Kirkwood et al. 2005; Raj & van Oudenaarden 2008). Such claims are, however, outside of the scope of our current study.   Rare sex-limited effects on wing form among mutations in EGFR and TGF-! signaling While much is known about the development of wing size and shape (Shingleton et al. 2005; Garc™a-Bellido et al. 1994; Weinkove et al. 1999; Day & Lawrence 2000; Prober & Edgar 2000), comparatively little is known about the sex-specific effects of the genes involved (Horabin 2005; Abbott et al. 2010; Gidaszewski et al. 2009). While these mutations represent only a subset of the almost innumerable potential mutations within and among genes, they serve as a lens through which we can view the sex-limited effects of mutations. It is now clear that only a handful of genes associated with growth may be acting in a sex-dependent manner. Indeed, these results call for a further investigation of the formerly understudied sex-effects of growth pathways.   One such study confirms a link between many of the patterning mutations used in the current study and the development of SSD in the wing (Horabin 2005). In her 2005 paper, Horabin demonstrated that components of the sex-determination pathway (specifically, Sxl) were responsible for activating size-regulating genes within the Hedgehog signaling pathway. In fact, of the handful of genes to display sex-limited effects on SSD or SShD, a few were associated with this pathway, including: Omb, dad !%%.!and Dpp (Horabin 2005; Abu-Shaar & Mann 1998). This does not appear to be coincidence as these are the only mutants in this pathway that we utilized for this study. Since these mutants only represent a subset of those with sex-limiting effects, we cannot assign causality to this pathway. Instead, this demonstrates that sex-limiting effects of genes interact with more complexity than previously understood; no one pathway appears to be acting in a sex-dependent manner to generate shape/size.   Another candidate pathway involved in the generation of SSD is the Insulin and Insulin-like growth factor (IIS)/Target of Rapamycin (TOR) pathway. Evidence suggests that components of this pathway, such as InR (Testa et al. 2013; Shingleton et al. 2005) and foxo (Carreira et al. 2011) can contribute to SSD and/or SShD.   Further studies, such as Takahashi & Blanckenhorn (2015) have found that most mutations appear to decrease the SSD of wing form. Our data appear to yield an interesting trend for the direction of SSD based on genetic background. Ostensibly, growth-pathway mutants in the Ore wild type background tend to decrease SSD, whereas mutants that affect SSD in Sam tend to increase it. At this point it is impossible to say if this trend is biologically meaningful, but given that Ore has a greater underlying magnitude of SSD (and is already in conflict with RenschÕs rule), these mutations may be interfering with genetic mechanisms influencing sexual dimorphism in the Ore background.   !%%%!Our data is somewhat inconsistent with the findings of another previous mutation screen study, namely those of Carreira et al. (2011), wherein the authors found a much greater proportion of random insertion mutations appeared to have sex-specific effects on wing shape. The reasons for this are as of yet unclear, but may reflect methodology, magnitude of mutational effects used or that in the current study all mutations were limited to two signaling pathways. First, our methods allowed us to effectively tease apart the sex-limited interactions of sex for each genotype pair by plotting them in a size-shape space. Second, the authors used a different wild type genetic background than either that were used in this study (Canton-S). It is clear from this study and others (Dworkin & Gibson 2006; Chandler et al. 2013) that genetic background has an appreciable effect on gene function. At least part of the variation in the number of genes affecting wing SSD must necessarily be due to genetic background effects; however, genetic background effects cannot wholly account for the differences observed. Third, we cannot rule out the effects of dominance when discussing the effects of gene function. The genotype of flies in the study by Carreira et al. (2011) was homozygous for all mutants used. Their lines were chosen specifically for their non-lethal homozygous phenotype, whereas mutation used in our study were chosen irrespective of lethality. Because of this, our flies necessarily had to be heterozygous in order to avoid lethality associate with the homozygous phenotype. Perhaps not all loss-of-function mutants within our study were sufficient to alter the phenotype in a sex-limited manner. Finally, because our mutants were deliberately selected based on their association with wing shape morphogenesis, our results are not strictly comparable to those of Carreira et al. (2011).  !%%&!Disentangling mutant-phenotype relationships Our findings suggest that in most cases when a mutational analysis is performed to understand the genetic architecture of SSD or SShD, it is important to assess whether the mutation is only affecting SSD/SShD or whether it is instead demonstrating some degree of sex-biased influence. Many genes may therefore appear to alter SSD/SShD, but are instead only affected by sex as one of several variables of its expression. This may seem like an arbitrary distinction, but it is important if we are to fully understand the genetic underpinnings of complex phenotypes. Many mutants, such as those in the EGFR signaling pathway used here, are either lethal or at least partially ablate development of certain organs as homozygotes, indicating that these genes are necessary for the development of the organ itself. If heterozygotes have sex-specific effects on size or shape, we cannot necessarily conclude that this gene affects SSD or SShD, but rather that the gene is important for formation of an organ and has sex-dependent effects. Only in the case of genes such as Maf1, a gene that has been demonstrated to directly effect SSD in Drosophila (Rideout et al. 2012), can we conclude that said gene is affecting SSD and not simply acting in a sex-limited manner.   To fully understand the scope of SSD and SShD, one must precisely define what is meant by size and shape. While the definition of size is relatively straightforward to interpret, shape is somewhat more nuanced. For many organs, shape can essentially be broken down into the relative size of component parts of the larger structure (given that all aspects are homologous). For instance, during development in Drosophila there are multiple quadrants of the developing wing imaginal disc whose individual sections !%%'!may grow more or less in relation to the others, thus altering the ÒshapeÓ of the wing. Mutant phenotypes may manifest as changes to large sections, such as a widening of the entire central portion of the wing (Ptc) or they may be subtler in effect, altering the placement of only a single crossvein (cv-2) (Dworkin & Gibson 2006). While these mutants may have local effects on size, such that they alter shape, what is less clear is whether these mutants are affecting size in a localized manner or the actual shape itself.   The effects of each pathway appear relatively consistent despite differences in genetic background. While mutations within the Egfr pathway tended to affect primarily SShD, those in the TGF-" pathway had a more mixed effect (more frequently affecting SSD). This pattern suggests that genetic background may only alter a mutationÕs quantitative effect, rather than its qualitative effect.   Ultimately, our results demonstrate the importance of distinguishing between the relative contributions of each mutation to sexual dimorphism for shape, size or both. Of those mutants with sex-limited effects, even fewer exclusively affect either shape or size dimorphism (Figure 4.2). While some studies have been successful in artificially altering SSD of specific traits through selection (Bird & Schaffer 1972; Douglas J. Emlen et al. 2005; Reeve & Fairbairn 1996), it is unclear whether whole trait size or simply trait shape (e.g. length) has been altered. Our results demonstrate the need to exercise caution when discussing the effect of mutants on size or shape dimorphism.   !%%(!Reassessing the assumption of common allometry One important method for quantifying ÒshapeÓ changes involves examining allometric relationships, specifically static allometry, which is the relationship among adult individuals between body size and organ size (Huxley 1932; Stern & Emlen 1999). In fact, one of the most obvious ways that males and females can differ is through differences in scaling relationships between body parts; these encompass some of the most obvious sources of variation in the natural world (Bonduriansky & Day 2003; Shingleton et al. 2009). By studying the relationship between two traits (e.g. body vs organ size), we can glean important information about the relative growth of traits and, therefore, the underlying mechanisms of differences in the growth of these traits. Consequently, allometry is an important tool for biologists to assess differences in size and shape dimorphism within (and across) a species. Our results support the claim for the importance of studying allometry by demonstrating that, while some mutants may have sex-limited effects on shape and/or size dimorphism (Figure 4.2), they do not necessarily affect the relationship between trait shape and size (Figure 4.4). Many mutants cause significant differences in sexual dimorphism of allometry, but do not necessarily alter SSD or SShD. These results may seem counterintuitive, but it is important to remember that, while changes in SSD or SShD may shift the direction of slope of allometry along one or more dimensions (in shape space), this does not necessarily alter the allometric slope itself (Frankino et al. 2005).   Since D'Arcy Thompson (1917) outlined his approach of how relative changes in body and organ size can be mapped out onto Cartesian coordinates to visualize relative !%%)!growth, the study of allometry and shape have been closely linked. Modern approaches use similar, albeit much more complicated methods to assess changes in relative landmark positions (Sanger et al. 2013; van der Linde & Houle 2009; Abbott et al. 2010). Ostensibly, one of the downfalls of shape analysis is that shape inherently carries information about its underlying relationship to size, despite the fact that geometric morphometric analyses partially separates it from shape (Gidaszewski et al. 2009; Mosimann 1970; Gould 1966; Nevill et al. 1995). More specifically, size itself is a measurement based on some aspect of shape. If size and shape do not scale isometrically (such that unit increase in size is accompanied by an equal increase in shape), then the underlying co-variation will be reflected in estimates of shape that are disproportionately affected by size (Mosimann 1970). This issue is implicit in the geometry of shapes themselves; as absolute size increases, surface area to volume ratios decrease (Gould 1966). This is particularly bad news for studies wishing to analyze induced changes in shape and size, because it means that the degree of independence between these two variables may be difficult to infer. However, by plotting size on shape and using the residuals from this model, Gidaszewski et al. (2009) were able to effectively eliminate the issue of non-independence with size and shape. These residuals represent the total variation in shape that is not due to allometric effects of size.   Allometric patterns of variation across sex and genotype are necessarily more complicated. While it is known that shape (and shape dimorphism) is strongly influenced by its relationship with size, it is not always clear that the assumption of a common !%%*!allometric relationship across sexes is met. Previous studies examining patterns of SSD and SShD (Gidaszewski et al. 2009) generally made the assumption of a common allometric relationship between males and females within each Drosophila species. This was despite their analysis suggesting that this assumption may not hold for all species. For the data we examined here, we could reject this assumption based on inferences based on statistical significance. Yet, it is clear that the magnitude of such differences were small, and allometric relationships were similar in most cases. Indeed, the allometric influence of size on shape appears to be largely consistent with respect to direction of effects, with a few notable exceptions (Figure 4.4). Regardless, we erred on the side of caution with this matter and decided to eschew analysis of SSD and SShD under assumption of common allometry. It is worth noting that the assumption of common allometry did not substantially alter the observed results (Supplementary Figure 4.1). As with other studies, we suggest that a rejection of this assumption simply based upon significance may not be optimal, and future work should determine what the consequences of making such assumptions might be for studies of sexual dimorphism and allometry.   Our results clearly demonstrate the effects of growth pathway mutants on SSD and SShD. Most notably, we cannot rule out the sex-specific effects of any genes involved in growth. Our results demonstrate the current lack of understanding of how growth-related genes interact with the sex of the individual. By visualizing the effects of each mutation within the framework of size/shape space we gain a previously unrealized understanding of the role each mutant plays in generating a sexÕs phenotype. While this !%%+!method is especially powerful for studying sexual dimorphism, its applications are not restricted to it. We therefore present this method as a means for dissecting the contributions of mutants to the development of size and shape.     !%%,! CHAPTER 5:  The Contribution of Candidate Genes to sexual size and shape dimorphism    !%%-!Introduction  In many species males and females can differ from one another in virtually all measurable traits including morphology, behavior, physiology, and life history. These differences are sometimes so distinct that individuals can be unambiguously assigned with minimal effort. Such qualitative distinctions between the sexes stem from the fact that males and females usually differ genetically (Dawkins & Krebs 1979; Mank 2009) and also contribute unequally to their gametes (Trivers 1972; Bateman 1948; De Lisle & Rowe 2015; Fairbairn 2005; Mank 2009). As each sex becomes adapted to its role as a large (female) or small (male) gamete contributor, specializations may arise that lead to the separation and definition of each sex.   Because both sexes still share the majority of a genome, most phenotypic differences between them must arise largely from sex-specific gene regulation rather than sex-specific genes. While in some (but not all) species, males and females differ genetically by their sex chromosomes, but these make up a relatively small portion of their respective genomes (Mank 2009). Despite a largely shared genome, this is not generally considered an impediment to prevent sexual dimorphism from evolving when there are sex-specific fitness optima. This is clear from the many examples of sexual dimorphism are exceedingly abundant in nature (Bininda-Edmonds & Gittleman 2000; Norman et al. 2002; Pietsch 2005; Badyaev 2002; Blanckenhorn et al. 2007; Huey et al. 2006). If males and females share most of a genome and sexual dimorphism is so !%&.!abundant, then phenotypic variation must therefore be due to how genes are being expressed, rather than which genes are being expressed.   Perhaps one of the most fundamental (in terms of its performance) phenotypic ÒtraitsÓ of an organism is its size. Like many other traits, size can be regulated to produce sex-specific means. Indeed, SSD has been the subject of possibly hundreds or thousands of studies, as well as many reviews (Lovich & Gibbons 1992; Allen et al. 2011; Fairbairn 1997b; Chown & Gaston 2010; Blanckenhorn 2005; Smith 1999; Badyaev 2002; Stillwell et al. 2010) and a dedicated edited volume (Fairbairn & Blanckenhorn 2007). However, most such studies are either descriptive with respect to SSD, or are studying them in the context of evolutionary, behavioral, or physiological questions. Few of these studies have investigated the proximate mechanisms that govern SSD (but see Cox et al. 2009; Tammaru et al. 2009; Zhao & Liu 2014; Esperk et al. 2007; Stillwell & Davidowitz 2010b). On the other hand, a substantial body of research has been dedicated to the study of size control, especially with respect to developmental genetics (Carreira et al. 2011; Oldham et al. 2000; Gokhale & Shingleton 2015; Mirth & Riddiford 2007; Nijhout 2003; Edgar 2006). In Chapter 3, I discussed in detail a number of growth-related candidate pathways and their effect on growth. We now know that both nutrients and nutrient-sensitive growth pathways are necessary for the proper development of sexual size dimorphism (SSD).   However, one must consider what is really meant by ÒsizeÓ. After all, an organism is more than a sum of its parts. Importantly we can consider the relative sizes of different !%&%!organs in an individual, as these do not necessarily scale isometrically with one another, and may differ among the sexes (Bonduriansky 2007a; Kodric-Brown et al. 2006; Bonduriansky & Day 2003; Gould 1966). Thus the shape of an individual or shape of organs within the individual may be just as important to consider as overall size. While there have been many ways of defining shape in the biological literature (see chapters 1, 4 for some overview), a commonly accepted approach is to define shape as the residual variation of homologous structures once the effects of location, scale, and rotation have been removed. In this definition, shape itself is a single geometric representation of a multi-dimensional trait, allowing for relatively intuitive descriptors across biological groups. Sex-specific variation in shape is important, because it allows for the development of more complex phenotypes than size. Importantly, there is substantial evidence of SShD (Dworkin & Gibson 2006; Gidaszewski et al. 2009; Sanger et al. 2013; Pitchers et al. 2013; Cheng & Kuntner 2015; Shine 1989; Valenzuela et al. 2004; Collyer et al. 2015) including species that do not necessarily differ in SSD (Schwarzkopf 2005). In Chapter 4, I demonstrated the extent to which certain wild type and mutant strains of Drosophila melanogaster wings can vary by size and shape. Indeed, in fruit flies alone, we know that proper wing shape is essential for predator avoidance (DeNieu et al. 2014), adaptation to various environmental conditions (Pitchers et al. 2013; Trotta et al. 2010), and courtship (Menezes et al. 2013; Hoikkala et al. 1998; Abbott et al. 2010). Because of its strong influence on fitness and the fact that shape and size covary considerably within and between species (Mitteroecker et al. 2004; Klingenberg 2016; Testa & Dworkin 2016; Gidaszewski et al. 2009), a more !%&&!complete understanding of the relationship between SSD and SShD is necessary, in particular with regards to whether they share common mechanisms that regulate them.   In Chapter 3, I discussed in detail a number of growth-related candidate pathways and their effect on overall growth and SSD. We now have some idea of how these candidate genes function to regulate size (Horabin 2005; Rideout et al. 2015; Mirth & Riddiford 2007; Shingleton et al. 2005; Stocker et al. 2003; Caldwell et al. 2005) and some insights into whether they influence sexual size dimorphism (chapter 4, Bıhni et al., 1999; Rideout et al., 2015). Chapter 4 demonstrated how some genes in growth-related signaling pathways (EGFR and TGF-§) might vary with respect to influencing size and shape dimorphism. Chapter 3 suggests that TOR signaling may potentially be both necessary for and sufficient for whole-body SSD, but what we donÕt know is the relative contribution of these pathways to organ (wing) size and shape. While we understand a lot about these sex-specific interactions, we still cannot point to any specific gene as being both necessary nor sufficient for SShD, despite a small set of studies examining this (Testa & Dworkin 2016; Abbott et al. 2010; Carreira et al. 2011).   Sex-biased gene expression itself is largely attributed to the sex-specific splicing of common genes. For instance, the core of the sex determination pathway can be summarized by the sex-specific splicing of three major genes: sex-lethal (Sxl), transformer (tra), and double-sex (dsx) (Fear et al. 2015). As a result of the sex-specific sex:autosome ratio, an active (female) or inactive (male) Sxl protein will ultimately be produced. Since male Sxl protein is non-functional, it does not properly splice tra, !%&'!resulting in a male tra splice variant that is also inactive. Flies that are mutant at the tra locus have been demonstrated to develop as normal males (Sturtevant 1945). It is precisely this inactivity though, that allows for male-specific splicing of the active male isoform of dsx, which then helps to regulate the development of a suite of male-specific phenotypes including pigmentation (Williams et al. 2008), sex combs (Tanaka et al. 2011), and the anal lobe (Glassford et al. 2015). In this way, genes that exist in both sexes can be activated or deactivated spatially and temporally in order to produce sex-specific phenotypes.   While we have a fairly complex understanding of the effects of downstream components of the sex determination pathway (Fear et al. 2015), we know very little about the specific effects of those near the top. For instance, we know that genes like transformer (tra) act to determine male vs female form (Sturtevant 1945; McKeown et al. 1988; Verhulst et al. 2010; Belote & Baker 1982) for discrete traits, but what is still unknown is at what level sex-specific size and shape are determined. By shape, we refer not to the primary sexual characteristics, but rather to the secondary. Both sexes have wings, yet we know that males and females have substantial dimorphism for both wing size and shape (Testa & Dworkin 2016; Sonnenschein et al. 2015; Gidaszewski et al. 2009; Abbott et al. 2010). Recent evidence has suggested that SSD is mediated in part by tra (Rideout et al. 2015); however, it is unclear whether tra of other components of the sex determination pathway may also influence SShD.  !%&(!To gain some insight into the relationship of the proximate mechanisms of SSD and SShD, I examined several growth pathways in tandem with the sex determination pathway to identify genes that modulate SSD, SShD, or both. We use the same candidate pathways discussed previously. We know that some pathways (IIS, Tor, and SDP) are necessary for SSD and some are sufficient (Tor), but how these pathways affect SSD and SShD in the wing are as of yet unknown. Specifically, my hypothesis is that SSD and SShD are being regulated by unequal sexual responses to growth pathway genes. If any candidate genes within these pathways are, in fact, involved in some combination of SSD or SShD regulation, then genetic manipulations should reveal their effects. Fortunately, the statistical tools available with geometric morphometrics will allow for the very precise quantification of size and shape. By quantifying SSD and SShD and examining their relationship to one another, we can investigate the contribution of each gene to SSD, SShD, or both. If SSD and SShD are indeed regulated by the same gene, tra, then we would expect these data to reflect that. Other candidate pathways would be expected to reveal similar patterns of SSD vs SShD if they are responsible. Our results suggest a general reluctance for candidate genes to decrease SSD or SShD, which differs considerably from previous studies (Takahashi & Blanckenhorn 2015; Carreira et al. 2009; Testa & Dworkin 2016). Regardless, our candidate pathways have demonstrated that control of SSD and SShD are only partially correlated to one another, and the results suggest that wing and overall body size SSD are potentially not mediated by similar mechanisms.   !%&)!Materials and Methods Fly strains and Maintenance All strains were chosen based on their ability to over-express or knockdown specific predetermined candidate genes. With few exceptions, all flies were obtained from the Bloomington Stock Center. See Chapter 3 for the list of strains used. Driver lines consisted of Actin5c-GeneSwitch-Gal4 (see Chapter 3), A9-Gal4, and Bx-Gal4 (see discussion). Flies were maintained on standard cornmeal-molasses food medium at either 24ûC or 18ûC on a 12/12 hour light/dark cycle.   Each over-expression line was introgressed into the common wild-type background, Samarkand (Sam), for seven generations to avoid background effects. We performed introgressions by backcrossing red-eyed females (carrying the insertion element) to males of the Sam background. Since knockdown lines utilized TRiP insertional elementsÑall of which are already in a common background, Òyellow, vermillionÓ (y,v)Ñwe did not introgress these into Sam. While some of the TRiP insertion lines also are segregating the X-linked scute allele, this was not the case for any of the lines except Sxl and InR.  For the experiments discussed in this chapter, we used wings from the flies in Chapter 3 where gene manipulation occurred organism wide, in addition to a suite of new genetic crosses using the wing-specific Gal4 lines so that the gene expression was only modulated in the wing.   !%&*!Modulating gene expression using the Gal4 sensitivity to temperature The Gal4-UAS system is known to be temperature-sensitive (Duffy 2002), with increasing Gal4 activity at warmer temperatures. As such, I generated all crosses at two temperatures: a high temperature (24ûC) to drive high levels of expression and a low temperature (18.5ûC) to drive lower levels of expression. These temperature treatments were replicated for both A9 and Bx drivers; since gene expression in GeneSwitch-crossed flies was already modulated, they were not included in the low-temperature replicate. Using this system, we were able to drive the overexpression and/or knockdown of genes that would otherwise have ablated the wings at higher temperatures.   It is worth noting that, while GeneSwitch crosses were expected to drive expression in all wing tissues, A9 and Bx are only known to drive expression in the dorsal pouch of the wing imaginal disc. Manipulations in the dorsal pouch will therefore not necessarily affect the ventral portion of the wing disc in an identical manner. In the case where the growth of wing imaginal discs are perturbed, particularly by nutrient-sensitive growth pathway manipulation, growth is coordinated within (and among) other developing organs (Gokhale et al. 2016; Stieper et al. 2008). Wing discs therefore respond to spatially restricted perturbations by coordinating aspects of wing growth. In this way, perturbations solely in the dorsal pouch are expected to affect both wing size and shape. However, it remains that for individual genes (whose effects are entirely cell autonomous) that the effects could be limited to the dorsal portion of the wing. However, such effects should generate a telltale phenotypic effect of ÒbendingÓ or ÒdelaminatingÓ !%&+!of the adult wing (as one layer differs in size from the other), were rarely observed in these crosses (table 5.1). Furthermore, in many RNAi knockdowns, the whole wing (and not just the dorsal layer) was ablated.  Titrating gene expression with GeneSwitch See Chapter 3 for full protocol.   Experimental Fly Crosses All crosses were maintained at 24ûC or 18.5ûC with a 12/12 hour light/dark schedule. For all experimental treatments, we crossed four virgin females carrying a wing driver (both drivers are X-linked) with male flies of the appropriate treatment (i.e. target genes whose expression was to be modulated). We used the following three drivers to driver expression into the wing: A9, Beadex (Bx), and Actin5C-GeneSwitch (GeneSwitch). The first two, A9 and Bx, were wing-specific, only driving expression in the wing imaginal disc. The third driver, GeneSwitch, was a ubiquitous driver (see Chapter 3 for full details).   Parental flies were allowed to oviposit on the food medium at their respective temperature treatment, allowing their progeny to be raised completely in the proper temperature. All vials were monitored for egg density to ensure that no vials were under-crowded (20 eggs) or over-crowded (50 eggs). Once adults had exclosed and wings sclerotized (based on the presence of a fully formed adult wing), flies were collected and placed together into 1.5ml Eppendorf tubes for storage per treatment  !%&,!Table 5.1. List of wings displaying aberrant phenotype and their relative frequencies. driver response temperature (ûC) aberrant phenotype  relative frequency A9 Akt-RNAi 24 blistered  common A9 Akt-RNAi 24 ablation rare A9 Ci 18 ablation full A9 Ci-RNAi 24 vein defects  common A9 Ci-RNAi 24 blistered rare A9 fru 24 crumpled rare A9 InR 18 vein defects uncommon A9 InR 24 vein defects common A9 InR-RNAi 18 vein defects common A9 InR-RNAi 24 vein defects uncommon A9 Ptc-RNAi 24 vein defects common A9 Sxl-RNAi 18 vein defects full (female) A9 Tor[TED] 24 partial-full ablation near full A9 Tor[wt] 24 small wing  common (female) A9 Tor[wt] 24 ablation uncommon (female); full (male) Bx Akt 18 crossveinless uncommon Bx Akt-RNAi 18 ablation full Bx Ci-RNAi 18 vein defects full Bx  Ci-RNAi 18 blistered uncommon Bx Ci-RNAi 24 blistered common !%&-!Table 5.1. (contÕd) Bx dsx-RNAi 18 crossveinless common Bx Foxo-RNAi 18 crossveinless common Bx Foxo-RNAi 24 vein defects common Bx Foxo-RNAi 24 crossveinless uncommon Bx fru-RNAi 18 crossveinless common Bx InR 18 vein defects uncommon Bx InR 18 crossveinless uncommon (female); near full (male) Bx InR-RNAi 18 crossveinless uncommon Bx InR-RNAi 24 vein defects common Bx InR-RNAi 24 crossveinless uncommon Bx ptc 18 crossveinless rare Bx ptc-RNAi 18 vein defects full Bx ptc-RNAi 18 & 24 crossveinless common Bx Ras 18 crossveinless common Bx Ras-RNAi 18 & 24 blistered common (female) Bx Ras-RNAi 18 & 24 partial-full ablation full (male) Bx S6K 18 & 24 crossveinless common Bx Sxl-RNAi 18 vein defects full (female) Bx Tor-RNAi 18 & 24 crossveinless common (female); full (male) Bx Tor[wt] 18 ablation/blistered full (male) Bx tra.F 18 crossveinless common Bx tra-RNAi 18 crossveinless common !%'.!group. Each tube was filled ~1ml of 70% ethanol to preserve all flies. Adult flies were dissected in a phosphate buffered solution and mounted on slides in a solution of 70% glycerol and 30% PBS. Wings were taken from the right side of each fly for consistency and mounted on labeled slides.   Imaging and data acquisition Once all wings were dissected, slides were then imaged using an Olympus BX51 microscope. Resulting wing images were cropped and resized for processing based on the software requirements of tpsDig (version 1.39; Rohlf 2003) and Wings (version 3.72; Houle et al. 2003). Using tpsDig, two reference landmarks were placed on each wing so that Wings could summarize each wingÕs shape with a series of splines. Wing spline data were then read into CPReader (version 1.12r; CPR 2015) and converted to a series of landmarks and sliding semi-landmarks. Since many treatments resulted in wings with aberrant crossvein phenotypes, crossvein landmarks were intentionally omitted from these analyses (Table 5.1). CPReader allowed us to calculate shape coordinates from the raw landmarks by using a Generalized Procrustes Analysis to remove the effects of rotation, while location, and scale are removed beforehand. Centroid size (i.e., the square root of the sum of the squared distances from each landmark to the centroid of the configuration) was then calculated and scaled based on original wing size (accounting for any changes due to cropping and resizing images for data acquisition).  !%'%!Morphometric Analysis While the scaling of the landmark configurations by their centroid size removes any isometric effects, the allometric component of size on shape remains. As has been shown previously (Gidaszewski et al. 2009; Pitchers et al. 2013; Debat et al. 2016) the allometric component of shape in Drosophila wings can be considerable and needs to be taken into account for studies of sexual dimorphism. In order to account for differences in allometric slopes within groups, we removed the effect of allometric (size vs shape) scaling on wing shapes for each. For each treatment group (separated by driver, response, and temperature), we fit a multivariate linear regression of shape on size and used the resulting shape residuals as our new shape coordinates. These values correspond to variation in shape that is ideally uncorrelated with variation in size. (Klingenberg, 2016; and references therein). Unless otherwise differentiated, all references to SShD will be of the allometry-corrected rather than uncorrected values.  Analysis of SSD and SShD Both SSD and SShD were estimated and statistical inferences made using the same approaches as detailed in Chapter 4.  All statistical analyses were conducted using R statistical software (version 3.2.2). Significance testing (specifically those utilizing RPPP) was conducted using functions with the geomorph package (v. 2.1.8) and with custom functions (see chapter 4). All error bars are 95% confidence intervals generated by non-parametric bootstraps. !%'&!Specific models differed slightly from Chapter 4; we used the following models to assess the difference in shape dimorphism for each driver, response, temperature, and dose (where ÒgenotypeÓ corresponds to control vs treatment):  Model 1: shape!sex + genotype  Model 2: shape!sex + genotype + sex:genotype   Results  Of the candidate gene pathways tested, few yielded significant differences in SShD; most either increased or decreased SSD (Figure 5.4). Surprisingly, and inconsistent with the previous literature, none of the crosses resulted in a decrease of SShD. Several genes did modulate SShD. When Sxl (A9 and Bx crosses at 18ûC; 24ûC were unviable), InR (Bx at 18ûC and GeneSwitch at 24ûC crosses), and Ci (A9 and GeneSwitch crosses at 24ûC) are knocked down, SShD increases. In all cases, increased SShD appears to be due to disproportionate response in female shape vs male shape (observation not quantified; Figure 5.1). For genes like Sxl and InR, knockdowns are usually accompanied by a decrease in SSD and increase in SShD. Since Ci knockdowns only consistently affect SShD and not SSD, it is a much better candidate for necessity of SShD.   Obviously, the effect of each treatment on size and shape in males and females was not always equal, suggesting sex-limited effects. Figure 5.3 demonstrates how male and female size (top) and shape (bottom) covary with one another for sex determination   !%''!Figure 5.1. Sexual Shape Dimorphism in knockdowns that caused an increase in SShD. Average direction of SShD in a Sxl knockdown (top left), Ci knockdown (top right), and InR knockdown (bottom left) wings. Landmark coordinates are connected to display typical wing veins and outlines to demonstrate shape. Colors represent both female (red) and male (blue) shapes (magnified 1x) from female to male wing shapes. Knockdowns that increase SShD tend to do so by causing an exaggerated female response to shape vs males.      !0.20.00.10.2!0.150.00xysxl knockdown!0.3!0.10.1!0.150.00xyci knockdown!0.20.00.10.2!0.150.00xyinr knockdown!%'(!pathway treatments. The resulting male/female scaling relationships were not always isometric, thus we corrected SShD values to remove the effects of allometry (see methods for details).  After correcting SShD values for allometric scaling effects, we noticed a negative trend between SSD and SShD (Figure 5.2). Before correction, however, this trend did not exist, suggesting that allometry-free values of SShD may be an artifact of rescaling the size/shape relationship. Since overall size and SSD are uncorrelated, their responses to SShD and allometry-corrected SShD tend to covary similarly. Figure 5.2 allows us to visualize the all such relationships between all treatments in the context of SSD, SShD, Allometry-free SShD, and Size (due to unequal scaling relationships; Figure 5.3).   Of those crosses that had an effect on SSD, most perturbations tended to increase SSD rather than decrease it (Figure 5.4). A total of 20 treatments significantly increased SSD, while only 11 significantly reduced it. We also found a driver-specific bias in those crosses that increased SSD. Seven of the overexpression crosses significantly increased SSD, but only two of these were non-Bx mediated. Results from knockdown crosses were even more dramatic. In fact, nine of the 11 total knockdowns that increased SSD were driven by Bx-Gal4. Because Bx-Gal4 is X-linked and also acts as a hypomorph, these results were not unexpected (see discussion for rationale).   Within the sex determination pathway, only Sxl and tra stand out. Sxl knockdowns produce flies with much smaller SSD and larger SShD, whereas overexpression of the   !%')! Figure 5.2. Pairwise comparision between SSD (first row/column), unmodified SShD (second row/column), allometry corrected SShD (third row/column), and Mean Size for males and females (fourth row/column). In allometry-corrected plots, there tends to be a negative relationship between SShD and both SSD and Mean Size. Uncorrected SShD does not share similar trends, but remains somewhat uncorrelated.    !!"#$%&&'()*+,-.'++).*)/0!!"#!!#1)%2-!34)!!"#$%&&'()*+,-.'++).*)/0!!"#!%'*! Figure 5.3. The allometry of SSD (above) and SShD (below) for feminized (left) and masculinized (right) crosses. Female scores are plotted on the x-axis, while male scores are on the y-axis. Centroid size was used as the size variable for SSD plots. Shape scores generated using methods outlined in Drake and Klingenberg (2008) were used to represent male and female shape. We recognize that there are issues with estimates of shape calculated this way (that they correlate with size); however, this method remains a valuable tool for summarizing overall shape. Red point outlines represent significant outliers from the Standardized Major-Axis Regression (sma), whose line is represented as blue. A dotted gray line in the SSD plots corresponds to the control group sma line. dsx was included in ÒfeminizedÓ crosses arbitrarily due to the   !"#$!%&'$(&)$(&)$*$+&)$*$+&)$!%'+!Figure 5.3. (contÕd)  fact that other sex determination pathway knockdowns were expected to cause masculinization, despite the fact that dsx knockdown was expected to affect both males and females.   !%',!    Figure 5.4. Magnitude of SSD and SShD for knockdown and overexpression crosses. The effect of each manipulation is mapped out in a size-and-shape dimorphism space. Genotypic means for each treatment type are indicated by point !%'-!Figure 5.4. (contÕd)  style (knockdown vs overexpression) and color (temperature). SSD is plotted on each x-axis for both plots and SShD is displayed on the y-axis. The plots above display the entire range of variation observed, while those below display only the area with the highest density of points. Lines with significant sex-by-genotype effects are highlighted as follows: effect on both size and shape, shape only and size only. Only significant genes (after sequential Bonferroni correction) from the linear models are colored. Few mutations in this study reduce sexual dimorphism of size or shape. Error bars are 95% confidence intervals (unadjusted alpha). All gene names are displayed lower-case, regardless of dominance.   !%(.!female isoform tra.F causes reduction of SSD, but no effect on SShD. Data for tra knockdowns was less consistent for wing-specific expression (A9-Gal4 over-expression caused reduction in SSD, but Bx-Gal4 knockdown caused increase in SSD. This is interpreted as a Bx-specific effect, since GeneSwitch-Gal4 over-expression was consistent with A9-Gal4).   These data are further supported by vector correlations between control and treatment wings. Vector correlations for SShD in Sxl were consistently below 60%, and those for tra even lower (Figure 5.5). Others treatments with consistently low vector correlations (and SSD vs SShD scores) include the following: Ci, InR, ptc. Occasionally other treatmentsÑsuch as akt, Foxo, Tor, S6K, and fruÑyielded low vector correlations in overexpression lines.   Discussion  Our data suggest low sensitivity for genetic perturbations to decrease SSD or SShD, counter to previously observed effects. Of the crosses that showed significant effects, somewhat fewer caused a reduction in SSD and none caused reduction of SShD. These data conflict with results from Testa and Dworkin (2016; V P Carreira et al., 2009; Takahashi & Blanckenhorn, 2015), wherein genetic perturbations appear to show no preference for increasing or decreasing SSD.     !%(%! Figure 5.5. Vector correlations to assess similarity of direction for sexual shape dimorphism (treatment vs. control). While temperature appears to have little effect !"#$%&'(%&&")*$+%,-!"#$%&'()*(*+,-./01#$,!'2342(56*,6*,6786*,9##,!#,!)&'()&'(:!%(&!Figure 5.5. (contÕd)  on the direction of SShD for most mutations, several stand out with more divergent directions of SShD. Those manipulations with significant effects on SShD in previous figures are also notable for their large effect on vector correlation. Lower values correspond to lower shape correlation, thus a larger overall effect. Error bars are 95% confidence intervals (unadjusted alpha).   !%('!For these experiments, we noticed a significant negative trend between SSD and SShD. Standardized Major Axis regressions reveal that as SSD increases, SShD decreases (Correlation = 0.20; p < 0.0001). While this trend holds true for SShD residuals (allometry-corrected values of shape), it is absent from other pairwise plots of shape on size (Figure 5.2). In this case, it may be a statistical artifact due to how the allometry correction may generate a spurious correlation for the effect of shape on size. Regardless, this trend is very curious from a biological standpoint. What this trend reveals is that perturbations of no genes increase or decrease both SSD and SShD at the same time. If this trend holds true on a species-level, it could mean that control of SSD and SShD is mutually exclusive.   The case for removing the assumption of common allometry Not surprisingly, variation in shape is influenced by variation in size, both isometrically and allometrically. Simply put, shape often covaries with size. Some shape configurations may simply be associated with larger or smaller sizes, despite scaling by centroid size removing the isometric effect of size (Mitteroecker et al. 2004).  In the case where shape can covary with more than one independent variable (e.g. when shape differs by sex and genotype), there may be more than one allometric slope along which shape can vary. For instance, if females had greater variation in size than males, then assuming a common allometric (shape vs size) slope would therefore not be valid. However as pointed out in Klingenberg (2016) a simple statement of sex:size interaction (and its significance) is insufficient, in particular with very large sample sizes. In such cases a very small (and biologically insignificant) alteration in the allometric relationship !%((!may be statistically significant with large sample sizes. Thus it is important to examine the vector correlations for these allometric vectors.  With the data I have generated here, males and females do not covary isometrically across all of the experimental treatments (Figure 5.3). If the covariance between males and females for each group is different, then we cannot assume a common allometric slope for all groups (i.e. shape may not react similarly to changes in shape for every group). Given the wide variation in size of our data, we can no longer assume a common allometric slope for all data and must therefore correct for it. Fortunately, there are methods of reducing the allometric effect of size on shape between each treatment group (Gidaszewski et al. 2009).   In order to account for differences in allometric slopes within groups, we therefore removed the effect of allometric (size vs shape) scaling on wing shapes for each. The process of removing the effects of allometry involves fitting a multivariate linear regression of shape on size. The resulting line will display a ÒpredictedÓ line of best fit, which represents the allometric component of shape on size. Movement along this line describes the change in shape as size varies. In order to remove that effect, we must use the shape (y-axis) residuals of this model. Ideally, residuals are supposed to be uncorrelated with the line of best fit, Òi.e. allometry.Ó This must be done individually for each treatment group, so as to ensure we are not assuming any common allometries. The resulting values will correspond to the shape variation that is independent of size once the effect of shape on size has been accounted for, yielding the best possible   !%()!approximation to shape uncorrelated with size. (Klingenberg, 2016; and references therein)  Reasons for cautious interpretation Because the insertion of driver and response p-elements into the genome is known to have phenotypic effects, we must cautiously interpret these data. In the case of Bx-Gal4, for instance, many more crosses yielded significant size effects than would be expected due to chance alone. This is likely due to the fact that Bx-Gal4 is an enhancer trap, meaning that the Bx-Gal4 driver itself is inserted into the native Bx locus. Since Bx-GAL4 is X-linked and acts as a recessive hypomorph (independent of its role as a GAL4) it was not unexpected to have different effects in males (who are hemizygous for the Bx-GAL4) and females (who are heterozygous for it). While much care went into generating proper crosses to control for the effects of p-element insertion, more controls would have been ideal. For example, we did not originally control for driver-specific effects of these crosses. We are currently generating the remaining necessary controls (and replication of key experimental treatments).  As mentioned earlier, A9-Gal4 and Bx-Gal4 are both wing-specific drivers, whose expression is concentrated more strongly in the dorsal than in the ventral pouch of the developing wing imaginal disc (Liang et al. 2014; Sun & Artavanis-Tsakonas 1997; Peterson & OÕConnor 2013; Bejarano et al. 2012). A characteristic over or undergrowth of only the dorsal layer, rather than the whole wing, might be expected in crosses aimed to affect size. While such a warped or blistered phenotype was observed in a few !%(*!crosses, most crosses appeared to yield the intended phenotype for its respective cross (table 5.1). Whereas some studies using A9-Gal4 and Bx-Gal4 have found that occasional curling and Òcanoe-shapedÓ wings can occur (Liang et al. 2014), many still find that these drivers adequately influence the full adult wing (Nguyen et al. 1998; Sun & Artavanis-Tsakonas 1997; Jackson et al. 1997; Takeo et al. 2005; Bejarano et al. 2012). For A9 at least, the ability to produce full wing phenotypes is because expression is ubiquitous in the wing disc until the late third instar, where it is then restricted to the dorsal compartment (Haerry et al. 1998; Nguyen et al. 1998). Regardless, we now have reason to believe that this issue may affect our interpretation of the data. The effects we observe here are potentially those that we set out to measure; conversely, they could be reflecting a sex-specific sensitivity to these genetic manipulations instead.   For the purposes of this paper, we will focus on SSD and SShD as the sex-specific response to dorsal wing disc manipulation, rather than assuming a whole-organ response to genetic manipulation. This interpretation may explain why a reduction in SSD or SShD was so uncommon. A given genetic manipulation could be acting to reduce dimorphism, but if the ventral layer of the wing is unaffected then the overall wing shape is also expected to warp.   It is for these reasons that we will only discuss the results from candidate genes that had consistent effects across multiple treatments.  !%(+!Consensus Among Crosses in Size/Shape Space: In 2005, Horabin put forth the hypothesis that SSD was in part regulated by the Hh pathway. Central to this hypothesis is the condition that ectopic Sxl protein can upregulate Hh signaling, increasing expression of the activating form of the transcription factor Ci. Supposedly an increase in either Sxl or Ci should be sufficient to increase SSD (because females are sensitive to Sxl, while males are not). Conversely, Sxl and Ci should also be necessary to maintain SSD; in the absence of either, SSD should decrease. Our data do not support the claim that Ci and Sxl are sufficient and necessary for SSD when driven endogenously, but rather show that they are necessary for SShD (both) and that Sxl is necessary for SSD.   As for HorabinÕs (2005) hypothesis, our support is somewhat tenuous. We were only able to demonstrate that Sxl was necessary for SSD, which is consistent with data from other studies where Sxl knockdowns produced small female wings relative to males (Yan & Perrimon 2015). What our data do show, however, is that Sxl and Ci appear to be necessary for modulating sex-specific wing shape development. Ostensibly, Sxl could be activating the Hh signaling pathway in such a way as to increase the sex-specific difference in the shape of the wing. This explanation might explain overgrowth in wings exposed to ectopic Sxl. Obviously more data are needed to substantiate this claim. If true, we would expect overexpressions of Ci to rescue the phenotype of Sxl knockdowns. Unfortunately, given the nature of Sex-lethal being lethal when lost, these experiments may prove challenging.   !%(,!While other components of the IIS pathway had little effect on SSD or SShD, InR knockdowns and overexpressions regularly yielded a decrease in SSD. These data are consistent with those in Chapter 3. Activation of IIS was not sufficient to increase whole body SSD (Chapter 3) or wing SSD, but our data suggest that it influences SShD. If this is the case, IIS activation not only controls proper male:female proportions in relative size, but likely also shape. This trend is noticeable before and after correction for allometry, though Foxo knockdowns tended to also influence SShD before allometric correction (data not shown). Differences in SShD attributed to Foxo, therefore, correlate with size.    Perhaps the most exciting find is that Tor has the ability to both increase and decrease SSD. Since Tor is a positive growth regulator, its function is to increase rather than decrease growth. The being said, increases in Tor appear to be increasing one sexÕs size (males) relative to the other (females). These data are also consistent with data from Chapter 3, where additively decreasing Tor activity yielded increased SSD (Figure 3.2.b). Furthermore, another element downstream of Tor, S6K, displayed weak evidence for controlling SSD. In Bx-Gal4 crosses, knockdowns at both temperatures increased SSD, whereas overexpression with A9-Gal4 decreased it. Since S6K does not show consistent support, we should remain skeptical, but given that it showed similar results at both temperatures and in two different drivers, it was worth mentioning.   In order to make a more compelling case for the TOR pathwayÕs ability to regulate wing SSD, further experiments would be necessary. For instance, if overexpression of S6K !%(-!were able to rescue SSD in Tor knockdowns, a much stronger case for TOR pathway involvement could certainly be made. Until we measure actual expression levels of TOR signaling molecules during growth, however, we cannot be entirely certain that TOR is directly involved. Perhaps the most convincing piece of evidence would be to measure expression levels in males and females during the early third larval instar. This is crucial, because according to Chapter 1 it is precisely during this phase that SSD is generated (Testa et al. 2013).   The results from nutrient-sensitive growth pathway manipulations such as IIS/TOR are to be expected if sexually dimorphic traits such as shape and size are condition-dependent. In Chapter 3, I demonstrated that whole body SSD is nutrient dependent AND nutrient-sensitive growth pathway-dependent. Here, the data also suggest that not only is SSD condition dependent, but so is SShD. Given the propensity of sexually dimorphic traits for being condition dependent, this revelation is not very surprising (Andersson 1994; Price et al. 1993; Johnstone 1995). For condition dependence to evolve, a system should have sexually selected traits (e.g. shape and size) that are condition dependent while simultaneously having a large amount of variance for overall condition (Bonduriansky 2007b; Rowe & Houle 1996; Johnstone et al. 2009).   When are sex-specific shape and size determined? Despite the vast literature on sex determination in Drosophila, we know comparatively little about how the sex determination pathway controls sex-specific phenotypes (Cline 1984; Kopp et al. 2000; DiBenedetto et al. 1987; Bopp et al. 1996; Baker & Belote 1983; !%).!Salz & Erickson 2010; Younger-Shepherd et al. 1992; Bopp et al. 1993; Hodgkin 1989; Horabin 2005). We know that Sxl controls dosage compensation and, when females lacking Sxl or it is overexpressed in males, they die. Downstream of Sxl, is the tra gene. This gene is known from knockdown experiments to cause transformations from female to male (Cline 1984; Verhulst et al. 2010). It is unclear though, if tra is also causing a transformation from female to male size.   Our data suggest that proper functioning of Sxl and tra are necessary for SSD; no other downstream elements demonstrated such a strong necessity. In the case of tra, overexpression of the female-specific isoform causes a reduction in SSD. Additionally, knockdowns in Sxl were shown to cause an increase in SShD. We believe this is related to the fact that Sxl mutants are lethal for females. With decreasing levels of Sxl, females become not only smaller, but developmentally impaired. Since both Sxl and InR had such similar effects, we feel these data may add support to the claim made by Fear et al. (2015), where InR expression was correlated with Sxl. Further research is needed to confirm the relationship between Sxl and InR.   Somewhat surprisingly, SShD was never reduced in any of these crosses. If tra was in fact controlling SShD, a knockdown would be expected to reduce SShD. However unlikely, it is feasible that tra does not control secondary sexual characteristics, but only primary sexual characteristics. Further experimentation is needed to be certain.   !%)%!Regardless, since no other components were capable of altering SSD, we can fairly confidently conclude that SSD is controlled at the level of tra. Since it is the farthest gene in the pathway to influence SSD, tra expression must be modulating sex-specific growth.  !Wing SSD vs Whole Body SSD: Insights into Regulation of Allometry Evidence from the literature suggests that the genetics of size and shape are often found to be uncorrelated (Gidaszewski et al. 2009; Gilchrist and Partridge 2001; Zimmerman et al. 2000). Because of this, it is not surprising that our results demonstrate few cases where both size and shape are affected and also show little support for similar regulation of organ and body size. That shape and size are genetically uncorrelated also has broader implications for the study of shape and size. First, size or shape cannot be used as proxies for one another, since they are uncorrelated. Second, selection may need to act specifically on either shape or size dimorphism in order to influence their evolution; selection on larger female relative to male body size is therefore not likely to alter aspects of shape dimorphism. Condition dependence in size and shape would then be more likely to evolve independently of one another.   Our data, when compared to Chapter 3Õs data, suggest that organ size and body size are controlled by very different mechanisms. While Chapter 3 argues that SSD cannot be increased with systemic gene overexpression, here the data paint a different picture. Our data confirm that overexpression of genes in specific tissues is sufficient to increase SSD and SShD. That ubiquitous expression fails to increase SSD suggests !%)&!that dimorphic trait values are generated not by systemic increases in specific genes/pathways, but rather by specific spatiotemporal effects.   Sex-specific isoforms of sex-determination genes are often generated through post-translational modification to the active protein. In addition, in the cases of tra and dsx, the mRNAs are alternatively spliced to produce proteins of varying activity. We know that sex-specific isoforms of proteins like these are already linked to cis-regulatory changes in sex-specific phenotypes. For example, male coloration is mediated in part by dsx and bric-a-brac (bab) expression in the thorax. Cis-regulatory mechanisms upstream of bab allow for male-specific isoform of dsx to drive expression of bab, increasing melanin content in the last few tergites (Williams et al. 2008).   Perhaps the most promising direction to search for the genetic mechanisms of SSD and SShD are in the cis-regulatory effects of these candidate genes, specifically those of the TOR pathway. Our data allow us to put forth the hypothesis that differences in TOR signaling are influencing SSD.       !%)'! CHAPTER 6:  Conclusion and future directions  !%)(!Summary of Chapters Chapter 1 ¥ Sexual Size Dimorphism (SSD), the difference in body size between males and females, is a common phenomenon. ¥ Males and females usually share most of their genomes, except for the sex chromosomes. Because both sexes largely share a genome, the evolution of SSD should theoretically be very difficult. Despite this, SSD is nearly ubiquitous among metazoans. ¥ SSD can ultimately evolve through one of three mechanisms: sexual selection, fecundity selection, or viability selection. While we also understand the proximate mechanisms of sex-specific gene expression, we do not yet understand the proximate mechanisms of SSD. ¥ Drosophila has specific developmental patterns for us to consider when teasing apart mechanisms. Drosophila are holometabolous and have discrete, hormonally controlled life stages. Critical size, the point at which starvation no longer delays pupariation, is a good proxy for us to study initial body size.  ¥ We know a lot about the genetics of how growth is controlled in Drosophila, but we do not know how sex-specific growth is regulated to produce SSD ¥ Sexual Shape Dimorphism (SShD) is another important factor to consider when studying sexual dimorphism. Shape, however, is very difficult to quantify. Geometric morphometrics is currently our best method of quantifying shape.  ¥ The genetics of SSD and SShD within organs is also relatively unknown. We have little idea how genes that contribute to growth affect either SSD vs SShD.  !%))! Chapter 2 ¥ Females reach their critical size at a larger weight than males do.  ¥ Females have a faster growth rate than males during their third larval instar.  ¥ Both males and females hit a peak larval weight approximately halfway through their third instar, then begin to lose weight before pupariating. Females lose significantly more weight than males during this period. ¥ Males and females both lose weight during their pupal stages, but not in a sex-specific manner. ¥ Flies mutant for the Insulin Receptor (InR) completely lose SSD.  Chapter 3 ¥ Flies reared on poor-quality diets lose SSD. SSD is therefore condition-dependent. ¥ When genes within nutrient-sensitive growth pathways are perturbed, SSD is often reduced or eliminated. Overexpression of growth pathway genes, however, causes an unexpected reduction in total body size for males and females; SSD is typically unaffected by this effect. ¥ IIS pathway genes are necessary for SSD, but not sufficient to increase it. Genes in the TOR pathway, however, demonstrate greater potential for regulating SSD.  !%)*!¥ Sex-specific control of whole body SSD is controlled at the level of tra within the sex determination pathway. Downstream gene, dsx, is sufficient to decrease SSD.  Chapter 4 ¥ Few mutants within EGFR and TGF-§ pathways affected SSD or SShD in a sex-specific manner. Only a handful of genes associated with growth act in a sex-dependent manner. ¥ Most mutations with sex-specific effects on SSD/SShD did so in a background-dependent manner. Many mutants only affected SSD or SShD within either Sam or Ore genetic backgrounds; very few were significant in both backgrounds.  ¥ The response to mutation is also somewhat background-dependent. Mutations in Ore tended to reduce SSD, while mutants in Sam tended to increase SSD. ¥ It is important to distinguish between relative contributions of SSD vs SShD when analyzing the effects of candidate genes.  Chapter 5 ¥ Most gene manipulations tended to increase SSD or SShD, rather than decrease. While some treatments reduced SSD, none reduce SShD. ¥ It is always important to investigate and assess the relationship between size and shape. In the event that size and shape covary differently across groups, the allometric effects of size on shape can and should be removed.  ¥ SShD increases in perturbations of the Hh pathway. This is caused by exaggerated female response to genetic knockdowns of Ci and Sxl.  !%)+!¥ Again, IIS pathway genes tend to be necessary for SSD, but not sufficient to increase it. TOR pathway genes, however, show additional support for controlling SSD.  ¥ Sex-specific control of wing SSD is also controlled at the level of tra within the sex determination pathway.  ¥ The upstream gene, Sxl, is necessary for SSD and proper SShD. Sxl knockdowns yielded some of the most dramatic sex-specific shape responses by females.  ¥ Our data support the claim made by Fear et al. (2015) that Sxl expression is correlated with InR, due to the similar effects of both.  ¥ Wing and body SSD are potentially controlled by different mechanisms.   Future directions to strengthen our arguments  The first way we intend to strengthen our argument would be to improve the quality and nature of our controls. Currently, our statistical estimates are conservative in such a way that only those crosses that are extremely different from the controls show significance. Specifically, this is because each treatment is compared to a pool of controls, which itself has a rather large variance. Additional controls would be sufficient to increase our power to detect significant effects of additional treatment groups. We are presently working on rearing up the proper controls to add to the analyses, which will be included in the eventual publication of these data.   !%),!Perhaps the most obvious way to further strengthen our arguments would be to specifically design more complicated crosses to detect each candidateÕs effect on SSD vs SShD. Currently, our data support the hypothesis that TOR signaling is involved in regulating SSD to some extent. If we could demonstrate that S6K overexpressions could rescue a Tor knockdown, then we would add a considerable level of support for the TOR activity hypothesis. Further crosses could refine and elucidate our understanding of how these pathways contribute to SSD and SShD.  An even greater amount of support would come from collecting gene expression data from some of the promising crosses in Chapters 3 and 5. By demonstrating the effect that our experiments are having on gene expression, we can be more certain that our experiments are having the intended effect. With these data, we could quantify the effect of each gene on SSD more quantitatively, rather than qualitatively. This will be an important piece of data that would be vital to our arguments. Additionally, and more importantly, we would be able to demonstrate that it is in fact increased TOR expression that is causing changes to SSD. Without this critical piece of data, we cannot be completely certain that TOR pathway genes are directly affecting SSD rather than indirectly influencing it.        !%)-!                     APPENDIX !%*.!Table A1a. Significance Table (raw p values) for effects on wing shape  where G= genotype, S= sex, B= background and Cs= (centroid) size    Mutant Line Allele Pathway G G x S G x B aos 2513 W11 Egfr 0.001 0.671 0.14 omb 3045 md653 TGF-" 0.001 0.001 0.003 cv-2 6342 225-3 TGF-" 0.092 0.363 0.093 GAP1 6372 mip-w[+] Egfr 0.001 0.034 0.001 ksr 10212 J5E2 Egfr 0.001 0.935 0.193 dad 10305 J1E4 TGF-" 0.001 0.054 0.593 drk 10372 k02401 Egfr 0.001 0.046 0.482 bs/DSRF 10413 k07909 Egfr 0.001 0.111 0.001 s 10418 k09530 Egfr 0.001 0.463 0.019 spi 10462 s3547 Egfr 0.001 0.518 0.003 mad 10474 k00237 TGF-" 0.001 0.56 0.001 ed 10490 k01102 Egfr 0.001 0.88 0.012 tsh 10842 A3-2-66 TGF-" 0.095 0.458 0.397 cos 11156 k16101 Hh 0.011 0.228 0.001 tkv 11191 k19713 TGF-" 0.001 0.121 0.001 babo 11207 k16912 TGF-"/Hh 0.054 0.428 0.047 trl 12088 S2325 TGF-" 0.001 0.136 0.315 rho-AP 12413 BG00314 ? 0.523 0.006 0.005 pka-C1 12752 BG02142 Hh 0.225 0.946 0.356 sbb 12772 BG01610 TGF-" 0.001 0.203 0.038 psq 12916 kg00811 Egfr 0.011 0.38 0.223 osa 12945 kg03117 Chromatin Remodeling 0.001 0.188 0.036 rasGAP 13311 kg02382 Egfr 0.541 0.371 0.001 pnt 13535 kg04968 Egfr 0.002 0.063 0.021 drk 13943 k02401 Egfr 0.095 0.223 0.001 cbl 13944 kg03080 Egfr 0.003 0.459 0.018 mam 14189 kg02641 N/Egfr 0.001 0.908 0.577 rho-6 14208 kg05638 Egfr 0.809 0.35 0.814 dpp 14268 kg04600 TGF-" 0.006 0.898 0.022 pka-C3 14345 kg00222 Hh 0.177 0.045 0.035 p38b 14364 kg01337 TGF-"/Egfr 0.247 0.59 0.031 tkv 14403 kg01923 TGF-" 0.041 0.111 0.009 wmd 14541 kg07581 Unknown 0.093 0.264 0.04 mad 14578 kg00581 TGF-" 0.002 0.651 0.153 !%*%!Table A1a. (contÕd)               ast 14638 kg07563 Egfr 0.014 0.972 0.074 dpp 14694 kg08191 TGF-" 0.002 0.822 0.074 rho1 14901 kg01774 Egfr? 0.082 0.059 0.353 sax 14920 kg07525 TGF-" 0.046 0.24 0.003 sax* 5404 sax4 TGF-" 0.04 0.001 0.026 egfr 10385 k05115 Egfr 0.001 0.098 0.023 src42A 13751 kg02515 Egfr 0.001 0.057 0.023 rho/stet 14321 kg07115 Egfr 0.001 0.653 0.305 !%*&!Table A1a. (contÕd)      Mutant Line Cs x G x S Cs x G x B G x S x B Cs x G x S x B aos 2513 0.863 0.827 0.324 0.093 omb 3045 0.004 0.001 0.001 0.086 cv-2 6342 0.023 0.822 0.027 0.677 GAP1 6372 0.674 0.881 0.405 0.428 ksr 10212 0.057 0.89 0.364 0.248 dad 10305 0.24 0.058 0.098 0.296 drk 10372 0.478 0.544 0.57 0.796 bs/DSRF 10413 0.157 0.163 0.437 0.165 s 10418 0.849 0.23 0.518 0.002 spi 10462 0.495 0.52 0.438 0.573 mad 10474 0.542 0.464 0.125 0.094 ed 10490 0.07 0.051 0.542 0.046 tsh 10842 0.947 0.775 0.165 0.262 cos 11156 0.267 0.096 0.288 0.36 tkv 11191 0.358 0.813 0.26 0.161 babo 11207 0.465 0.51 0.429 0.1 trl 12088 0.046 0.82 0.997 0.23 rho-AP 12413 0.983 0.815 0.196 0.098 pka-C1 12752 0.032 0.89 0.825 0.601 sbb 12772 0.323 0.647 0.02 0.539 psq 12916 0.979 0.732 0.03 0.01 osa 12945 0.876 0.373 0.043 0.123 rasGAP 13311 0.037 0.038 0.286 0.307 pnt 13535 0.173 0.752 0.504 0.609 drk 13943 0.514 0.22 0.107 0.222 cbl 13944 0.765 0.57 0.947 0.195 mam 14189 0.298 0.035 0.245 0.066 rho-6 14208 0.448 0.72 0.152 0.25 dpp 14268 0.285 0.058 0.912 0.025 pka-C3 14345 0.587 0.276 0.1 0.018 p38b 14364 0.566 0.906 0.771 0.902 tkv 14403 0.222 0.835 0.501 0.269 wmd 14541 0.651 0.157 0.039 0.378 mad 14578 0.815 0.199 0.216 0.753 ast 14638 0.211 0.934 0.666 0.781 dpp 14694 0.162 0.878 0.421 0.424 !%*'!Table A1a. (contÕd)     rho1 14901 0.083 0.677 0.082 0.03 sax 14920 0.31 0.519 0.5 0.389 sax* 5404 0.205 0.701 0.808 0.795 egfr 10385 0.12 0.112 0.077 0.316 src42A 13751 0.054 0.212 0.617 0.13 rho/stet 14321 0.064 0.558 0.102 0.327 !%*(!Table A1b. Sample size of treatments, where M= male, F= female, w= wild-type, m= mutant, O= Oregon-R background, S= Samarkand background       Mutant Line M, w, O F, w, O M, w, S F, w, S M, m, O F, m, O M, m, S F, m, S aos 2513 21 20 21 20 20 20 20 22 omb 3045 19 18 20 20 12 19 19 20 cv-2 6342 10 11 20 20 10 10 20 20 GAP1 6372 19 18 20 19 18 19 20 20 ksr 10212 17 20 19 20 20 20 22 20 dad 10305 20 20 18 20 20 20 20 20 drk 10372 20 20 20 20 20 20 20 21 bs/DSRF 10413 19 19 20 20 20 19 20 20 s 10418 20 20 18 21 20 20 20 20 spi 10462 18 20 20 20 20 19 21 21 mad 10474 19 20 20 22 20 20 18 22 ed 10490 20 20 19 20 20 20 21 21 tsh 10842 19 21 20 20 20 19 22 19 cos 11156 19 20 19 20 22 19 20 18 tkv 11191 20 20 20 21 20 21 18 17 babo 11207 19 21 20 20 20 20 20 20 trl 12088 20 20 21 20 20 21 20 20 rho-AP 12413 20 20 20 20 20 19 20 19 pka-C1 12752 20 20 18 20 20 20 21 21 sbb 12772 20 20 20 19 20 22 20 21 psq 12916 21 20 21 20 20 20 20 21 osa 12945 31 30 20 20 30 24 20 20 rasGAP 13311 19 19 20 19 21 21 20 20 pnt 13535 20 20 20 19 20 19 20 20 drk 13943 20 19 20 21 16 17 21 21 cbl 13944 20 20 21 20 20 20 20 20 mam 14189 10 8 21 21 10 10 20 20 rho-6 14208 19 21 20 20 19 20 20 21 dpp 14268 21 20 19 20 20 18 22 18 pka-C3 14345 21 20 20 20 21 20 20 20 p38b 14364 10 10 20 20 10 6 20 20 tkv 14403 11 10 10 12 10 11 20 20 !%*)!Table A1b. (contÕd)     wmd 14541 20 20 20 21 20 21 21 20 mad 14578 20 19 22 21 21 19 20 20 ast 14638 20 20 20 20 20 18 20 20 dpp 14694 20 19 19 20 20 20 20 20 rho1 14901 20 20 20 21 20 20 20 20 sax 14920 19 19 20 20 20 20 21 20 sax* 5404 20 21 18 19 20 20 20 20 egfr 10385 20 19 20 20 20 20 20 20 src42A 13751 19 20 19 20 19 20 19 20 rho/stet 14321 19 19 19 20 19 21 19 19 !%**!Table A2. Significance Table (raw p values) for effects on wing shape after the effects of allometry are removed where G= genotype, S= sex, B= background and Cs= (centroid) size.    Mutant Line ID G G x S G x B Cs x G x S Cs x G x B G x S x B Cs x G x S x B aos 2513 0.098 0.221 0.001 0.664 0.31 0.236 0.625 omb 3045 0.001 0.072 0.001 0.6 0.311 0.002 0.264 cv-2 6342 0.066 0.215 0.171 0.494 0.412 0.619 0.208 GAP1 6372 0.002 0.235 0.001 0.075 0.012 0.177 0.299 ksr 10212 0.008 0.005 0.002 0.165 0.24 0.165 0.526 dad 10305 0.16 0.009 0.797 0.158 0.003 0.962 0.205 drk 10372 0.005 0.174 0.031 0.115 0.021 0.021 0.607 bs/DSRF 10413 0.001 0.126 0.001 0.204 0.021 0.005 0.427 s 10418 0.001 0.002 0.106 0.377 0.188 0.009 0.107 spi 10462 0.001 0.527 0.002 0.558 0.371 0.025 0.359 mad 10474 0.001 0.019 0.005 0.594 0.656 0.005 0.746 ed 10490 0.001 0.778 0.591 0.998 0.225 0.393 0.164 tsh 10842 0.008 0.216 0.002 0.837 0.828 0.051 0.988 cos 11156 0.001 0.019 0.037 0.025 0.03 0.036 0.375 tkv 11191 0.001 0.004 0.006 0.404 0.947 0.301 0.135 babo 11207 0.048 0.427 0.02 0.208 0.517 0.721 0.092 trl 12088 0.001 0.001 0.04 0.776 0.891 0.005 0.38 rho-AP 12413 0.075 0.444 0.017 0.805 0.69 0.194 0.009 pka-C1 12752 0.001 0.227 0.107 0.081 0.513 0.002 0.204 sbb 12772 0.001 0.537 0.004 0.703 0.001 0.396 0.069 psq 12916 0.004 0.283 0.143 0.677 0.127 0.048 0.823 osa 12945 0.001 0.255 0.02 0.699 0.492 0.728 0.368 rasGAP 13311 0.027 0.666 0.031 0.565 0.587 0.792 0.607 pnt 13535 0.016 0.199 0.111 0.439 0.199 0.659 0.293 drk 13943 0.002 0.005 0.001 0.414 0.905 0.081 0.051 cbl 13944 0.138 0.07 0.329 0.148 0.002 0.754 0.518 mam 14189 0.001 0.15 0.043 0.001 0.109 0.043 0.386 rho-6 14208 0.086 0.091 0.433 0.664 0.959 0.009 0.099 dpp 14268 0.017 0.322 0.08 0.798 0.109 0.871 0.822 !%*+!Table A2. (contÕd)  pka-C3 14345 0.019 0.024 0.567 0.151 0.237 0.345 0.23 p38b 14364 0.003 0.498 0.171 0.542 0.001 0.466 0.6 tkv 14403 0.021 0.225 0.023 0.403 0.673 0.454 0.395 wmd 14541 0.028 0.118 0.435 0.248 0.599 0.001 0.172 mad 14578 0.001 0.038 0.026 0.639 0.456 0.357 0.11 ast 14638 0.078 0.091 0.002 0.995 0.317 0.007 0.132 dpp 14694 0.001 0.299 0.045 0.476 0.027 0.56 0.801 rho1 14901 0.221 0.362 0.436 0.366 0.329 0.394 0.244 sax 14920 0.002 0.394 0.38 0.758 0.272 0.135 0.338 sax* 5404 0.007 0.007 0.047 0.442 0.929 0.299 0.845 egfr 10385 0.001 0.125 0.03 0.742 0.393 0.014 0.211 src42A 13751 0.031 0.37 0.002 0.384 0.195 0.411 0.586 rho/stet 14321 0.002 0.156 0.19 0.02 0.12 0.207 0.421    !%*,!     Figure A1. Magnitude of SSD and SShD for 42 mutants in Oregon-R (left) and Samarkand (right) wild type backgrounds, after correcting for the influence of allometry (shape on size). A common allometry relationship was assumed across genotype and sex within each background and line combination. Residuals from the allometric model were then used for the analysis. This figure is otherwise identical to Figure 2 (which does not correct for allometry). The effect of each mutant is mapped out in a size-and-shape dimorphism space. Genotypic means for each mutant are indicated by point style and connected by a solid line. SSD is plotted on each x-axis for all plots and SShD is displayed on the y-axis. The plots above display the entire range of variation observed, while those below display only the area with the highest density of points. Lines with significant sex-by-genotype effects are highlighted as follows: effect on both size and shape, shape only and size only. Only significant genes (after sequential Bonferroni correction) from the linear models are colored. Error bars are 95% confidence intervals (unadjusted alpha). All gene names are displayed lower-case, regardless of dominance.   !%*-!                       REFERENCES  !%+.!REFERENCES  Abbott, J.K., Bedhomme, S. & Chippindale,  a. K., 2010. Sexual conflict in wing size and shape in Drosophila melanogaster. Journal of Evolutionary Biology, 23(9), pp.1989Ð1997.  Abouheif, E. & Fairbairn, D.J., 1997. A comparative analysis of allometry for sexual size dimorphism: assessing RenschÕs rule. The American Naturalist, 149(3), pp.540Ð562.  Abu-Shaar, M. & Mann, R.S., 1998. Generation of multiple antagonistic domains along the proximodistal axis during Drosophila leg development. Development (Cambridge, England), 125(19), pp.3821Ð30.  Adkins-Regan, E. & Reeve, H.K., 2014. Sexual dimorphism in body size and the origin of sex-determination systems. The American naturalist, 183(4), pp.519Ð36.  Alberts, S.C., Altmann, J. & Wilson, M.L., 1996. Mate guarding constrains foraging activity of male Baboons. Animal Behaviour, 51(Enomoto 1974), pp.1269Ð1277. Allen, C.E., Zwaan, B.J. & Brakefield, P.M., 2011. Evolution of sexual dimorphism in the Lepidoptera. Annual review of entomology, 56, pp.445Ð464.  Andersson, M. & Simmons, L.W., 2006. Sexual selection and mate choice. Trends in Ecology & Evolution, 21(6), pp.296Ð302.  Andersson, M.B., 1994. Sexual Selection, Princeton University Press. Armesilla, A.L. et al., 2004. Novel functional interaction between the plasma membrane Ca2+ pump 4b and the proapoptotic tumor suppressor Ras-associated factor 1 (RASSF1). Journal of Biological Chemistry, 279(30), pp.31318Ð31328. Arnqvist, G. & Henriksson, S., 1997. Sexual cannibalism in fishing spider and a model for the evolution of sexual cannibalism based on genetic constraints. Evolutionary Ecology, (11), pp.255Ð273. Aron, L. et al., 2010. Pro-survival role for ParkinsonÕs associated gene DJ-1 revealed in trophically impaired dopaminergic neurons. PLoS Biology, 8(4). Ashburner, M., 1989. Drosophila. A laboratory handbook., Cold Spring Harbor Laboratory Press. Azevedo, R.B.R., French, V. & Partridge, L., 2002. Temperature modulates epidermal cell size in Drosophila melanogaster. Journal of Insect Physiology, 48(2), pp.231Ð237. Azevedo, R.B.R., Vernon French & Linda Partridge, 1997. Life!History Consequences of !%+%!Egg Size in Drosophila melanogaster. The American Naturalist, 150(2), pp.250Ð282.  Badyaev, A. V., 2002. Growing Apart: An Ontogenetic Perspective on the Evolution of Sexual Size Dimorphism. Trends in Ecology & Evolution, 17(8), pp.369Ð378. Baker, B.S. & Belote, J.M., 1983. Sex determination and dosage compensation in Drosophila melanogaster. Ann. Rev. Genet., 17, pp.345Ð393. Baker, B.S. & Ridge, K.A., 1980. Sex and the single cell. I. On the action of major loci affecting sex determination in Drosophila melanogaster. Genetics, 94(2), pp.383Ð423. Baker, B.S. & Wolfner, M.F., 1988. A molecular analysis of doublesex, a bifunctional gene that controls both male and female sexual differentiation in Drosophila melanogaster. Genes & Development, 2(4), pp.477Ð489.  Barmina, O. et al., 2005. Sex- and segment-specific modulation of gene expression profiles in Drosophila. Developmental biology, 288(2), pp.528Ð44.  Barmina, O. & Kopp, A., 2007. Sex-specific expression of a HOX gene associated with rapid morphological evolution. Developmental biology, 311(2), pp.277Ð86.  Bateman, A.J., 1948. Intra-sexual selection in Drosophila. Heredity, 2, pp.349Ð368. Bejarano, F. et al., 2012. A genome-wide transgenic resource for conditional expression of Drosophila microRNAs. Development, 139(15), pp.2821Ð2831. Bell, G., 1978. The evolution of anisogamy. Journal of Theoretical Biology, 73(2), pp.247Ð270. Belote, J.M. & Baker, B.S., 1982. Sex determination in Drosophila melanogaster: analysis of transformer-2, a sex-transforming locus. Proceedings of the National Academy of Sciences of the United States of America, 79(5), pp.1568Ð72.  Bininda-Edmonds, O.R.P. & Gittleman, J.L., 2000. Are Pinnipeds Functionally Different from Fissiped Carnivores? The Importance of Phylogenetic Comparative Analyses. Evolution, 54(3), pp.1011Ð1023. Bird, M. a. & Schaffer, H.E., 1972. A study of the genetic basis of the sexual dimorphism for wing length in Drosophila melanogaster. Genetics, 72(3), pp.475Ð487. Birkhead, T.R. & Clarkson, K., 1980. Mate Selection and Precopulatory Guarding in Gammarus pulex. Zeitschrift fr Tierpsychologie, 52, pp.365Ð380. Birkhead, T.R., Lee, K.E. & Young, P., 1988. Sexual Cannibalism in the Praying Mantis Hierodula Membranacea. Behaviour, 106(1), pp.112Ð118. !%+&!Blair, S.S., 2003. Genetic mosaic techniques for studying Drosophila development. Development (Cambridge, England), 130(21), pp.5065Ð72.  Blanckenhorn, W.U., 2005. Current Issues Ð Perspectives and Reviews Behavioral Causes and Consequences of Sexual Size Dimorphism. Ethology, 1016, pp.977Ð1016. Blanckenhorn, W.U. et al., 2007. Proximate Causes of RenschÕs Rule: Does Sexual Size Dimorphism in Arthropods Result from Sex Differences in Development Time? The American Naturalist, 169(2), pp.245Ð57. Blanckenhorn, W.U., 2011. The evolution of Body Size: What keeps organisms small? The Quarterly Review of Biology, 75(4), pp.385Ð407. Blanckenhorn, W.U. et al., 2006. When Rensch meets Bergmann: does sexual size dimorphism change systematically with latitude? Evolution; international journal of organic evolution, 60(10), pp.2004Ð11.  Bıhni, R. et al., 1999. Autonomous Control of Cell and Organ Size by CHICO, a Drosophila Homolog of Vertebrate IRS1-4. Cell, 97(7), pp.865Ð75. Bolstad, G.H. et al., 2015. Complex constraints on allometry revealed by artificial selection on the wing of Drosophila melanogaster. Proceedings of the National Academy of Sciences, 112(43), p.201505357.  Bonduriansky, R., 2007a. Sexual selection and allometry: a critical reappraisal of the evidence and ideas. Evolution; international journal of organic evolution, 61(4), pp.838Ð49.  Bonduriansky, R., 2007b. The evolution of condition-dependent sexual dimorphism. The American naturalist, 169(1), pp.9Ð19. Bonduriansky, R. & Chenoweth, S.F., 2009. Intralocus sexual conflict. Trends in ecology & evolution, 24(5), pp.280Ð8.  Bonduriansky, R. & Day, T., 2003. The evolution of static allometry in sexually selected traits. Evolution, 57(11), pp.2450Ð8.  Bonduriansky, R. & Rowe, L., 2016. Intralocus Sexual Conflict and the Genetic Architecture of Sexually Dimorphic Traits in Prochyliza xanthostoma (Diptera: Piophilidae). Society for the Study of Evolution, 59(9), pp.1965Ð1975. Bookstein, F.L., 1985. Morphometrics in evolutionary biology: the geometry of size and shape change, with examples from fishes. Academy of Natural Sciences. Bopp, D. et al., 1993. Expression of the Sex-lethal gene is controlled at multiple levels during Drosophila oogenesis. Development (Cambridge, England), 118(3), pp.797Ð812.  !%+'!Bopp, D. et al., 1996. Sex-specific control of Sex-lethal is a conserved mechanism for sex determination in the genus Drosophila. Development (Cambridge, England), 122(3), pp.971Ð82.  Brennan, P.L.R., Clark, C.J. & Prum, R.O., 2010. Explosive eversion and functional morphology of the duck penis supports sexual conflict in waterfowl genitalia. Proceedings. Biological sciences / The Royal Society, 277(1686), pp.1309Ð14.  Britton, J.S. et al., 2002. DrosophilaÕs Insulin/PI3-Kinase Pathway Coordinates Cellular Metabolism with Nutritional Conditions. Developmental Cell, 2(2), pp.239Ð249. Brogiolo, W. et al., 2001. An evolutionarily conserved function of the Drosophila insulin receptor and insulin-like peptides in growth control. Current biology": CB, 11(4), pp.213Ð21.  Brosseau, G.E., 1960. Genetic analysis of the male fertility factors on the Y chromosome of Drosophila melanogaster. Genetics, 45(3), pp.257Ð274. Burtis, K.C. & Baker, B.S., 1989. Drosophila doublesex gene controls somatic sexual differentiation by producing alternatively spliced mRNAs encoding related sex-specific polypeptides. Cell, 56(6), pp.997Ð1010.  Caldwell, P.E., Walkiewicz, M. & Stern, M., 2005. Ras Activity in the Drosophila Prothoracic Gland Regulates Body Size and Developmental Rate via Ecdysone Release. Current Biology, 15(23), pp.1785Ð95. Carreira, V.P. et al., 2011. Genetic basis of wing morphogenesis in Drosophila: sexual dimorphism and non-allometric effects of shape variation. BMC developmental biology, 11(1), p.32.  Carreira, V.P., Mensch, J. & Fanara, J.J., 2009. Body size in Drosophila: genetic architecture, allometries and sexual dimorphism. Heredity, 102(3), pp.246Ð56.  Carvalho, A.B., 2002. Origin and evolution of the Drosophila Y chromosome. Current opinion in genetics & development, 12(6), pp.664Ð668.  Chandler, C.H., Chari, S. & Dworkin, I., 2013. Does your gene need a background check? How genetic background impacts the analysis of mutations, genes, and evolution. Trends in Genetics, 29(6), pp.358Ð366.  Chapman, T. et al., 1995. Cost of Mating in Drosophila melanogaster females is mediated by male accessory gland products. Nature, 373, pp.241Ð244. Chapman, T., 2001. Seminal fluid-mediated fitness traits in Drosophila. Heredity, 87(Pt 5), pp.511Ð21.  Chen, C., Jack, J. & Garofalo, R., 1996. The Drosophila Insulin Receptor is Required for Normal Growth*. Endocrinology, 137(3), pp.846Ð856. !%+(!Cheng, R.C. & Kuntner, M., 2015. Disentangling the Size and Shape Components of Sexual Dimorphism. Evolutionary Biology, 42(2), pp.223Ð234. Chown, S.L. & Gaston, K.J., 2010. Body size variation in insects: a macroecological perspective. Biological reviews of the Cambridge Philosophical Society, 85(1), pp.139Ð69.  Cline, T.W., 1984. Autoregulatory functioning of a Drosophila gene product that establishes and maintains the sexually determined state. Genetics, 107(2), pp.231Ð77.  Cline, T.W., 1993. The Drosophila sex determination signal: how do flies count to two? Trends in Genetics, 9(11), pp.385Ð390. Cline, T.W. & Meyer, B.J., 1996. Vive la Difference: Males vs females in flies vs worms. Annual Review of Genetics, 30(31), pp.637Ð702.  Collyer, M.L., Sekora, D.J. & Adams, D.C., 2015. A method for analysis of phenotypic change for phenotypes described by high-dimensional data. Heredity, 115(4), pp.357Ð365.  Cox, R.M. & Calsbeek, R., 2010. Sex-specific selection and intraspecific variation in sexual size dimorphism. Evolution; international journal of organic evolution, 64(3), pp.798Ð809.  Cox, R.M. & Calsbeek, R., 2009. Sexually Antagonistic Selection, Sexual Dimorphism, and the Resolution of Intralocus Sexual Conflict. The American Naturalist, 173(2), pp.176Ð187. Cox, R.M., Stenquist, D.S. & Calsbeek, R., 2009. Testosterone, growth and the evolution of sexual size dimorphism. Journal of evolutionary biology, 22(8), pp.1586Ð98.  CPR, 2015. Houle Lab: Software. Di Cristofano, A. & Pandolfi, P.P., 2000. The Multiple Roles of PTEN in Tumor Suppression. Cell, 100(4), pp.387Ð390.  DÕAmico, L.J., Davidowitz, G. & Nijhout, H.F., 2001. The Developmental and Physiological Basis of Body Size Evolution in an Insect. Proceedings of the Royal Society B: Biological Science, 268(1476), pp.1589Ð93. Darwin, C., 1874. The Descent of Man: And Selection in Relation to Sex Revised ed. J. Wanamaker, ed., Philadelphia. David, J.R. et al., 2003. Genetic variability of sexual size dimorphism in a natural population of Drosophila melanogaster: an isofemale-line approach. Journal of genetics, 82(3), pp.79Ð88.  !%+)!Davidowitz, G. et al., 2004. The effects of environmental variation on a mechanism that controls insect body size. Evolutionary Ecology, 6(1), pp.49Ð62.  Davidowitz, G., DÕAmico, L.J. & Nijhout, H.F., 2003. Critical Weight in the Development of Insect Body Size. Evolution & Development, 5(2), pp.188Ð197. Dawkins, R. & Krebs, J.R., 1979. Arms Races between and within Species. Proceedings of the Royal Society B: Biological Sciences, 205(1161), pp.489Ð511.  Day, S.J. & Lawrence, P. a, 2000. Measuring dimensions: the regulation of size and shape. Development (Cambridge, England), 127(14), pp.2977Ð87.  Debat, V. et al., 2016. HSP90 and the Quantitative Variation of Wing Shape in Drosophila melanogaster. Society for the Study of Evolution, 60(12), pp.2529Ð2538. DeNieu, M., Pitchers, W. & Dworkin, I., 2014. Adaptation to a novel predator in Drosophila melanogaster: How well are we able to predict evolutionary responses? bioRxiv, pp.1Ð56. Available at: http://biorxiv.org/content/early/2014/05/19/005322.abstract. Devlin, R.H. & Nagahama, Y., 2002. Sex determination and sex differentiation in fish: An overview of genetic, physiological, and environmental influences. Aquaculture, 208(3-4), pp.191Ð364. DiBenedetto,  a J. et al., 1987. Sequences expressed sex-specifically in Drosophila melanogaster adults. Developmental biology, 119(1), pp.242Ð51.  Van Doorn, G.S. & Kirkpatrick, M., 2010. Transitions between male and female heterogamety caused by sex-antagonistic selection. Genetics, 186(2), pp.629Ð645. Dougherty, M.K. et al., 2005. Regulation of Raf-1 by direct feedback phosphorylation. Molecular Cell, 17(2), pp.215Ð224. Drake, A.G. & Klingenberg, C.P., 2008. The pace of morphological change: historical transformation of skull shape in St Bernard dogs. Proceedings. Biological sciences / The Royal Society, 275(1630), pp.71Ð76. Duffy, J., 2002. GAL4 system in Drosophila: a fly geneticistÕs Swiss army knife. genesis, (34), pp.1-15. Dworkin, I. & Gibson, G., 2006. Epidermal Growth Factor Receptor and Transforming Growth Factor-§ Signaling Contributes to Variation for Wing Shape in Drosophila melanogaster. Genetics, 1431(July), pp.1417Ð1431. Dworkin, I., Anderson, J. A., Idaghdour, Y., Parker, E. K., Stone, E. A. & Gibson, G. 2011 The effects of weak genetic perturbations on the transcriptome of the wing imaginal disc and its association with wing shape in Drosophila melanogaster. !%+*!Genetics 187, 1171Ð1184. (doi:10.1534/genetics.110.125922) Edgar, B. a, 2006. How flies get their size: genetics meets physiology. Nature reviews. Genetics, 7(12), pp.907Ð16. Emlen, D.J. et al., 2012. A mechanism of extreme growth and reliable signaling in sexually selected ornaments and weapons. Science (New York, N.Y.), 337(6096), pp.860Ð4. Emlen, D.J., 1997. Alternative reproductive tactics and male-dimorphism in the horned beetle Onthophagus acuminatus (Coleoptera: Scarabaeidae). Behavioral Ecology and Sociobiology, 41(5), pp.335Ð341.  Emlen, D.J., Hunt, J. & Simmons, L.W., 2005. Evolution of sexual dimorphism and male dimorphism in the expression of beetle horns: phylogenetic evidence for modularity, evolutionary lability, and constraint. American Naturalist, 166(4), pp.42Ð68.  Emlen, D.J., Hunt, J. & Simmons, L.W., 2005. Evolution of Sexual Dimorphism and Male Dimorphism in the Expression of Beetle Horns: Phylogenetic Evidence for Modularity, Evolutionary Lability, and Constraint. The American Naturalist, 166(4), pp.42Ð68.  Erdman, S.E. & Burtis, K.C., 1993. The Drosophila doublesex proteins share a novel zinc finger related DNA binding domain. The EMBO journal, 12(2), pp.527Ð535. Esperk, T. et al., 2007. Achieving High Sexual Size Dimorphism in Insects: Females Add Instars. Ecological Entomology, 32(3), pp.243Ð256. Fabian, D.K. et al., 2012. Genome-wide patterns of latitudinal differentiation among populations of Drosophila melanogaster from North America. Molecular ecology, 21(19), pp.4748Ð69. Fagegaltier, D. et al., 2014. A Genome-Wide Survey of Sexually Dimorphic Expression of Drosophila miRNAs Identifies the Steroid Hormone-Induced miRNA let-7 as a Regulator of Sexual Identity. Genetics, 198(2), pp.647Ð68.  Fairbairn, D.J., 1997a. Allometry for Sexual Size Dimorphism: Pattern and Process in the Coevolution of Body Size in Males and Females. Annual Review of Ecology and Systematics, 28(1), pp.659Ð687. Fairbairn, D.J., 2005. Allometry for Sexual Size Dimorphism: Testing Two Hypotheses for RenschÕs Rule in the Water Strider Aquarius remigis. The American Naturalist, 166(4), pp.69Ð84.  Fairbairn, D.J., 1997b. Allometry For Sexual Size Dimorphism:Pattern and Process in the Coevolution of Body Size in Males and Females. Annual Review of Ecology and Systematics, 28(1997), pp.659Ð687.  !%++!Fairbairn, D.J., 1990. Factors Influencing Sexual Size Dimorphism in Temperate Waterstriders. American Naturalist, 136(1), pp.61Ð86. Fairbairn, D.J. & Blanckenhorn, W.U., 2007. Sex, size, and gender roles": evolutionary studies of sexual size dimorphism, Oxford University Press. Fairbairn, D.J., Blanckenhorn, W.U. & Sz”kely, T., 2007. Sex, size, and gender roles, Oxford University Press.  Fairbairn, D.J. & Roff, D.A., 2006. The quantitative genetics of sexual dimorphism: assessing the importance of sex-linkage. Heredity, 97(5), pp.319Ð328. Fear, J.M. et al., 2015. The wright stuff: reimagining path analysis reveals novel components of the sex determination hierarchy in Drosophila melanogaster. BMC systems biology, 9, p.53.  Ferguson & Fairbairn, 2000. Sex-specific selection and sexual size dimorphism in the waterstrider Aquarius remigis. Journal Of Evolutionary Biology, 13(2), pp.160Ð170. Fischer, K. & Fiedler, K., 2001. Sexual Differences in Life-History Traits in the Butterfly Lycaena tityrus: A Comparison Between Direct and Diapause Development. Entomologia Experimentalis et Applicata, 100, pp.325Ð330. Frankino, W.A. et al., 2005. Natural selection and developmental constraints in the evolution of allometries. Science (New York, N.Y.), 307(5710), pp.718Ð20.  Freeman, M., 1997. Cell determination strategies in the Drosophila eye. Development (Cambridge, England), 124(2), pp.261Ð70. Fuller, R.C., Houle, D. & Travis, J., 2005. Sensory bias as an explanation for the evolution of mate preferences. American Naturalist, 166(4), pp.437Ð446.  Gao, X. & Pan, D., 2001. TSC1 and TSC2 tumor suppressors antagonize insulin signaling in cell growth. Genes & development, 15(11), pp.1383Ð92.  Garami, A. et al., 2003. Insulin activation of Rheb, a mediator of mTOR/S6K/4E-BP signaling, is inhibited by TSC1 and 2. Molecular Cell, 11(6), pp.1457Ð1466. Garc™a-Bellido,  a, Cort”s, F. & Mil⁄n, M., 1994. Cell interactions in the control of size in Drosophila wings. Proceedings of the National Academy of Sciences of the United States of America, 91(21), pp.10222Ð6.  Garcia-Bellido, A. & Merriam, J.R., 1971. Parameters of the Wing Imaginal Disc Development of Drosophila melanogaster. Developmental Biology, 24, pp.61Ð87. Ghosh, S.M., Testa, N.D. & Shingleton, A.W., 2013. Temperature-size rule is mediated by thermal plasticity of critical size in Drosophila melanogaster. Proceedings of the Royal Society B: Biological Sciences, 280(1760), pp.20130174Ð20130174.  !%+,!Gidaszewski, N. a, Baylac, M. & Klingenberg, C.P., 2009. Evolution of sexual dimorphism of wing shape in the Drosophila melanogaster subgroup. BMC Evolutionary Biology, 9(1), p.110.  Gilbert, S.F., 1987. In friendly disagreement: Wilson, Morgan, and the embryological origins of the gene theory. American Zoologist, 27(3), pp.797Ð806. Gilchrist, a S. & Partridge, L. 2001 The contrasting genetic architecture of wing size and shape in Drosophila melanogaster. Heredity (Edinb). 86, 144Ð152. (doi:10.1046/j.1365-2540.2001.00779.x) Glassford, W.J. et al., 2015. Co-option of an Ancestral Hox-Regulated Network Underlies a Recently Evolved Morphological Novelty Article Co-option of an Ancestral Hox-Regulated Network Underlies a Recently Evolved Morphological Novelty. Developmental Cell, 34(5), pp.520Ð531.  Gokhale, R.H. et al., 2016. Intra-Organ Growth Coordination in Drosophila Is Mediated By Systemic Ecdysone Signaling. Developmental Biology.  Gokhale, R.H. & Shingleton, A.W., 2015. Size control: the developmental physiology of body and organ size regulation. Wiley Interdisciplinary Reviews: Developmental Biology, (July), p.n/aÐn/a.  Gotoh, H. et al., 2014. Developmental Link between Sex and Nutrition; doublesex Regulates Sex-Specific Mandible Growth via Juvenile Hormone Signaling in Stag Beetles. PLoS Genetics, 10(1). Gould, S.J., 1966. Allometry and size in ontogeny and phylogeny. Biological reviews of the Cambridge Philosophical Society, 41(4), pp.587Ð640.  Haerry, T.E. et al., 1998. Synergistic signaling by two BMP ligands through the SAX and TKV receptors controls wing growth and patterning in Drosophila. Development, 125(20), pp.3977Ð3987. Head, G., 1995. Selection on Fecundity and Variation in the Degree of Sexual Size Dimorphism Among Spider Species (Class Araneae). Evolution, 49(4), pp.776Ð781.  Hedrick, A. V & Temeles, E.J., 1989. The evolution of sexual dimorphism in animals: Hypotheses and tests. Trends In Ecology & Evolution, 4(5), pp.136Ð138. Heimpel, G.E. & de Boer, J.G., 2008. Sex determination in the Hymenoptera. Annual Review of Entomology, 53, pp.209Ð230.  Herrel,  a et al., 1999. Sexual dimorphism of head size in Gallotia galloti: testing the niche divergence hypothesis by functional analyses. Functional Ecology, 13(1), pp.289Ð297.  Hodgkin, J., 1989. Drosophila sex determination: a cascade of regulated splicing. Cell, !%+-!56(6), pp.905Ð6.  Hoikkala, A., Aspi, J. & Suvanto, L., 1998. Male courtship song frequency as an indicator of male genetic quality in an insect species, Drosophila montana. Proceedings. Biological sciences / The Royal Society, 265(1395), pp.503Ð8.  Holland, B. & Rice, W.R., 1999. Experimental removal of sexual selection reverses intersexual antagonistic coevolution and removes a reproductive load. Proceedings of the National Academy of Sciences of the United States of America, 96(9), pp.5083Ð8.  Horabin, J.I., 2005. Splitting the Hedgehog signal: sex and patterning in Drosophila. Development, 132(21), pp.4801Ð10. Houle, D. et al., 2003. Automated measurement of Drosophila wings. BMC Evolutionary Biology, 3(1), p.25.  Huey, R.B. et al., 2006. Sexual size dimorphism in a Drosophila clade, the D. obscura group. Zoology (Jena, Germany), 109(4), pp.318Ð30.  Huxley, J., 1932. Problems of Relative Growth. Ikeya, T. et al., 2002. Nutrient-dependent expression of insulin-like peptides from neuroendocrine cells in the CNS contributes to growth regulation in Drosophila. Current biology": CB, 12(15), pp.1293Ð300. Jackson, S.M. et al., 1997. dally, a Drosophila glypican, controls cellular responses to the TGF-§-related morphogen, Dpp. Development, 124(November), pp.4113Ð4120. Johnstone, R. a, 1995. Sexual selection, honest advertisement and the handicap principle: reviewing the evidence. Biological reviews of the Cambridge Philosophical Society, 70, pp.1Ð65. Johnstone, R. a, Rands, S. a & Evans, M.R., 2009. Sexual selection and condition-dependence. Journal of evolutionary biology, 22(12), pp.2387Ð94.  De Jong, G. & Bochdanovits, Z., 2003. Latitudinal clines in Drosophila melanogaster: body size, allozyme frequencies, inversion frequencies, and the insulin-signalling pathway. Journal of genetics, 82(3), pp.207Ð23. Kamakura, M., 2011. Royalactin induces queen differentiation in honeybees. Nature, 473(7348), pp.478Ð83.  Karubian, J. & Swaddle, J.P., 2001. Selection on females can create Òlarger malesÓ. Proceedings. Biological sciences / The Royal Society, 268(1468), pp.725Ð8.  Kijimoto, T. & Moczek, A.P., 2016. Hedgehog signaling enables nutrition-responsive inhibition of an alternative morph in a polyphenic beetle. Proceedings of the !%,.!National Academy of Sciences of the United States of America, p.1601505113Ð.  Kijimoto, T., Moczek, A.P. & Andrews, J., 2012. Diversification of doublesex function underlies morph-, sex-, and species-specific development of beetle horns. Proceedings of the National Academy of Sciences of the United States of America, 109(50), pp.20526Ð31.  Kirkwood, T.B.L. et al., 2005. What accounts for the wide variation in life span of genetically identical organisms reared in a constant environment? Mechanisms of ageing and development, 126(3), pp.439Ð43.  Klingenberg, C.P., 2016. Size, shape and form: concepts of allometry in geometric morphometrics. Development Genes and Evolution, (April), pp.1Ð25.  Kodric-Brown, A., Sibly, R.M. & Brown, J.H., 2006. The allometry of ornaments and weapons. Proceedings of the National Academy of Sciences of the United States of America, 103(23), pp.8733Ð8.  Kokko, H. & Jennions, M.D., 2014. The Relationship between Sexual Selection and Sexual Conflict. Cold Spring Harbor Perspectives in Biology, 6(9), pp.a017517Ða017517.  Kopp,  a et al., 2000. Genetic control and evolution of sexually dimorphic characters in Drosophila. Nature, 408(6812), pp.553Ð9.  Koyama, T., Mendes, C.C. & Mirth, C.K., 2013. Mechanisms regulating nutrition-dependent developmental plasticity through organ-specific effects in insects. Frontiers in Physiology, 4 SEP(September), pp.1Ð12. Lande, R., 1981. Models of speciation by sexual selection on polygenic traits. Proceedings of the National Academy of Sciences of the United States of America, 78(6), pp.3721Ð3725. Lande, R., 1980. Sexual Dimorphism, Sexual Selection, and Adaptation in Polygenic Characters. Evolution, 34(2), pp.292Ð305. Lavine, L. et al., 2015. Exaggerated Trait Growth in Insects. Annual Review of Entomology, 60(1), pp.453Ð472.  Liang, L. et al., 2014. Functional genomic analysis of the periodic transcriptome in the developing drosophila wing. Developmental Cell, 29(1), pp.112Ð127.  van der Linde, K. & Houle, D., 2009. Inferring the Nature of Allometry from Geometric Data. Evolutionary Biology, 36(3), pp.311Ð322.  Lindenfors, P., 2002. Sexually antagonistic selection on primate size. Journal of Evolutionary Biology, 15(4), pp.595Ð607. !%,%!Lindenfors, P., Gittleman, J. & Jones, K., 2007. Sexual size dimorphism in mammals. Sex, size and gender roles: evolutionary studies of sexual size dimorphism, pp.19Ð26. Lindenfors, P., Tullberg, B. & Biuw, M., 2002. Phylogenetic analyses of sexual selection and sexual size dimorphism in pinnipeds. Behavioral Ecology and Sociobiology, 52(3), pp.188Ð193.  De Lisle, S.P. & Rowe, L., 2013. Correlated evolution of allometry and sexual dimorphism across higher taxa. The American naturalist, 182(5), pp.630Ð9.  De Lisle, S.P. & Rowe, L., 2015. Ecological Character Displacement between the Sexes. The American Naturalist, 186(6), pp.000Ð000.  Liu, L. et al., 2006. Hypoxia-Induced Energy Stress Regulates mRNA Translation and Cell Growth. Molecular cell, 21(4), pp.521Ð31. Lovich, J.E. & Gibbons, J.W., 1992. A Review of Techniques Quantifying Sexual Size Dimorphism. Growth, Development, and Aging, 56, pp.269Ð281. Macrae, M. et al., 2005. A conditional feedback loop regulates Ras activity through EphA2. Cancer Cell, 8(2), pp.111Ð118. Maine, E.M. et al., 1985. The Sex-lethal gene of Drosophila: DNA alterations associated with sex-specific lethal mutations. Cell, 43(2 Pt 1), pp.521Ð9.  Mank, J.E., 2009. Sex Chromosomes and the Evolution of Sexual Dimorphism: Lessons from the Genome. The American Naturalist, 173(2), pp.141Ð150. Manning, B.D., 2004. Balancing Akt with S6K: Implications for both metabolic diseases and tumorigenesis. Journal of Cell Biology, 167(3), pp.399Ð403. Marshall, L., Rideout, E.J. & Grewal, S.S., 2012. Nutrient/TOR-Dependent Regulation of RNA Polymerase III Controls Tissue and Organismal Growth in Drosophila. The EMBO Journal, pp.1Ð15. Matsuda, H. & Abrams, P.A., 1999. Why are equally sized gametes so rare? The instability of isogamy and the cost of anisogamy. Evolutionary Ecology Research, 1(7), pp.769Ð784. McKeown, M., Belote, J.M. & Baker, B.S., 1987. A molecular analysis of transformer, a gene in Drosophila melanogaster that controls female sexual differentiation. Cell, 48(3), pp.489Ð99.  McKeown, M., Belote, J.M. & Boggs, R.T., 1988. Ectopic expression of the female transformer gene product leads to female differentiation of chromosomally male Drosophila. Cell, 53(6), pp.887Ð95.  !%,&!Menezes, B.F. et al., 2013. The influence of male wing shape on mating success in Drosophila melanogaster. Animal Behaviour, 85(6), pp.1217Ð1223.  Merila, A.J. et al., 2011. Quantitative Genetics of Sexual Size Dimorphism in the Collared Flycatcher, Ficedula albicollis. Society, 52(3), pp.870Ð876. Mesnick, S.L. & Le Boeuf, B.J., 1991. Sexual behavior of male northern elephant seals: II. Female response to potentially injurious encounters. Behaviour, 117(3), pp.262Ð280.  Mirth, C., Truman, J.W. & Riddiford, L.M., 2005. The Role of the Prothoracic Gland in Determining Critical Weight for Metamorphosis in Drosophila melanogaster. Current Biology, 15(20), pp.1796Ð1807. Mirth, C.K. & Riddiford, L.M., 2007. Size Assessment and Growth Control: How Adult Size is Determined in Insects. BioEssays, 29(4), pp.344Ð55. Mitteroecker, P. et al., 2004. Comparison of cranial ontogenetic trajectories among great apes and humans. Journal of Human Evolution, 46(6), pp.679Ð698. Moczek, A.P. & Kijimoto, T., 2014. Development and evolution of insect polyphenisms: Novel insights through the study of sex determination mechanisms. Current Opinion in Insect Science, 1, pp.52Ð58.  de Moed, G.H., De Jong, G. & Scharloo, W., 1997. Environmental effects on body size variation in Drosophila melanogaster and its cellular basis. Genetical research, 70(1), pp.35Ð43.  Molleman, F. et al., 2011. Sexual Differences in Weight Loss upon Eclosion are Related to Life History Strategy in Lepidoptera. Journal of Insect Physiology, 57(6), pp.712Ð22. Moller, A.P. et al., 1999. Good-genes effects in sexual selection. Proceedings of the Royal Society B-Biological Sciences, 266(1414), pp.85Ð91.  Montagne, J. et al., 1999. Drosophila S6 kinase: a regulator of cell size. Science (New York, N.Y.), 285(5436), pp.2126Ð9.  Moore, E.C. & Roberts, R.B., 2013. Polygenic sex determination. Current Biology, 23(12), pp.R510ÐR512.  Mosimann, J.E., 1970. Size Allometry: Size and Shape Variables with Characterizations of the Lognormal and Generalized Gamma Distributions. Journal of the American Statistical Association, 65(330), pp.930Ð945. Moya-LaraŒo, J., Halaj, J. & Wise, D.H., 2002. Climbing to reach females: Romeo should be small. Evolution; international journal of organic evolution, 56(2), pp.420Ð5.  !%,'!Nevill, A.M., Holder, R.L. & Alan, M., 1995. Scaling, normalizing and per ratio an allometric modeling approach standards$: an allometric modeling approach. Journal of Applied Physiology, 79(3), pp.1027Ð1931. Nguyen, M. et al., 1998. Interpretation of a BMP activity gradient in Drosophila embryos depends on synergistic signaling by two type I receptors, SAX and TKV. Cell, 95(4), pp.495Ð506. Nicholson, L. et al., 2008. Spatial and temporal control of gene expression in Drosophila using the inducible GeneSwitch GAL4 system. I. Screen for larval nervous system drivers. Genetics, 178(1), pp.215Ð34.  Nijhout, H., 1981. Physiological Control of Molting in Insects. American Zoologist, 21(3), pp.631Ð640. Nijhout, H.F., 2008. Size matters (but so does time), and itÕs OK to be different. Developmental cell, 15(4), pp.491Ð2.  Nijhout, H.F., 2003. The Control of Body Size in Insects. Developmental Biology, 261(1), pp.1Ð9. Nijhout, H.F., Davidowitz, G. & Roff, D.A., 2006. A Quantitative Analysis of the Mechanism that Controls Body Size in Manduca sexta. Journal of Biology, 5, pp.16.0Ð16.15. Nijhout, H.F. & Williams, C.M., 1974. Control of Moulting and Metamorphosis in the Tobacco Hornworm, Manduca sexta (L.): Growth of the Last-Instar Larva and the Decision to Pupate. The Journal of Experimental Biology, 61(2), pp.481Ð91. Norman, M. et al., 2002. First encounter with a live male blanket octopus: the worldÕs most sexually size-dimorphic large animal. New Zealand Journal of Marine and Freshwater Research, 36, pp.733Ð736. Ogden, S. K., Ascano, M., Stegman, M. A. & Robbins, D. J. 2004 Regulation of Hedgehog signaling: A complex story. Biochem. Pharmacol. 67, 805Ð814. (doi:10.1016/j.bcp.2004.01.002) Okamoto, N. et al., 2009. A fat body-derived IGF-like peptide regulates postfeeding growth in Drosophila. Developmental cell, 17(6), pp.885Ð91. Oldham, S. et al., 2000. Genetic control of size in Drosophila. Philosophical transactions of the Royal Society of London. Series B, Biological sciences, 355(1399), pp.945Ð52.  Orme, M.H. & Leevers, S.J., 2005. Flies on steroids: the interplay between ecdysone and insulin signaling. Cell metabolism, 2(5), pp.277Ð8.  Paaby, A.B. & Rockman, M. V., 2014. Cryptic genetic variation: evolutionÕs hidden !%,(!substrate. Nature Reviews Genetics, 15(4), pp.247Ð258.  Palsson, A. & Gibson, G., 2004. Association between nucleotide variation in Egfr and wing shape in Drosophila melanogaster. Genetics, 167(3), pp.1187Ð1198. Partridge, L. et al., 1994. Evolution and Development of Body Size and Cell Size in Drosophila melanogaster in Response to Temperature. Evolution, 48(4), pp.1269Ð1276. Pearson, D., Shine, R. & How, R., 2006. Sex-specic niche partitioning and sexual size dimorphism in Australian pythons (Morelia spilota imbricata). Biological Journal of the Linnean Society, 77, pp.113Ð125. Peterson, A.J. & OÕConnor, M.B., 2013. Activin receptor inhibition by Smad2 regulates Drosophila wing disc patterning through BMP-response elements. Development (Cambridge, England), 140(3), pp.649Ð59.  Pietsch, T.W., 2005. Dimorphism, parasitism, and sex revisited: modes of reproduction among deep-sea ceratioid anglerfishes (Teleostei: Lophiiformes). Ichthyological Research, 52(3), pp.207Ð236.  Pitchers, W., Pool, J.E. & Dworkin, I., 2013. Altitudinal clinal variation in wing size and shape in african Drosophila melanogaster: One cline or many? Evolution, 67(2), pp.438Ð452. Pitnick, S. et al., 2001. MalesÕ evolutionary responses to experimental removal of sexual selection. Proceedings. Biological sciences / The Royal Society, 268(1471), pp.1071Ð80.  Pitnick, S. & Garc™a-Gonz⁄lez, F., 2002. Harm to females increases with male body size in Drosophila melanogaster. Proceedings. Biological sciences / The Royal Society, 269(1502), pp.1821Ð8.  Prasad, N.G. et al., 2007. An evolutionary cost of separate genders revealed by male-limited evolution. The American naturalist, 169(1), pp.29Ð37. Preziosi, R.F. et al., 1996. Body Size and Fecundity in the Waterstrider Aquarius remigis: A Test of DarwinÕs Fecundity Advantage Hypothesis. Oecologia, 108(3), pp.424Ð431. Preziosi, R.F. & Fairbairn, D.J., 2000. Lifetime Selection on Adult Body Size and Components of Body Size in a Waterstrider: Opposing Selection and Maintenance of Sexual Size Dimorphism. Evolution, 54(2), pp.558Ð566. Price, T., Schluter, D. & Heckman, N.E., 1993. Sexual selection when the female directly benefits. Biological Journal of the Linnean Society, 48(3), pp.187Ð211.  Prober, D.A. & Edgar, B.A., 2000. Ras1 Promotes Cellular Growth in the Drosophila !%,)!Wing. Cell, 100(4), pp.435Ð46. Prokop, Z.M. et al., 2012. Meta-analysis suggests choosy females get sexy sons more than Ògood genes.Ó Evolution, 66(9), pp.2665Ð2673.  Raj, A. & van Oudenaarden, A., 2008. Nature, nurture, or chance: stochastic gene expression and its consequences. Cell, 135(2), pp.216Ð26.  Rea, S.L. et al., 2005. A stress-sensitive reporter predicts longevity in isogenic populations of Caenorhabditis elegans. Nature genetics, 37(8), pp.894Ð898. Reeve, J.P. & Fairbairn, D.J., 1999. Change in sexual size dimorphism as a correlated response to selection on fecundity. Heredity, 83(6), pp.697Ð706.  Reeve, J.P. & Fairbairn, D.J., 2001. Predicting the evolution of sexual size dimorphism. Journal of Evolutionary Biology, 14(2), pp.244Ð254.  Reeve, J.P. & Fairbairn, D.J., 1996. Sexual Size Dimorphism as a Correlated Response to Selection on Body Size$: An Empirical Test of the Quantitative Genetic Model. Society, 50(5), pp.1927Ð1938. Rice, W.R. et al., 2006. Assessing sexual conflict in the Drosophila melanogaster laboratory model system. Philosophical transactions of the Royal Society of London. Series B, Biological sciences, 361(1466), pp.287Ð99.  Rideout, E.J., Marshall, L. & Grewal, S.S., 2012. Drosophila RNA polymerase III repressor Maf1 controls body size and developmental timing by modulating tRNAiMet synthesis and systemic insulin signaling. Proceedings of the National Academy of Sciences of the United States of America, 109(4), pp.1139Ð44. Rideout, E.J., Narsaiya, M.S. & Grewal, S.S., 2015. The Sex Determination Gene transformer Regulates Male-Female Differences in Drosophila Body Size. PLoS Genetics, 11(12), pp.1Ð23. Rohlf, F.J., 2003. TpsDig. Rowe, L. & Houle, D., 1996. The lek paradox and the capture of genetic variance by condition dependent traits. Proceedings: Biological Sciences, 263(1375), pp.1415Ð1421. Rui, L. et al., 2001. Insulin/IGF-1 and TNF-% stimulate phosphorylation of IRS-1 at inhibitory Ser307 via distinct pathways. Journal of Clinical Investigation, 107(2), pp.181Ð189. Salz, H. & Erickson, J., 2010. Sex Determination in Drosophila: The View From the Top. Fly (Austin), 4(1), pp.60Ð70. Sanger, T.J. et al., 2013. Convergent Evolution of Sexual Dimorphism in Skull Shape !%,*!Using Distinct Developmental Strategies. Evolution, 67(8), pp.2180Ð2193.  Schluter, D., Price, T.D. & Rowe, L., 1991. Conflicting Selection Pressures and Life History Trade-Offs. Proceedings of the Royal Society B: Biological Sciences, 246(1315), pp.11Ð17. Schwarzkopf, L., 2005. Sexual Dimorphism in Body Shape Without Sexual Dimorphism in Body Size in Water Skinks (Eulamprus Quoyii). Herpetologica, 61(2), pp.116Ð123. Seger, R. & Krebs, E.G., 1995. The MAPK signaling cascade. The FASEB Journal, 9, pp.726Ð735. Serrano-Meneses, M.A. et al., 2007. Sexual size dimorphism in the American rubyspot: male body size predicts male competition and mating success. Animal Behaviour, 73(6), pp.987Ð997. Shah, O.J., Wang, Z. & Hunter, T., 2004. Inappropriate Activation of the TSC/Rheb/mTOR/S6K Cassette Induces IRS1/2 Depletion, Insulin Resistance, and Cell Survival Deficiencies. Current Biology, 14, pp.1650Ð1656. Shen, J. et al., 2009. A screen of apoptosis and senescence regulatory genes for life span effects when over-expressed in Drosophila. Aging, 1(2), pp.191Ð211.  Shine, R., 1989. Ecological Causes for the Evolution of Sexual Dimorphism: A Review of the Evidence on JSTOR. Quarterly Review of Biology, 64(4), pp.419Ð461.  Shine, R., 1988. The Evolution of Large Body Size in Females: A Critique of DarwinÕs ÒFecundity AdvantageÓ Model. The American Naturalist, 131(1), pp.124Ð131. Shingleton, A.W., 2005. Body-size Regulation: Combining Genetics and Physiology. Current Biology, 15(20), pp.R825Ð7. Shingleton, A.W. et al., 2009. Many ways to be small: different environmental regulators of size generate distinct scaling relationships in Drosophila melanogaster. Proceedings. Biological sciences / The Royal Society, 276(1667), pp.2625Ð33.  Shingleton, A.W. et al., 2007. Size and Shape: The Developmental Regulation of Static Allometry in Insects. BioEssays, 29(6), pp.536Ð48. Shingleton, A.W. et al., 2005. The Temporal Requirements for Insulin Signaling During Development in Drosophila. PLoS biology, 3(9), p.e289. Shingleton, A.W., Mirth, C.K. & Bates, P.W., 2008. Developmental Model of Static Allometry in Holometabolous Insects. Proceedings of the Royal Society B: Biological Sciences, 275(1645), pp.1875Ð85. Shirangi, T.R., Taylor, B.J. & McKeown, M., 2006. A double-switch system regulates !%,+!male courtship behavior in male and female Drosophila melanogaster. Nature genetics, 38(12), pp.1435Ð9.  Simmons, L.W. & Garc™a-Gonz⁄lez, F., 2008. Evolutionary reduction in testes size and competitive fertilization success in response to the experimental removal of sexual selection in dung beetles. Evolution, 62(10), pp.2580Ð2591. Slaidina, M. et al., 2009. A Drosophila Insulin-like Peptide Promotes Growth During Nonfeeding States. Developmental cell, 17(6), pp.874Ð84. Smith, R.J., 1999. Statistics of sexual size dimorphism. Journal of Human Evolution, 36(4), pp.423Ð458.  Sonnenschein, A. et al., 2015. An image database of Drosophila melanogaster wings for phenomic and biometric analysis. GigaScience, 4(1), p.25.  Stern, D.L. & Emlen, D.J., 1999. The developmental basis for allometry in insects. Development (Cambridge, England), 126(6), pp.1091Ð101.  Stevens, N.M., 1905. Studies in Spermatogenesis with Especial Reference to the ÒAccessory Chromosome,Ó Carnegie Institution of Washington. Stieper, B.C. et al., 2008. Imaginal Discs Regulate Developmental Timing in Drosophila melanogaster. Developmental Biology, 321(1), pp.18Ð26. Stillwell, R.C. et al., 2011. Experimental Manipulation of Body Size to Estimate Morphological Scaling Relationships in Drosophila. Journal of Visualized Experiments, (56), pp.4Ð7. Stillwell, R.C. et al., 2010. Sex Differences in Phenotypic Plasticity Affect Variation in Sexual Size Dimorphism in Insects: From Physiology to Evolution. Annual Review of Entomology, 55, pp.227Ð45. Stillwell, R.C. & Davidowitz, G., 2010a. A Developmental Perspective on the Evolution of Sexual Size Dimorphism of a Moth. Proceedings of the Royal Society B: Biological Science, 277(1701), pp.3819Ð3826. Stillwell, R.C. & Davidowitz, G., 2010b. Sex Differences in Phenotypic Plasticity of a Mechanism that Controls Body Size: Implications for Sexual Size Dimorphism. Proceedings of the Royal Society B: Biological Sciences, (July), pp.3819Ð3826.  Stillwell, R.C., Daws, A. & Davidowitz, G., 2012. The ontogeny of sexual size dimorphism of a moth: when do males and females grow apart? Stocker, H. et al., 2003. Rheb is an essential regulator of S6K in controlling cell growth in Drosophila. Nature cell biology, 5(6), pp.559Ð65.  Sturtevant,  a H., 1945. A gene in Drosophila melanogaster that transforms females into !%,,!males. Genetics, (May), pp.297Ð299. Stutt, A.D. & Siva-Jothy, M.T., 2001. Traumatic insemination and sexual conflict in the bed bug Cimex lectularius. Proceedings of the National Academy of Sciences of the United States of America, 98(10), pp.5683Ð5687. Sun, X. & Artavanis-Tsakonas, S., 1997. Secreted forms of DELTA and SERRATE define antagonists of Notch signaling in Drosophila. Development (Cambridge, England), 124(17), pp.3439Ð3448. Sutter, N. & Bustamante, C., 2007. A Single IGF1 Allele is a Major Determinant of Small Size in Dogs. Science, 316(5821), pp.112Ð115. Suzuki, Y. et al., 1999. A BMP homolog acts as a dose-dependent regulator of body size and male tail patterning in Caenorhabditis elegans. Development (Cambridge, England), 126(2), pp.241Ð250. Sz”kely, T., Freckleton, R.P. & Reynolds, J.D., 2004. Sexual selection explains RenschÕs rule of size dimorphism in shorebirds. Proceedings of the National Academy of Sciences of the United States of America, 101(33), pp.12224Ð7.  Sz”kely, T., Lislevand, T. & Figuerola, J., 2007. Sexual Size Dimorphism in Birds. In Sex, Size & Gender Roles: Evolutionary Studies of Sexual Size Dimorphism. pp. 27Ð37. Takahashi, K.H. & Blanckenhorn, W.U., 2015. Effect of genomic deficiencies on sexual size dimorphism through modification of developmental time in Drosophila melanogaster. Heredity, 115(2), pp.140Ð145.  Takeo, S. et al., 2005. Expression of a secreted form of Dally, a Drosophila glypican, induces overgrowth phenotype by affecting action range of Hedgehog. Developmental Biology, 284(1), pp.204Ð218. Tammaru, T. et al., 2009. Proximate sources of sexual size dimorphism in insects: locating constraints on larval growth schedules. Evolutionary Ecology, 24(1), pp.161Ð175. Tanaka, K. et al., 2011. Evolution of Sex-Specific Traits through Changes in HOX- Dependent doublesex Expression. PLoS Biology, 9(8). Tang, H.Y. et al., 2011. FOXO regulates organ-specific phenotypic plasticity in Drosophila. PLoS genetics, 7(11), p.e1002373.  Tatar, M. et al., 2001. A mutant Drosophila insulin receptor homolog that extends life-span and impairs neuroendocrine function. Science (New York, N.Y.), 292(5514), pp.107Ð10.  Testa, N.D. & Dworkin, I., 2016. The sex-limited effects of mutations in the EGFR and !%,-!TGF-" signaling pathways on shape and size sexual dimorphism and allometry in the Drosophila wing. Development Genes and Evolution, pp.1Ð13.  Testa, N.D., Ghosh, S.M. & Shingleton, A.W., 2013. Sex-Specific Weight Loss Mediates Sexual Size Dimorphism in Drosophila melanogaster. F. Pichaud, ed. PLoS ONE, 8(3), p.e58936.  Teuschl, Y., Reim, C. & Blanckenhorn, W.U., 2007. Correlated responses to artificial body size selection in growth, development, phenotypic plasticity and juvenile viability in yellow dung flies. Journal of Evolutionary Biology, 20(1), pp.87Ð103. Thompson, D.W., 1917. On Growth and Form, Cambridge: Cambridge University Press. Trivers, R.L.L., 1972. Parental investment and sexual selection. Sexual selection and the descent of man, 12(2), pp.136Ð179.  Trotta, V. et al., 2010. Thermal plasticity of wing size and shape in Drosophila melanogaster, D. simulans and their hybrids. Climate Research, 43(1), pp.71Ð79.  Um, S.H. et al., 2004. Absence of S6K1 protects against age- and diet-induced obesity while enhancing insulin sensitivity. Nature, 431(7005), pp.200Ð205. Valenzuela, N. et al., 2004. Geometric Morphometric Sex Estimation for Hatchling Turtles: A Powerful Alternative for Detective Subtle Sexual Shape Dimorphism. Copeia, 2004(4), pp.735Ð742. Verhulst, E.C., Zande, L. Van De & Beukeboom, L.W., 2010. Insect sex determination$: it all evolves around transformer. Current Opinion in Genetics & Development, 20(4), pp.376Ð383.  Waterbury, J. a et al., 2000. Sex determination in the Drosophila germline is dictated by the sexual identity of the surrounding soma. Genetics, 155(4), pp.1741Ð56.  Weatherbee, S.D. et al., 1998. Ultrabithorax regulates genes at several levels of the wing-patterning hierarchy to shape the development of the Drosophila haltere. Genes & Development, 12(10), pp.1474Ð1482.  Weatherhead, P.J. & Robertson, R.J., 1979. Offspring Quality and the Polygyny Threshold: ÒThe Sexy Son Hypothesis.Ó The American Naturalist, 113(2), pp.201Ð208. Webb, T.J. & Freckleton, R.P., 2007. Only half right: species with female-biased sexual size dimorphism consistently break RenschÕs rule. PloS one, 2(9), p.e897.  Weber, K., 2005. Many P-Element Insertions Affect Wing Shape in Drosophila melanogaster. Genetics, 169(3), pp.1461Ð1475.  Weinkove, D. et al., 1999. Regulation of imaginal disc cell size, cell number and organ !%-.!size by Drosophila class I(A) phosphoinositide 3-kinase and its adaptor. Current biology": CB, 9(18), pp.1019Ð29.  Williams, T.M. et al., 2008. The Regulation and Evolution of a Genetic Switch Controlling Sexually Dimorphic Traits in Drosophila. Cell, 134(4), pp.610Ð623. Wilson, E., 1905. The chromosomes in relation to the determination of sex in insects. Science, 22(564), pp.500Ð502.  Wong, D.M. et al., 2014. Insulin- and warts-dependent regulation of tracheal plasticity modulates systemic larval growth during hypoxia in Drosophila melanogaster. PLoS ONE, 9(12), pp.1Ð26. Wullschleger, S., Loewith, R. & Hall, M.N., 2006. TOR signaling in growth and metabolism. Cell, 124(3), pp.471Ð84. Yamanaka, N., Rewitz, K.F. & O Õconnor, M.B., 2013. Ecdysone Control of Developmental Transitions: Lessons from Drosophila Research. Annu. Rev. Entomol, 58, pp.497Ð516. Yan, D. & Perrimon, N., 2015. spenito is required for sex determination in Drosophila melanogaster. Proceedings of the National Academy of Sciences of the United States of America, 112(37), pp.11606Ð11.  Younger-Shepherd, S. et al., 1992. deadpan, an essential pan-neural gene encoding an HLH protein, acts as a denominator in Drosophila sex determination. Cell, 70(6), pp.911Ð22.  Zelditch, M., Fink, W. & Swidersky, D., 1995. Morphometrics, Homology, and Phylogenetics: Quantified Characters as Synapomorphies. Syst. Biol. 44(2):179-189, 199, 44(2), pp.179Ð189. Zhao, W. & Liu, N., 2014. The Proximate Causes of Sexual Size Dimorphism in Phrynocephalus przewalskii. , 9(1), pp.1Ð9. !Zimmerman, E., Palsson, A. & Gibson, G. 2000 Quantitative trait loci affecting components of wing shape in Drosophila melanogaster. Genetics 155, 671Ð83.