THE CAUSES AND EFFECTS OF GENITAL HYPOALLOMETRY IN DROSOPHILA
By
Austin P. Dreyer

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Zoology – Doctor of Philosophy
Ecology, Evolutionary Biology and Behavior – Dual Major
2014

ABSTRACT
THE CAUSES AND EFFECTS OF GENITAL HYPOALLOMETRY IN DROSOPHILA
By
Austin P. Dreyer
The study of size covariation between traits has a long history of describing
morphological variation. For over a century, scientists have recognized variation in the
proportional size of traits, and have searched to explain the patterns from an
evolutionary perspective. Research on the scaling relationships between traits, called
allometries, has established the interaction of traits within an organism plays a crucial
role in the adaptation of species to their environment. The evolutionary forces that give
rise to changes in the proportional size of traits have been more difficult to elucidate.
Using the model organism, Drosophila melanogaster, I have focused my
research on the scaling of male genitalia in relation to overall body size to explore
proximate and ultimate causes of allometries in general. Most traits scale at or near a
1:1 ratio to overall body size, called isometry. In contrast, the male genitalia of many
groups scale hypoallometrically to body size, remaining a constant size across a range
of body sizes. Determining the factors that drive the atypical allometric relationship of
the male genitalia promises to reveal principles of size control across all traits.
To investigate the developmental mechanisms underlying genital hypoallometry,
I first compared the effects of genetic variation on genital traits (hypoallometric) to
somatic traits (isometric). Previous research has shown that genital traits are less
sensitive to environmental variation than somatic traits and here I demonstrate that
genital traits are also less sensitive to variation in genetic factors that affect trait size. I

also showed that genitalia have low levels of developmental stability than somatic traits,
measured as the response in trait size to stochastic developmental errors. Next, I used
targeted gene expression of insulin-signaling genes in developing genital tissues of
Drosophila to allometrically engineer male flies with extreme genital sizes. Females
were exposed to males with different genital sizes, and demonstrated a preference for
copulating, and fertilizing progeny, with males that had larger genitalia. To expand the
scope of these results, a stochastic mathematical model of allometry evolution was
designed that incorporated the developmental regulation of size. Results of simulated
allometry evolution showed that the underlying factors controlling final trait size largely
determine how scaling relationships respond to selection and evolve.
Collectively, my dissertation represents a significant step forward in our
understanding of trait size regulation between covarying traits. Additionally, my research
demonstrates the novel use of Drosophila melanogaster to modify existing levels of trait
variation to test selection hypotheses. Scaling relationships between traits are an
important component of morphological evolution that we can continue learning about
only via multifaceted research as demonstrated here.

ACKNOWLEDEMENTS
After just over six years at Michigan State University working on my dissertation,
I feel qualified to say several things about what it takes to complete a PhD in the life
sciences. First, a good research program is essential. Second, perseverance in the face
of constant failure, or at the very least, things constantly failing to go as planned. And
third, surrounding yourself with good people. It is the third point that I wish to elaborate
on in this section.

The journey to my PhD really started during my undergraduate career while doing
research on subterranean insects at Hope College. For the first time in my life it was
suggested to me by my research advisor, Thomas Bultman, that it might be possible for
me to have a PhD someday. Ever since that little seedling of an idea was planted in my
mind, I have worked toward making a PhD happen. Other professors at Hope College
supported my aspirations and I thank K. Greg Murray and Kathy Winnet-Murray for
reminding me that I am capable. Dr. Bultman provided me with so much support after
introducing me to the concept of a PhD, including taking me to Switzerland in the
summer of 2008 for some field research, and I will forever be in his debt for stoking the
fire of research in me. Two other professors at Hope College had a very formative
impact on my research, William Mungle and Wally Fu. Both forced me to stay the
course over two semesters of challenging organic chemistry labs, and I have often
measured my ability for hard work based on those two semesters.

iv

Once I decided to attend Michigan State University, I was introduced to a new group of
academic colleagues with whom I would develop as a scientist. To my wonderful
committee members, thank you for your constant support during my time at MSU: Ian
Dworkin, Jenny Boughman, Jim Miller and Mark Scriber. My first real contact with MSU
was through my PhD mentor, Alex Shingleton, and I will always remember those initial
meetings as I was introduced to the concept of proportional trait size and allometry. I will
never meet a more intelligent, more dedicated, more unwilling to accept good-as-goodenough person, and I am so thankful for the tremendous impact he has had on me as a
scientist and as a human being. Lastly, to Sue and Dick Hill, for their demonstration of
how to be an educator in every sense of the word and how to balance work with life
outside of the lab, thank you so much.

The friends I have made while at MSU are the kind born out of a crucible. Ever since the
first introduction at the Biological Sciences TA retreat in the fall of 2008, Eli Swanson
and Carlos Anderson have been brothers to me and I could not have made it half as far
without them. Nick Testa joined the Shingleton Lab in the fall of 2010, and we quickly
embarked on years of metal music, video games, shared research woes and a
commitment to completing what we started. His help was invaluable throughout my
dissertation. Aaron Florn and I first crossed paths while he was an undergraduate in a
course I was a TA for, Comparative Vertebrate Anatomy. Since that time, he joined the
Zoology department at MSU as a Masters student, and as they say, the rest is history.
The importance of our friendship to the completion of my degree can easily be summed

v

up by the unsolicited comments of people in the department identifying us as ‘joined at
the hip.’ A heartfelt thank you to Chelsea Van Eck, David Green, Rachael Eaton, and
Bene Bachelot for their support as well, through thick and thin. And to Nora Lewin,
thank you for everything.

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 years of missed holidays, vacations and last minute changes of plan. In
particular, thank you to my Uncle Dean and Aunt Connie, and a gigundous thank you to
my brother Jamin, who was also completing a PhD for much of my time at MSU. Jamin
was always willing and able to commiserate with, to talk things through, and to offer an
ear as someone who could understand almost exactly what I was going through at any
given time. And to my parents Jim and Cathy, none of this would have happened
without you. Thank you for your love and unwavering support.

vi

TABLE OF CONTENTS
LIST OF TABLES ............................................................................................................ ix
LIST OF FIGURES ........................................................................................................... x
CHAPTER 1: General Introduction..................................................................................1
Introduction...................................................................................................................2
Background ...........................................................................................................2
Definitions of Allometry ..........................................................................................3
Variation in Allometry.............................................................................................4
Summary of Dissertation ..............................................................................................6
Summary of Chapter 2 .................................................................................................7
Summary of Chapter 3 .................................................................................................8
Summary of Chapter 4 .................................................................................................8
General Methods ...................................................................................................9
Study Subjects.......................................................................................................9
Morphological Data Collection .............................................................................11
Writing Style of This Dissertation ...............................................................................12
CHAPTER 2: The effect of genetic and environmental variation on genital size in male
Drosophila: canalized but developmentally unstable .....................................................13
Introduction.................................................................................................................14
Materials and Methods ...............................................................................................17
Fly Stocks ............................................................................................................17
Fly Rearing ..........................................................................................................18
Genetic Variation .................................................................................................18
Developmental Instability.....................................................................................18
Morphology ..........................................................................................................18
Analysis ......................................................................................................................19
Genetic Variation .................................................................................................19
Genetic Static Allometry ......................................................................................20
Developmental Instability.....................................................................................21
Results .......................................................................................................................23
Discussion ..................................................................................................................26
CHAPTER 3: Female preference and the hypoallometry of male genitalia in
Drosophila ......................................................................................................................32
Introduction.................................................................................................................33
Methods......................................................................................................................40
Allometric Engineering.........................................................................................40
Fly Assays ...........................................................................................................43
Morphology ..........................................................................................................46
Analysis ......................................................................................................................47
Static Allometry....................................................................................................47

vii	
  

Mating Assay – Behavioral Traits ........................................................................47
Mating Assay – Female Preference ....................................................................48
Mating Assay – Reproductive Success ...............................................................48
Developmental Stability .......................................................................................49
Results .......................................................................................................................50
Allometric Engineering.........................................................................................50
Static Allometry....................................................................................................52
Mating Assay – Behavioral Traits ........................................................................52
Mating Assay – Female Preference ....................................................................52
Developmental Stability .......................................................................................58
Discussion ..................................................................................................................62
APPENDIX .................................................................................................................68
CHAPTER 4: Trait development and the evolution of allometries in silico ....................72
Introduction.................................................................................................................73
Methods......................................................................................................................79
Developmental Basis ...........................................................................................79
Model ...................................................................................................................79
IndividualSA vs PopulationSA .............................................................................82
Selection ..............................................................................................................82
Population Structure ............................................................................................86
Results .......................................................................................................................88
Discussion ..................................................................................................................96
CHAPTER 5: Conclusion and future directions ...........................................................101
Summary of Chapters ..............................................................................................102
Chapter 1..................................................................................................................102
Chapter 2..................................................................................................................102
Chapter 3..................................................................................................................103
Chapter 4..................................................................................................................104
Conclusion................................................................................................................105
Genital Hypoallometry in Drosophila .................................................................105
Future Directions ...............................................................................................106
REFERENCES .............................................................................................................108

viii	
  

LIST OF TABLES
Table A1. Pairwise comparisions by genotype of mating behaviors. No significant
difference between genotypes for early stages of courtship. Significant differences in
behaviors are indicated by (*). Direct competition regime was not analyzed because
males could not be differentiated in video recording of mating trial. All p-values
calculated by Tukey’s HSD tests
71
Table A2. No relationship between the number of progeny sired and genital size. Linear
regression of total number of offspring compared to genital size for each mating regime.
No correlation or significant relationship between genital size and reproductive success
across all males that sired offspring
71

ix

LIST OF FIGURES
Figure 2.1. Genetic variation, allometric coefficient and fluctuating asymmetry of
somatic and genital traits in male Drosophila melanogaster. (A) Genital traits had low
levels of total genetic variation (light gray bars, CVT) but not low levels of organautonomous genetic variation (dark gray bars, CVI). The difference between total
genetic variation and organ-autonomous variance is an estimate of genetic variation
that is correlated with variation in other traits (‘systemic’ genetic variation). Columns with
the same letter are not significantly different for total genetic variation (CVT) using
Tukey’s HSD (P > 0.05). Traits do not differ for organ-autonomous genetic variation
(CVI) using Tukey’s HSD (P > 0.05 for all) (B) The low systemic genetic variance of the
genital traits reflected their low multivariate allometric coefficient compared to most
somatic traits, although these differences are not significant for multiple comparisons
(Tukey’s HSD, P>0.05 for all). Grey horizontal line is isometry. (C) Genital traits did not
show low levels of fluctuating asymmetry. Light grey bars, line 303, white bars, line 324,
dark grey bars, line 335. Within a line, columns with the same letter are not significantly
different for FA using Tukey’s HSD (P 0.05). All error bars are 95% confidence
intervals
24
Figure 2.2 Model of a selection regime that alters the slope of the genital-body scaling
relationship while maintaining genital-autonomous genetic variation. Selection is for
proportionally smaller genitalia in large males and proportionally larger genitalia in small
males (black arrows). Implicit to such a regime is that there is selection or constraint
maintaining variation in body size (gray arrows)
25
Figure 3.1. Elav-GAL80 expressed with Poxn rescues wild-type mating behavior.
Courtship latency of IIS manipulated flies (Poxn > dFOXO.3x) had significantly longer
courtship latency than control male flies (Poxn > UAS-GFP). Coexpression of an ElavGAL80 that prevented expression of GAL4 in neuronal tissues recovered wild-type
courtship latency (Elav-GAL80:Poxn > FOXO.3x). The symbol (a) denotes significance
(Tukey’s HSD p < 0.0001)
42
Figure 3.2. IIS perturbation using Poxn changes size of the genitalia only. Size of the
posterior lobe of the genital arch was significantly smaller than the large or wild-type
genital size classes (Tukey HSD < 0.001, represented by symbol ‘a’). There were no
significant differences in size of the wings or puparial between the genital size classes
(Tukey HSD > 0.05). Data for genital size classes were pooled by the IIS transgene
present in the fly (large: InR.CA, wild-type: UAS-GFP, small: dFOXO.3x) because there
was no difference in size between flies with or without the Ubi-GFP (Tukey HSD > 0.05).
Error bars represent SEM
51
Figure 3.3. Males with larger genitalia are more likely to mate in single male regime.
Logistic regression of male mating success (probability = 1.0) or failure (probability = 0)
by log-transformed genital size of males. Females were more likely to mate with males
that had larger genitalia (p = 0.009). Males with small genitalia were able to successfully

x

mate with females, but did so at a lower frequency than males with larger genital sizes.
Histogram bars in grey (top) show frequency of successfully copulation and black bars
(bottom) show frequency of failure to copulate by genital size. Line represents logistic
regression probability curve
53
Figure 3.4. Males with larger genitalia are more likely to sire offspring in single mating
regime. Logistic regression of a female fertilizing (probability = 1.0) or not fertilizing
(probability = 0) eggs by log-transformed genital size. Females were more likely to
fertilize eggs using sperm from males with larger genitalia (p = 0.001). Females fertilized
eggs using sperm form males with small genitalia, but at a lower frequency than sperm
from males with larger genital sizes. Histogram bars in grey (top) show frequency of
successful fertilization and black bars (bottom) show frequency of failure to fertilize by
genital size. Line represents logistic regression probability curve
54
Figure 3.5. Males with larger genitalia were more likely to mate in the direct competition
regime. Logistic regression of a male mating success (probability = 1.0) or failure
(probability = 0) by log-transformed genital siz. Females were more likely to mate with
males that had larger genitalia (p = 0.002). Males with small genitalia were able to
successfully mate with females, but did so at a much lower frequency than males with
larger genital sizes. Histogram bars in grey (top) show frequency of successful
copulation and black bars (bottom) show frequency of failure to copulate by genital size.
Line represents logistic regression probability curve
56
Figure 3.6. Males with larger genitalia were more likely to sire progeny in the indirect
competition regime. Logistic regression of female fertilizing (probability = 1.0) or not
fertilizing (probability = 0) eggs by log-transformed absolute genital size. There was a
significant positive relationship between genital size and fertilization probability (p =
0.031). Sperm from males with small genitalia was used to fertilize eggs, but at a much
lower frequency than sperm from males with larger genital sizes. Histogram bars in gray
(top) show frequency of fertilization and black bars (bottom) show frequency of failure to
fertilize by genital size. Line represents logistic regression probability curve
57
Figure 3.7. Fluctuating asymmetry (FA) is higher in genital traits than the wings in the
single male regime. There was no difference in FA values by trait across the genital size
classes, but there was a difference between the traits. FA was calculated using the
FA10b value which accounts for measurement error in log-transformed traits. Error bars
are 95% confidence intervals
59
Figure 3.8. No difference in FA between flies that mated and did not mate in the direct
competition regime. Error bars are 95% confidence intervals
60
Figure 3.9. No difference in FA between flies that sired progeny and those that did not
in the indirect competition regime. Error bars are 95% confidence intervals
61

xi

Figure 3.10. Response in genitalia and wings to IIS manipulation to make large and
small traits. Wing significantly increased and decreased in size in response to upregulating (large: NP6333>InR.CA) and down-regulating (small: NP6333>InR.DN25) IIS
in those tissues (Tukey’s HSD < 0.05). Posterior lobe of the genital arch only decreased
in size in response to down-regulating IIS (small: Poxn>dFOXO.3x) but did not increase
in size when IIS was up-regulated (large: Poxn>InR.CA) suggesting the genitalia are
insensitive to positive nutritional growth regulators or are at a maximum size in
response to nutrition. Error bars are 95% confidence intervals. Control flies were > UASGFP for both traits
65
Figure A1. Complete mating assay consisting of 18 – 35mm x 10mm mating chambers
with mating pairs visible in each chamber
69
Figure A2. There was a difference in allometric coefficient between, but not within,
traits. The allometric coefficient was lower in the genitalia thatn the wing, as expected.
he allometric coefficients do not vary across size class, but are different between traits.
Error bars are 95% bootstrap confidence intervals from 1000 replicates
70
Figure 4.1. Population static allometry and the individual static allometry. (A) A
hypothetical populationSA demonstrating the relationship between covarying traits T1
and T2 comprised of 25 individuals, represented by unique colors. (B) Individuals
express only one phenotype (colored circle) of the possible individualSA possible for
their genotype (colored lines) based on the environmental conditions they are exposed
to during development
78
Figure 4.2. Patterns of individualSA. (A, A’) When µG = 0, the populationSA appears
normally distributed (A) and the individualSA have equal levels of variation as the size
of T1 increases or decreases (A’) creating a bowtie shaped pattern. (B, B’) When µG >
0, the populationSA appears similar in distribution to populations with µG = 0, including
similar slope and covariation between traits T1 and T2 (B), but the underlying individuallevel static allometries exhibit more variation as the size of T1 increases (B’) producing
a speedometer shaped pattern. (C, C’) When µG < 0, the population-level static
allometry has the same appearance and statistical properties as populations with µG > 0
or µG = 0, but the individual-level static allometries exhibit more variation as the size of
T1 decreases (C’) creating a broomstick shaped pattern
84
Figure 4.3. Four selection regimes: Stabilizing selection (A), Directional selection (B),
Proportional selection (C) and Correlational selection (D). Grey points indicate selected
individuals red points indicate individuals removed from the population prior to
reproduction. The positive form of selection is shown in B-D with negative selection for
those types removing individuals from the opposite orientation in the population
87
Figure 4.4. Evolution of populationSA for ten generations of selection on bowtie
populations (µG = 0). Inset plots display the change in the allometric coefficient (α)
through time (generation). (A) The individualSA pattern when µG = 0. Generations 0 –

xii

10 represented by colored lines blue – red. The effect on the populationSA from
stabilizing selection (B), positive and negative directional selection (C,C’), positive and
negative proportional selection (D,D’) and positive and negative correlational selection
(E,E’)
90
Figure 4.5. Evolution of populationSA for ten generations of selection on speedometer
populations (µG > 0). Inset plots display the change in the allometric coefficient (α)
through time (generation). (A) The individualSA pattern when µG > 0. Generations 0 –
10 represented by colored lines blue – red. The effect on the populationSA from
stabilizing selection (B), positive and negative directional selection (C,C’), positive and
negative proportional selection (D,D’) and positive and negative correlational selection
(E,E’)
91
Figure 4.6. Evolution of populationSA for ten generations of selection on broomstick
populations (µG < 0). Inset plots display the change in the allometric coefficient (α)
through time (generation). (A) The individualSA pattern when µG < 0. Generations 0 –
10 represented by colored lines blue – red. The effect on the populationSA from
stabilizing selection (B), positive and negative directional selection (C,C’), positive and
negative proportional selection (D,D’) and positive and negative correlational selection
(E,E’)
92
Figure 4.7. Initial (blue) and final (red) bowtie populations (µG = 0) after 10 generations
of selection. All lines are of the corresponding colored populationSA calculated using
major axis regression. Represented are stabilizing selection (B) positive and negative
directional selection (C, C’) positive and negative proportional selection (D, D’), positive
and negative correlational selection (E, E’)
93
Figure 4.8. Initial (blue) and final (red) speedometer populations (µG > 0) after 10
generations of selection. All lines are of the corresponding colored populationSA
calculated using major axis regression. Represented are stabilizing selection (B)
positive and negative directional selection (C, C’) positive and negative proportional
selection (D, D’), positive and negative correlational selection (E, E’)
94
Figure 4.9. Initial (blue) and final (red) broomstick populations (µG < 0) after 10
generations of selection. All lines are of the corresponding colored populationSA
calculated using major axis regression. Represented are stabilizing selection (B)
positive and negative directional selection (C, C’) positive and negative proportional
selection (D, D’), positive and negative correlational selection (E, E’)
95

xiii

CHAPTER 1:
General Introduction

1	
  

Introduction
Background
Understanding why organisms are the size and shape that they are has been a
focus of research for well over a century. How populations adapt to their environment
through morphological diversification is of great interest to evolutionary biologists, driven
by the visually apparent variation in body plan across taxonomic groups (Peters 1983,
Wagner and Lynch 2010). Some of the most common classes of adaptations by
organisms to their environment are phenotypic modification, and modifications to the
degree of integration between traits within an organism (Schmidt-Nielsen 1984, Brown
and West 2000, Klingenberg and Marugan-Lobon 2013). Much remains to be
discovered about how integrated traits in a population respond to selection. Specifically,
how is selection affecting change in the phenotypes of covarying traits in a population?
To what degree does the development of tissues constrain or facilitate response to
selection? How can we elucidate the selection pressures acting on populations and
generate predictions for the evolutionary outcomes from different forms of selection?
Research on the evolution of morphology spans many disciplines and includes
behavioral, developmental, genetic, and physiological research. The goal of this
dissertation is to use a multifaceted approach to test questions of morphological
evolution, first through empirical research on a specific trait size/body size relationship
and more broadly through theoretical work that uses a mathematical model to explore
how selection can change the relationships between traits within an organism.

The evolution of size and shape springs from how traits are related to one another
developmentally and functionally (Calder 1984). As organisms adapt to their

2	
  

environments, the maintenance of trait functionality is a potentially powerful mechanism
for their survival. The balance between maintaining utility and adapting to new
conditions often hinges on the relative size of traits to one another (Pelabon et al. 2014).
For example, the size of a flying insect’s wing needs to be large enough to provide
sufficient lift for a given body size to fly, but not so large as to be energetically costly to
generate and maintain. An observable pattern throughout natural systems is that much
of the morphological diversity across taxa can be distilled to variation in the proportional
size of the same or similar traits (Thompson 1942, Gould 1966, Emlen and Allen 2003).

Definitions of Allometry
How traits are developmentally connected to one another can be described in several
contexts. Julian Huxley (1932) first used the term allometry to refer to any trait that
deviated from maintaining a 1:1 consistent size ratio with body size. Presently, allometry
is often used to refer to the study of any scaling relationships between two traits (Gould
1966, Pelabon et al.). Implicit to the concept of allometry is that there is covariation
between the two traits being studied, indicative of a biologically meaningful link such as
shared developmental control mechanisms or a behavioral integration of traits that must
work in conjuction with one another for survival and reproduction.

There are three categories of allometries that describe different aspects of trait
covariation: evolutionary, ontogentic and static allometries (Cock 1966, Gould 1966,
Cheverud 1982). An evolutionary allometry compares the scaling relationships between
traits across related taxanomic groups at the same developmental stage (typically

3	
  

species or populations). An ontogenetic, or growth, allometry compares individuals of
the same species across developmental stages. The third type of allometry is a static
allometry and compares the size of traits in individuals of the same species all at the
same developmental stage. Static allometries most often compare a trait to overall body
size of an organism but can be applied to any co-varying traits within an organism. The
source of size variation further distinguishes the classification (Shingleton et al. 2007).
For traits that vary due to environmental conditions, such as nutrition levels or
temperature, the comparison is referred to as an environmental static allometry.
Phenotypic variation that is due to genetic variation in organisms that have all been
exposed to the same environmental conditions are called genetic static allometries.

Variation in Allometry
Allometric relationships are often described mathematically using the allometric
α

equation y = bx which relates two traits, x and y (Huxley 1932). The allometric equation
is often log-transformed, making the relationship a linear comparison between traits and
scale-independent for comparison across trait scales (Kerkhoff and Enquist 2009). As a
result of the log-transformation, the allometric equation becomes linear as log(y) = α *
log(x) + log(b), where the slope is defined by α, the allometric coefficient, and the
allometric intercept is log(b). Much of the morphological variation between pairs or sets
of traits can be encapsulated by changes in these two parameters (Gould 1966). For
instance, traits with an allometric coefficient (α) of 1 are classified as isometric. Any
increase in the size of one trait results in a proportional change in the associated trait.
Traits with α > 1 are classified as hyperallometric and increase disproportionately in size

4	
  

with their associated trait. Traits with α < 1 are classified as hypoallometric and remain
more or less the same size across a range of their associated trait sizes. The intercept
term describes the relative size of a trait that is independent of the associated trait. An
increase or decrease in only the intercept of the allometric equation results in relatively
larger or smaller traits across a range of the covarying trait, respectively (Shingleton
2012).

The causes of deviation from the norm of isometry are intimately tied to the general
principles of growth control. Proximate causes for different allometric relationships
involve a modification of the mechanisms that link environmental and genetic size
regulators with individual trait growth (Shingleton et al. 2005, 2007). For example,
hyperallometric traits could be the result of hyper-sensitivity to growth cues in response
to environmental inputs (such as nutrition levels) resulting in disproportionate growth as
compared to traits that have similar sensitivities to growth cues as the body in general
(Emlen et al. 2012). Hypoallometric traits, on the other hand, are less sensitive to
growth factors resulting in a more constant size in the face of environmental variation
(Tang et al. 2011). The ultimate causes for different scaling relationships lie in the types
of selection that can give rise to allometries, although the effects of specific types of
selection continue to be a question with no clear answer (Bonduriansky and Day 2003,
Kodric-Brown et al. 2006, Eberhard et al. 2009, Cayetano et al. 2011, Arendt and
Fairbairn 2012, Berger et al. 2012, Fromhage and Kokko 2014, Pelabon et al. 2014).
Theoretical predictions suggest that the same scaling relationships can be the possible
outcomes of different forms of selection, and that different forms of selection can result

5	
  

in the same scaling relationships. While this very well may be the case, the field of
allometric research is in need of an updated framework that synthesizes our
understanding of trait development with the evolution of proportional trait size to
continue moving forward.
Summary of Dissertation
My doctoral dissertation research focuses on how scaling relationships evolve
and the integral role they play in developing morphological diversity. My dissertation
consists of four chapters following this introductory chapter. The first research chapter,
Chapter 2, describes the phenomenon of genital hypoallometry within male Drososphila
melanogaster and suggests possible causes for the atypical scaling relationship
observed in male genitalia across arthropods. Chapter 3 is a test for the type of
selection acting on male genitalia in D. melanogaster using the genetic tools available to
manipulate genital size beyond what would be found in natural populations. Chapter 4
outlines a mathematical model developed to test the effects of different forms of
selection on populations and generate predictions of how allometries can evolve.
Chapter 5 is a conclusion chapter that comments on the causes of genital
hypoallometry in Drosophila.

Collectively, my dissertation presents a novel approach to the ongoing question of how
allometries evolve and attempts to test theories using a known biological example of
hypoallometry. My dissertation is among the first to use the GAL4-UAS system in
Drosophila to test questions about selection on trait size by artificially modifying
phenotypic variation without physical manipulation. In total, my dissertation represents a

6	
  

significant step forward in understanding how phenotypic traits have evolved into the
sizes observed today, and how they will continue to evolve in the future.

Summary of Chapter 2
The genitalia of most male arthropods scale hypoallometrically with body size,
that is, they are typically the same size among large and small individuals in a
population. Such scaling is expected to arise when genital traits show less variation
than somatic traits in response to factors that generate size variation among individuals
in a population. Nevertheless, there have been few studies directly examining the
relative sensitivity of genital and somatic traits to factors that affect their size. Such
studies are key to understanding genital evolution, and more generally, the evolution of
morphological scaling relationships. Previous studies indicate that the size of genital
traits in male D. melanogaster show a relatively low response to variation in
environmental factors that affect trait size. In this chapter the size of genital traits in
male fruit flies is shown to also exhibit a relatively low response to variation in genetic
factors that affect trait size. Importantly, however, this low response is only to genetic
factors that affect body and organ size systemically, not those that affect organ size
autonomously. Furthermore, it is shown that the genital traits do not exhibit low levels of
developmental instability, i.e. the response to stochastic developmental errors that also
influence organ size autonomously. These results are discussed in the context of
current hypotheses on the proximate and ultimate mechanisms that generate genital
hypoallometry.

7	
  

Summary of Chapter 3
Across a large portion of the animal kingdom, male genital size remains near
constant among individuals in a population, despite considerable phenotypic variation in
body size and the size of other organs. Several competing hypotheses have been
proposed to explain the selective pressures underlying this commonly observed
phenomenon. However, a lack of experimental data makes it difficult to distinguish
between the different hypotheses. For any test of how selection is acting using a natural
population, morphological variation that exceeds the natural range of size is required.
To satisfy this requirement, “allometric engineering” in D. melanogaster was used to
modify a nutrient-sensing pathway in specific tissues and produce males with extreme
genital morphologies. Males of different genital sizes were then used to assay the
effects of genital size on various measures of reproductive success. The results of
these experiments indicate females prefer males with larger genital sizes as compared
to smaller genitalia for copulation and fertilization but are able to copulate with males of
any genital size. Females choose to fertilize eggs at the same frequency regardless of
the genital size of the male from which the sperm came. The ability to engineer male
flies with smaller but not larger genitalia, despite a demonstration of increase and
decrease in wing size due to perturbation of the same nutrient-sensing signals,
suggests male genitalia may be less sensitive to signals of increasing size or at a
maximum size in response to nutrition-based growth regulators.
Summary of Chapter 4
How scaling relationships evolve is an ongoing question in the field of
morphological evolution. Theory predicts that the effects of selection on scaling

8	
  

relationships is difficult to anticipate, and empirical work has done little to resolve the
question while focusing on exaggerated traits typical of secondary sexual
characteristics. In chapter 4 the understanding of development in holometabolous
insects is used to construct a stochastic mathematical model that simulates populations
of individuals with a set of hypothetical covarying traits that are free to evolve in
response to different types of selection. The developmental parameters included in the
model represent general classes of growth factors that are known to control trait
development. The results of simulations show that the underlying growth factors have a
profound effect on the predicted response to selection. Based on which mechanisms
are controlling size, populations can have different possible responses to selection on
trait size. It is possible for the same scaling relationships to be the result of multiple
selection regimes. Conversely, it is also possible for the same type of selection to result
in multiple scaling relationships based entirely on how the individuals in a population are
able to respond to selection. Additionally, the results suggest scaling relationships that
are ostensibly identical can respond to the same selection pressure in different ways.
The underlying developmental control mechanisms that are responsible for regulating
trait size, and the covariation between integrated traits, is a primary factor in the
determination of how scaling relationships evolve.
General Methods
Study Subjects
Drosophila melanogaster are a well-recognized model system for a variety of
biological applications. They are holometabolous insects, defined as progressing
through three distinct life stages: larva, pupa and adult (Klowden 2007). D.

9	
  

melanogaster are often reared under standard laboratory conditions, accepted to be
25°C and either 24 hr constant light exposure or 12 hr : 12 hr light : dark cycles (Stocker
and Gallant 2008). At 25°C, the D. melanogaster life cycle takes nine days from egg to
sexually mature adult making them highly amenable to research that requires large
numbers of individuals (Stocker and Gallant 2008). A standard cornmeal/molasses fly
food was made for all flies at Michigan State University as needed.

One of the most powerful genetic tools available to D. melanogaster is the GAL4/UAS
system, which allows controlled expression of transgenes in flies (Duffy 2002).
GAL4/UAS is a transcriptional regulation system isolated from the yeast
Saccharomyces cerevisiae. GAL4 encodes a protein that binds to the Upstream
Activating Sequence (UAS) and initiates transcription of a gene associated with the
UAS. GAL4 is expressed in a tissue or temporal specific manner, and when combined
with a UAS transgene construct in a single fly, allows for targeted expression of a gene
that is localized spatially and/or temporally. There are a number of GAL4 and UAS
transgenes that are publicly available, having been created and screened for expression
through the use of a P-transposase promoter. An additional level of expression control
is possible through the use of the protein GAL80 that is also expressed in a tissue
specific manner and competitively binds to GAL4 to prevent any transcriptional activity
(Elliot and Brand 2008). While many more genetic tools are possible using D.
melanogaster, the GAL4/UAS system and the GAL80 protein were the primary
techniques used in this dissertation.

10	
  

Fly strains used in laboratory experiments were based on the standard isogenic lab
strain Samarkand (Sam) or on wild caught lines from Raleigh, NC (Ral). Flies were
obtained from the Bloomington Drosophila Stock Center (BDSC) at Indiana University
unless otherwise noted. Controlling for the possible confounding effects of different
genetic backgrounds on the expression of a gene is a crucial step in any study
manipulating gene expression (Chari and Dworkin 2013). To minimize the effect of
genetic background, all transgenes were introgressed into the Sam line for a minimum
of five generations. The introgression of transgenes into a common genetic background
is designed such that a gene of interest is repeatedly inserted into a genetically
homogenous line. After a sufficient number of generations, only the transgene that is
traceable through genetic markers should be present in the otherwise homogenous
background.
Morphological Data Collection
All morphological trait measurements were dissected using micro forceps in 75%
ethanol. Flies used in any behavioral assays were reared socially isolated as individuals
in 2 mL microfuge tubes. Because all growth is fixed during the pre-adult stages in
holometabolous inects, puparial case size is tightly correlated to adult body size (Chiang
and Hodson 1950). Therefore, puparial case area was used as an estimate of body size
whenever possible. To trace puparial cases to their respective adult fly, flies were
allowed to eclose in marked 2 mL microfuge tubes and the puparial cases were saved
for measurement.

11	
  

Writing Style of This Dissertation
My dissertation is the direct result of collaboration with others who have
contributed critical pieces to the work. I could not have completed this body of work
without their direct input and guidance on my research. In recognition of their efforts, I
will use the first person plural and not the first person singular for the remainder of the
dissertation.

12	
  

CHAPTER 2:
Dreyer, AD and Shingleton, AW. 2011. The effect of genetic and environmental
variation on genital size in male Drosophila: canalized but developmentally unstable.
PloS One Vol. 6(12): pp.e28278

13	
  

Introduction
Within a population or species, variation in body size is expected to be
accompanied by approximately equivalent variation in the size of individual
morphological traits. Such covariation is necessary to maintain correct body proportion
across the range of body sizes observed in animal populations. A notable exception to
this pattern, however, is the relationship between genital size and body size in
arthropods. The genitalia of most arthropods are more-or-less the same size in both
large and small individuals (Eberhard et al. 1998, Bernstein and Bernstein 2002, Funke
and Huber 2005, Hosken et al. 2005, Kawano 2006). Consequently, smaller males have
proportionally larger genitalia than larger males. While the phenomenon is most obvious
in males, it has also been observed in female arthropods (Eberhard et al. 1998,
Palestrini et al. 2000, Uhl and Vollrath 2000, Tatsuta et al. 2007), as well as some
mammals (Patterson 1982, Oosthuizen and Miller 2000, Eberhard 2009).

The scaling relationship between two traits among individuals of the same
developmental stage in a population is called a static allometry, and is typically
described using the allometric equation, y = bxα , where x and y are the size of two traits
(Huxley and Teissier 1936). Log transformation of this equation produces the simple
linear equation log( y) = α log( x)€+ log(b) , and log-log plots of the size of different traits
among individuals in a population typically reveal linear scaling with a slope of α, called
€
the allometric
coefficient (Huxley and Teissier 1936). When α = 1, the relationship

between x and y is called isometry, with the ratio of y to x remaining constant across a

€ or hyperallometric, respectively,
range of x. When α < 1 or > 1 the relationship is hypo-

€

€

14	
  

with relative size of y decreasing (hypoallometry) or increasing (hyperallometry) with an
increase in x. Fundamental to the concept of allometry is that x and y covary; that is the
factors that generate variation in x also generate variation in y. The allometric coefficient
therefore captures the extent to which these factors affect y relative to x. If a factor that
generates size variation affects both traits equally, then y will scale isometrically to x
(assuming that all size variation is due to the factor). If the factor has a lesser or greater
effect on y than x, y will scale with x hypo- or hyperallometrically respectively
(Shingleton et al. 2007).

The observed hypoallometry of the male genitalia in arthropods suggests that genital
traits are relatively insensitive to the factors that generate size variation among
individuals in a population. Size variation may be generated by environmental variation
(plasticity), genetic variation and developmental instability (variation due to stochastic
developmental perturbations within an individual, (Van Valen 1962, Palmer and
Strobeck 1986, Nijhout 2003). Consequently, we might expect the genitalia to be
environmentally canalized, genetically canalized and/or developmentally stable. Here,
we define canalization as the property of a trait to resist genetic or environmental
variation (Waddington 1961, Stearns et al. 1995, Debat 2001), and developmental
stability as the property of a trait to resist stochastic developmental perturbations that
generate fluctuating asymmetry (FA) in a bilaterally symmetrical organism (Parsons
1990, Debat 2001). Previous studies have demonstrated that the male genitalia of
Drosophila are environmentally canalized, at least with respect to developmental

15	
  

nutrition, temperature and larval crowding (Shingleton et al. 2009). It is unclear,
however, whether they are also genetically canalized and developmentally stable.

In contrast to our lack of understanding of the developmental mechanisms that underlie
genital hypoallometry, there are a number of hypotheses as to its adaptive significance
(Arnqvist 1997, Bonduriansky and Day 2003, Hosken and Stockley 2004). A general
theme of many of these hypotheses is that there is stabilizing selection on male genital
size, either because females are physically unable to mate with males bearing
inappropriately-sized genitalia (Dufour 1936) or because females prefer males with
genitalia of a specific size (Eberhard et al. 1998). Alternatively, hypoallometry may arise
because there is directional selection on increased genital size that is strong in small
males but weak or absent in large males (Eberhard et al. 2009). These different
hypotheses, while not mutually exclusive, serve to emphasize the observation that the
form of selection on genitalia can be difficult to infer from patterns of allometry (Bertin
and Fairbairn 2007).

Elucidating the proximate mechanisms that generate genital hypoallometry may help
clarify the ultimate evolutionary processes that cause it. This is because different
evolutionary hypotheses suggest different patterns of genetic and environmental
variation in genital size. For example, if genital hypoallometry were a consequence of
elevated levels of stabilizing selection on genital size, we would expect to see a
reduction in the level of genetic variation in genital size relative to other traits: that is
they should be genetically canalized (Falconer and Mackay 1996, Eberhard 2009). We

16	
  

might also expect the genitalia to be environmentally canalized and developmentally
stable (Scharloo et al. 1967, Kaufman et al. 1977, Gavrilets and Hastings 1994,
Falconer and Mackay 1996, Eshel and Matessi 1998, Gibson and Wagner 2000), but
see (Siegal and Bergman 2002).

Here we measure the level of genetic variation and developmental stability in genital
and somatic traits in D. melanogaster. Consistent with our understanding of the
mechanisms that generate morphological scaling relationships, we find that genital traits
are genetically canalized. However, the genitalia are only canalized with respect to
genetic factors that affect the size of all organs in the body systemically. Genital traits
are not canalized with respect to genetic factors that affect the size of individual organs
autonomously. Further, we find that genital traits are not developmentally stable as
indicated by elevated levels of fluctuating asymmetry relative to some somatic traits. We
discuss these findings in light of current theories of genital evolution and argue that
stabilizing selection on genital size alone is insufficient to explain their hypoallometric
relationship with body size.

Materials and Methods
Fly Stocks
Male flies were from 38 of the Core40 isogenic wild-type lines from the
Drosophila Genetic Reference Panel (DGRP).

17	
  

Fly Rearing
Genetic Variation
Genetic variation was assayed among the Core40 isogenic DGRP lines. Larvae
from each line were reared in vials at low density (≤ 50 larvae) on standard
cornmeal:molasses medium at 25˚C in constant light. We collected, dissected and
measured males from at least three vials per line, totaling ten males per line.

Developmental Instability
We used three of the Core40 DGRP lines to assay developmental instability
(lines 303, 324, 335). Larvae were reared at low density (≤ 50 larvae) in ten vials per
line, as described above. We selected, dissected and measured five males from each
vial, totaling 50 males per line.

Morphology
Five organs were dissected from each male fly: three somatic traits (the wing, the
femur of the first leg, and the maxillary palp) and two genital traits (the posterior lobe of
the genital arch and the anal plate). Organs were mounted in dimethyl hydantoin
formaldehyde for imaging. Organ measurements were taken as area for the wing, the
maxillary palp, the posterior lobe of the genital arch, and the anal plate and as the
length of the femur, using a Leica DM6000B compound microscope and Retiga 200R
digital camera. We also measured a fourth somatic trait, thorax length, as the distance
between the attachment of the neck to the posterior tip of the scutellum using a Leica
MZ16FA dissecting microscope and a Leica DFC250 digital camera. Images were

18	
  

analyzed using ImagePro. All linear measurements were squared prior to analysis to
convert them to the same dimension as area measurements. All data were then natural
log transformed to allow the fitting of the linear allometric equation. For the
measurement of FA, we measured the wing, femur, maxillary palp and posterior lobe of
the genital arch from both sides of the fly three times, and calculated measurement
error using the methods of Palmer and Strobeck (Palmer and Strobeck 2003).

Analysis
Genetic Variation
We fit the data to the following linear model:

Y = u +G+e

where Y is the morphological€measurement, u is the intercept term, G is the effect of
line (random factor) and e is remaining non-genetic variation. We used the lmer function
in the lme4 package in R (R-Development-Core-Team 2009) to extract the variance
components using REML for G, which is a measure of the total genetic variation of Y,
here referred to as VT. Each VT was then converted into a coefficient of variation (CVT)
using the formula CVT = eVT −1 (Wright 1952). CVT was used as a measure of a trait’s
total genetic variation.
€

We reanalyzed the data but statistically controlled for variation in other traits by
including them as covariates in our model. This allowed us to estimate the amount of

19	
  

genetic variation in a trait that was orthogonal to and independent of variation in all
other traits, that is a trait’s organ-autonomous genetic variation. The final model was:

Y = u + A + B + ... + G + e

€ of all other traits. We then extracted the variance
where A, B etc. are the size
components for G, (VI) which is a measure of the organ-autonomous genetic variation
of Y, here referred to as VI, from the analysis. We used VI to calculate the organautonomous coefficient of variation (CVI) using the formula CVI = eVI −1 .
Each dataset was sampled with replacement to generate 1000 bootstrap datasets,
€
which were analyzed and used to construct a 95 percent
confidence interval of each

trait’s total (CVT) and organ-autonomous (CVI) genetic variation.

Genetic Static Allometry
The allometric coefficient of the genetic static allometry (where size variation is
solely a consequence of genetic variation) was calculated from the mean logtransformed trait measurements for each line. We used these data to calculate the
variance-covariance matrix for traits among lines, and extracted the first eigenvector
from this matrix using the svd function in the base package of R (R-Development-CoreTeam 2009). The allometric coefficient is reflected by the loadings of the first
eigenvector. Isometry occurs when all loadings of the vector equal 1/ n , where n is the
number of traits measured. Multiplying the loadings by

n gives the bivariate allometric

€
coefficient for each trait against a measure of overall body size (Klingenberg 1996). We
€
20	
  

used a random-variable bootstrap method to generate 95 percent confidence intervals
for the allometric coefficients for each trait (Shingleton et al. 2009).

Developmental Instability
Here we define developmental instability as the imprecision that results from
developmental noise, the random developmental processes that cause a trait to deviate
from its expected growth trajectory given its genotype and environment. Conversely, the
capacity of the growing organ to counteract developmental noise is defined as
developmental stability. Fluctuating asymmetry is therefore a measure of developmental
instability, and is reduced in organs that are developmentally stable. We used the
FA10b index, which corrects for measurement error, to quantify fluctuating asymmetry
for the wing, maxillary palp, femur and posterior lobe of the genital arch (Palmer and
Strobeck 2003). To calculate FA10b we fit the repeated measurement of each trait to
the following model:

Y = u + S + I + SI + e

€ S is the effect of body side, left or right (fixed factor), I is
where u is the intercept term,
the effect of the individual (random factor), SI is the interaction between individual and
side and e is measurement error. We used the lmer function in the lme4 package in R
(R-Development-Core-Team 2009) to estimate the variance components for SI ( σ2SI ),
which is used to calculate FA10b:

€

21	
  

FA10b = 0.798 2σ2SI

We used the MCMCglmm€function in the MCMCglmm package in R (R-DevelopmentCore-Team 2009) to generate values of 95 percent support for each trait’s level of
fluctuating asymmetry. We used a prior equal to the variation in wing size
measurements to generate parameter estimates and compared the results to those
using a non-informative prior and found no difference in parameter estimates.

All traits were tested for antisymmetry and directional asymmetry by assaying the
distribution of trait size on the right (R) and left (L) side of an individual. For
antisymmetry we tested the (R – L) distribution for normality and for directional
asymmetry we compared the mean of the signed (R – L) to zero (Palmer and Strobeck
2003). (R – L) for almost all traits was normally distributed (Shapiro-Wilk test, P > 0.004
with Bonferonni correction). The only exceptions were the maxillary palps of line 303
and wings of line 335 (P<0.004 for both). Plotting R versus L for the size of both these
traits suggested three maxillary palp measurements from line 303 and seven wing
measurements from 335 were outliers. Removal of these data normalized the
distribution of (R – L) for both these traits, although their inclusion had no effect on the
analysis (not shown). The maxillary palp of line 324 also showed evidence of slight
directional asymmetry with mean (R – L) deviating significantly from zero (t-test,
p<0.004 with Bonferoni correction). However, the mean (R – L) was less than FA4a
(where FA4a = 0.798 σ2(R −L) ,), and so any directional asymmetry was considered to be a
consequence of developmental instability (Palmer and Strobeck 2003).

€
22	
  

Results
The total amount of genetic variation (CVT) was lower for the genital traits (genital
arch and anal plate) than for the somatic traits (wing, maxillary palps and thorax) (Figure
2.1A), although when correcting for multiple comparisons this reduction in genetic
variation was significant only for the genital arch (Tukey’s HSD, p < 0.05). In contrast,
none of the traits differed in their level of organ-autonomous genetic variation (CVI) –
that is the amount of genetic variation in trait size that is not correlated with variation in
the size of other traits – when correcting for multiple comparisons (Tukey’s HSD, p >
0.05) (Figure 2.1B). These results suggest that very little of the variation in genital trait
size is a response to variation in genetic factors that affect all traits systemically. It is the
response to these systemic genetic factors that controls the slope of an organ’s scaling
relationship with body size on a genetic static allometry: traits with low response should
scale hypoallometrically with body size. Correspondingly, we found that the genital traits
were significantly more hypoallometric to overall body size than most somatic traits
(Figure 2.1B). Interestingly, the femur of the first leg, like the genital traits, displayed
both low levels of genetic variation and scaled more hypoallometrically to overall body
size than other somatic traits (Figures 2.1A & 2.1B).

Although the genital traits were genetically canalized with respect to factors that affect
organ size systemically, they did not show low levels of developmental instability. In
contrast, within each of the three lines examined, the maxillary palp and the genital arch
had significantly higher levels of fluctuating asymmetry than either the wing or the femur
(Tukey’s HSD p > 0.05) (Figure 2.1C).

23	
  

Figure 2.1. Genetic variation, allometric coefficient and fluctuating asymmetry of
somatic and genital traits in male Drosophila melanogaster. (A) Genital traits had low
levels of total genetic variation (light gray bars, CVT) but not low levels of organautonomous genetic variation (dark gray bars, CVI). The difference between total
genetic variation and organ-autonomous variance is an estimate of genetic variation
that is correlated with variation in other traits (‘systemic’ genetic variation). Columns with
the same letter are not significantly different for total genetic variation (CVT) using
Tukey’s HSD (P > 0.05). Traits do not differ for organ-autonomous genetic variation
(CVI) using Tukey’s HSD (P > 0.05 for all) (B) The low systemic genetic variance of the
genital traits reflected their low multivariate allometric coefficient compared to most

24	
  

Figure 2.1 (con’t). somatic traits, although these differences are not significant for
multiple comparisons (Tukey’s HSD, P>0.05 for all). Grey horizontal line is isometry.
(C) Genital traits did not show low levels of fluctuating asymmetry. Light grey bars, line
303, white bars, line 324, dark grey bars, line 335. Within a line, columns with the same
letter are not significantly different for FA using Tukey’s HSD (P 0.05). All error bars are
95% confidence intervals.

Figure 2.2 Model of a selection regime that alters the slope of the genital-body scaling
relationship while maintaining genital-autonomous genetic variation. Selection is for
proportionally smaller genitalia in large males and proportionally larger genitalia in small
males (black arrows). Implicit to such a regime is that there is selection or constraint
maintaining variation in body size (gray arrows).

25	
  

Discussion
Elucidating the causes of the unusual scaling relationship between genital size
and body size in arthropods is an active but unresolved area of research (Eberhard et
al. 1999, Green 1999, Huber 2003, Bertin and Fairbairn 2007, Eberhard 2009). The goal
of our study was to begin to explore the proximate mechanisms that underlie genital
hypoallometry, specifically the response of male genital size to genetic variation and to
stochastic developmental errors.

The slope of a scaling relationship between body and organ size captures the extent to
which factors that generate variation in body size also generate variation in organ size
(and vice versa). Consequently, traits that scale hypoallometrically to body size, such as
the genitalia, are expected to show low levels of variation in response to genetic and
environmental factors that affect both body and organ size. Previous studies have
shown that, as expected, genital traits show low levels of variation in response to
environmental factors that affect body and organ size; the genitalia are thus
environmentally canalized (Shingleton et al. 2009). Our data show that genital traits
also show low levels of variation in response to genetic factors that affect body and
organ size, that is the genitalia are genetically canalized (Figure 2.1A). Importantly,
however, genital traits do not show low levels of variation in response to genetic factors
that autonomously affect their size (Figure 2.1A). These genetic factors presumably
affect organ size at the level of individual organs and not through systemic mechanisms.
The genitalia also do not appear to show low levels of variation in response to
environmental factors that affect organ size autonomously. Fluctuating asymmetry (FA)

26	
  

arises through stochastic perturbations in the developmental process at the molecular,
chromosomal and epigenetic level (Parsons 1990) and is, by definition, not coordinated
across the body. Implicit to the concept of FA is that, since both sides of a bilateral
organism are influenced by identical genes, non-directional differences between the two
sides must be environmental in origin (Waddington 1942). FA can therefore be
considered a reflection of environmental variation that acts at the level of individual
organs (and tissues within those organs) rather than through systemic mechanisms.
Our finding that FA for genital traits is the same or higher than for somatic traits
suggests that genital traits do not have reduced sensitivity to environmental factors that
act autonomously on organs or tissues.

Our results suggest that there are two broad classes of developmental mechanisms that
regulate organ size in Drosophila: (1) systemic mechanisms that regulate organ and
body size as a whole, for example the level of circulating growth hormone; and (2)
organ autonomous mechanisms that affect the size of organs individually, for example
the expression of genes that pattern individual organs. The genitalia appear to have
reduced their response to the former but not the latter affectors of size (Figure 1A). This
is to be expected. The slope of an organ-body size scaling relationship captures the
extent to which factors that generate variation in body size also generate variation in
organ size; the evolution of hypoallometry (or hyperallometry) should therefore involve
changes in the response of an organ to these factors.
How the genitalia reduce their response to systemic regulators of size is unclear but is
an area of active research. For example, the developing genitalia are insensitive to

27	
  

changes in insulin signaling, the primary developmental mechanism through which
nutrition regulates growth in all animals (Shingleton et al. 2005). Changes in nutrition
during development affect the level of circulating insulin-like peptides that in turn affects
the rate of cell proliferation in growing tissues. Because the growth rate of the genitalia
is relatively insensitive to changes in insulin-signaling, their final size is less sensitive to
changes in nutrition and the genitalia are nutritionally canalized (Shingleton et al. 2009).
Importantly, insulin-insensitivity could also account for the genetic canalization of the
genitalia and the low slope of their genetic static allometry. Genetic variation in body
size has been linked to allelic variation within the insulin-signaling pathway (Sutter et al.
2007). Organs that are insensitive to changes in insulin-signaling caused by nutritional
variation should also be insensitive to changes in insulin-signaling caused by genetic
variation. More generally, if genetic variation in body and organ size is primarily
mediated by genes involved in the environmental regulation of size, then environmental
and genetic canalization may reflect the same developmental processes.

A deeper understanding of the developmental mechanisms that underlie the genetic
and environmental canalization of Drosophila genitalia will help clarify the adaptive
significance of their low allometric slope. There are a number of alternative hypotheses
to account for genital hypoallometry (Arnqvist 1997). The ‘lock-and-key hypothesis’
argues that male genitalia need to be of a particular size in order to physically fit with
the female genitalia, with strong stabilizing selection for genitalia of an intermediate size
(Eberhard 1985). The ‘one-size-fits-all’ hypothesis is similar but proposes that there is
stabilizing sexual selection rather than natural selection for genitalia of an intermediate

28	
  

size, with females favoring males with such genitalia (Eberhard 1985). These models
have been criticized more recently, in part because empirical studies have revealed
directional selection on genital size in male water striders (Aquarius remigis) despite
being hypoallometric to body size (Bertin and Fairbairn 2007). In response to this
criticism the models have been extended from their original implications of stabilizing
selection to include directional selection. Specifically, hypoallometry may result if
positive directional selection on genital size is more intense for small males than large
males (Eberhard et al. 2009).

Our data suggest that genital hypoallometry is not a consequence of stabilizing
selection on genital size alone. Stabilizing selection should not only reduce the genetic
variation in genital size that is correlated with variation in the size of other traits but also
the genetic variation in genital size that is organ autonomous (Falconer and Mackay
1996, Eberhard et al. 2009), which we did not find. Further, stabilizing selection might
also be expected to reduce the developmental instability of the genital traits (Gavrilets
and Hastings 1994) but see (Siegal and Bergman 2002), also not supported by our
data. Rather, the finding that the genitalia are only canalized with respect to genetic and
environmental factors that generate systemic variation in body and organ size suggests
that selection for hypoallometry has targeted the mechanisms that regulate the
response of the genitalia to these factors. These mechanisms ultimately regulate the
relationship between genitalia and body size (Dreyer and Shingleton 2011).

29	
  

What form of selection would target these mechanisms preferentially? One hypothetical
selection regime favors large genitalia in small males and small genitalia in large males
(Figure 2.2). Such selection is not expected to reduce genital-autonomous genetic
variation. This is because alleles that make the genitalia autonomously large will be
selected against in large males but selected for in small males, maintaining overall
genetic variation. The inverse is true for alleles that make the genitalia autonomously
small. In contrast, alleles that reduce the relative sensitivity of the genitalia to systemic
genetic and environmental regulators of organ size will be favored in both large and
small males. Implicit to this selection regime is the assumption that variation in body and
somatic trait size is maintained, either through selection or constraint (Figure 2.2).
Directional selection on genital size that is more negative in large males than small
males, or more positive in small males than large males, should similarly target genes
that influence the relationship between genitalia and body size and change the slope of
their scaling relationship (Eberhard et al. 2009). However, like stabilizing selection, such
directional selection might be expected to also reduce organ-autonomous genetic
variation in the genitalia (Bulmer 1976, Crnokrak and Roff 1995). On the other hand, of
all morphological traits genital traits may be most closely related to fitness. Fitness traits
seem to have elevated levels of variance (Merila and Sheldon 1999) and this may
counter the effects of directional selection on organ-autonomous genetic variation in
genital size.

Interestingly, the femur of the first leg of male Drosophila, like the genital traits, also
showed low levels of total genetic variation and scaled hypoallometrically to body size.

30	
  

The first legs of male Drosophila carry the sex-combs, thought to be used for grasping
the female genitalia prior to intromission (Cook 1977, Ng and Kopp 2008). One
hypothesis for the reduced total genetic variation of the femur therefore is that similar
selective pressures are acting on the first leg and genitalia in male Drosophila.
In general, the slope of allometric scaling relationships is a multivariate trait that reflects
variation in organ size, body size and the relationship between the two. Specifically, it
describes the extent to which environmental or genetic factors influence trait size
relative to body size. Implicit to the concept of allometry is that these factors should
affect both trait size and body size. Theories as to how allometric slopes evolve must
therefore consider selection on organ size relative to body size, rather than organ size
alone. Consequently, future studies should explore how selection acts on genital size in
males of varying sizes. Our prediction is that, if there is selection for hypoallometry, the
strength and direction of selection on genital size will depend on the size of the male the
genital is attached to.

31	
  

CHAPTER 3:
Female preference and the hypoallometry of male genitalia in Drosophila

32	
  

Introduction
Morphological diversity is one of the primary manifestations of biological
variation. Responses to selection in morphology are observable expressions of the
many intricate pathways and systems that produce a functional organism. A
fundamental property of morphological diversity in living organisms is variation in size
(McMahon and Bonner 1983, Brown and West 2000). The range of sizes across taxa
extends over many orders of magnitude, from single cell organisms up to
conglomerations of many trillions of cells in the largest aquatic and terrestrial organisms
(Schmidt-Nielsen 1984). A shared property across all organisms is that each must be
able to counteract the physical and physiological limitations of their specific habitats.
Organisms must also maintain functionality within the set of traits that defines them,
subsets of which often act in a coordinated fashion with one another (Wagner 1996,
Klingenberg and Marugan-Lobon 2013). To a large degree, the study of morphological
diversification can be summarized by the relative size among traits that combine to form
an organism (Bonduriansky and Day 2003).

Scaling relationships, or how traits are related in size, have been the focus of research
on the evolution of morphology for the past century. The term allometry was first used to
describe how differences in growth rates between traits could result in abnormal scaling
relationships (Huxley 1932). Allometry now typically refers to the proportional change in
size of one trait in relation to another trait, or often in relation to a measure of overall
body size (Gould 1966). Implicit to this definition of allometry is covariation between the
two traits. Historically, the most common method of quantifying scaling relationships has

33	
  

α

been through the use of the allometric equation, y = b*x , where x and y are the sizes of
two given traits (Huxley and Teissier 1936). When log-transformed, the equation
becomes the linear relationship log(y) = α * log(x) + log(b) with a slope of α, called the
allometric coefficient, and an allometric intercept of log(b). Log transformation of trait
values is done to linearize the relationship between the two traits, quantifying variation
of the scaling relationship in the two linear parameters of slope, α, and intercept, log(b)
(Huxley and Teissier 1936, Gould 1966). Log transformation also allows for trait
comparison independent of scale between traits (Kerkhoff and Enquist 2009, Shingleton
2012). Traits that maintain proportion to one another, that is a 1:1 ratio, have an
allometric slope of 1 and are called isometric. Traits that have slopes < 1 or > 1, referred
to as hypo- and hyperallometric respectively, have disproportionate size changes in
relation to one another. Hypoallometric traits maintain a consistent size across a range
of body sizes, meaning smaller individuals have proportionally larger traits and larger
individuals have proportionally smaller traits. Hyperallometric traits are those that are
proportionally larger in large individuals and proportionally smaller in small individuals.

Concurrent with the study of how traits relate to one another in size, the mechanisms
that control size continue to be a burgeoning area of research (Shingleton 2010, Emlen
et al. 2012, Koyama et al. 2013, Lavine et al. 2015). The regulation of size is controlled
on at least two levels: systemically (affecting all tissues in an organism), and organautonomously (affecting tissues independently)(Brogiolo et al. 2001, Emlen and Allen
2003, Shingleton et al. 2007, Tang et al. 2011). For traits to have different final sizes
within a single organism, there must be trait-specific regulation, or response, in the

34	
  

common size determining mechanisms that affect all tissues. A primary mechanism of
size control includes the pathways that transcribe nutritional inputs from an organism’s
environment to growth signals that are conveyed to developing tissues (Oldham and
Hafen 2003, Grewal 2009, Nijhout et al. 2013). The insulin/insulin-growth factor (IIS)
pathway works in concert with several other signaling pathways (e.g. ecdysone/juvenile
hormone (Mirth and Shingleton 2012), sex determining pathway (Kijimoto et al. 2012,
Gotoh et al. 2014) to control how traits respond to environmental variation. It has been
shown that the unusual scaling relationships in some species, such as hyperallometry of
horns in male rhinocerous beetles and the hypoallometry of male genitalia in Drosophila
melanogaster can be correlated with either increased or decreased sensitivity to
signaling of the IIS pathway (Tang et al. 2011, Emlen et al. 2012). While nutrition is only
one of several mechanisms through with final trait size is controlled (Shingleton 2010) it
is of paramount importance in the regulation of coordinated trait development within
organisms and across taxa.

Traits related to sex have experienced a particularly high level of interest in evolutionary
research. This is in part due to the role sexual characteristics play in charismatic sexual
selection processes and speciation (Kraaijeveld et al. 2011) but is also a direct
representation of how unusual sexual characteristics tend to scale in relation to the rest
of the body. Most traits scale at or near the 1:1 ratio of isometry, but sexual traits often
deviate from that pattern (Harvey and Pagel 1991, Green 1992, Bonduriansky 2007,
Eberhard 2009). Across the arthropods, many taxa exhibit hypoallometric male
genitalia, meaning small males have proportionally larger genitalia and larger individuals

35	
  

have proportionally smaller genitalia (Eberhard 2009). During mating, male genitalia are
often physically required to interact with the female reproductive organs meaning she
has direct contact with the male and can assess the male based on his genitalia before
fertilization (Eberhard 1996). If genitalia are important for mating success, why are male
genitalia hypoallometric? Traditionally there have been four hypotheses proposed to
account for the observed morphology of male genitalia:
1. Lock-and-key hypothesis. Male and female genitalia must be a particular size to
physically fit together. Individuals with genitals that deviate from this size are
physically unable to copulate and will be selected against (Dufour 1936).
2. Pleiotropy. Genital morphology is selectively neutral but genital size and shape are
invariant due to stabilizing selection on a pleiotropic trait or traits (Mayr 1963).
3. Sexual selection. Genital size indicates male quality and females have a preference
for males with extreme genital size. Under specific conditions this can theoretically
produce hypoallometry (Eberhard 1985, Bonduriansky and Day 2003).
4. One-size-fits-all. The male genital size that stimulates the broadest range of female
genitalia has the highess fitness. Intermediate-sized male genitalia is expected to
maximize compatibility with females, and is therefore the optimum genital size
(Eberhard et al. 1998).

Male genitalia have been widely recognized as some of the most diverse morphological
structures of animals since before Darwin (Hosken and Stockley 2004). Based on the
incredible variation in genital structures, Dufour (1844) proposed the lock-and-key
hypothesis, genital structures act as a mechanism of species-specific recognition, to

36	
  

account for the divergence of this trait. His hypothesis stated that the male genitalia
were a ‘key’ to the female ‘lock’ and only two individuals of the same species would be
able to physically achieve sperm transfer. The effect of the lock-and-key hypothesis on
genital evolution is generally considered to be weak if present at all, but remains
unfalsified as a contributing factor to diversification of genitalia (Eberhard 1985, Shapiro
and Porter 1989, Mutanen et al. 2006).

The pleiotropy hypothesis states that the diversity of genital morphology is a result of
selection on pleiotropic traits (Mayr 1963). The pleiotropy hypothesis was later refined to
account for the unique prediction of hyper-variability in genitalia and not in other traits
that may also be exposed to pleiotropic selection (Arnold 1973). Pleiotropic effects in
the genitalia were predicted to cause more rapid evolution than other traits because as
one sex developed a modification of genital morphology due to pleiotropy, the genitalia
of the other sex would be expected to adapt to match the new morphology of the first
sex. Thus, there would be a positive feedback of morphological changes in the genitalia
resulting in increased levels of diversification. Additional predictions for the pleiotropy
hypothesis include an increase in genetic variability of genital traits and strong condition
dependency of the genitalia because they are selectively neutral (Arnqvist 1997).
Empirical tests for the effect of pleiotropy on genital evolution have proven difficult,
although pleiotropic effects on genitalia may be possible (Arnqvist and Thornhill 1998).
In general, the effect of pleiotropy on genital morphology is also thought to be weak.
Additionally, when compared to somatic traits the male genitalia of Drosophila are
insensitive to changes in nutrition (Tang et al. 2011) and to systemic genetic factors

37	
  

(Dreyer and Shingleton 2011), both of which are predicted to be equal to somatic trait
levels under pleiotropy, suggesting pleiotropy may have no significant effect on genital
morphology in some organisms.

The hypotheses for genital divergence that have received much more attention are the
sexual selection and one-size-fits-all hypotheses. Both of these hypotheses involve a
preference by females for particular male genital sizes. Sexual selection for genital
evolution includes four mechanisms: Fisherian selection, good genes, sperm
competition and sexual conflict (Hosken and Stockley 2004). Fisherian selection and
good genes both predict directional selection for larger genital sizes as direct indicators
of male quality (Eberhard 1985). Sperm competition occurs when sperm from multiple
males competes to fertilize eggs inside the female reproductive tract (Simmons 2001).
Sexual conflict occurs when the two sexes struggle to maximize their own fitness, with
males typically adapting to remove conspecific sperm and females adapting to retain
sperm from the most fit males (Arnqvist and Rowe 2005).

The one-size-fits-all hypothesis states that intermediate male genital sizes are favored
because they are compatible with the most typical female genital size (Eberhard et al.
1998). Female and male genitalia physically interact, therefore females may choose
males based on the ability of the male genitalia to physically stimulate her reproductive
tract. As a result of selection based on a physical stimulation rather than a visual
inspection of male traits by females to assess male condition, intermediate sized
genitalia could be the preferred size (Eberhard et al. 1998). Ultimately, the one-size-fits-

38	
  

all hypothesis of genital evolution is a specific form of sexual selection where females
are basing their preference for male genital size on tactile feedback and not on the size
of the male trait as an indicator of the male’s condition. Because it is necessary for the
female to physically interact with the male genitalia, the female would be selecting for
fertilization of progeny based on male phenotype. The ability of females to bias
fertilization success of males they mated with is called cryptic female choice (Birkhead
and Pizzari 2002). There are a number of mechanisms by which female Drosophila may
be able to ‘select’ male sperm including sperm dumping of undesired sperm, differential
sperm storage and differential abortion of embryos (Eberhard 1996, Birkhead 1998,
Snook 2005).

An additional caveat to the evolution of genital morphology is that the majority of genital
traits measured in arthropods, and in some vertebrate taxa as well, exhibit
hypoallometry (Eberhard et al. 2009). One of the predictions of the one-size-fits-all
hypothesis is that selection for male genitalia is for the intermediate size, i.e. stabilizing
selection. Hypoallometry is thought to be a possible outcome of stabilizing selection on
one of two covarying traits (Eberhard et al. 1998, Hosken et al. 2005). Sexual selection,
in contrast, has traditionally been considered to exclusively produce hyperallometric
scaling relationships (Green 1992, Kodric-Brown et al. 2006), but this concept has come
under increasing challenge (Bonduriansky 2007). Specifically, Bertin and Fairbairn
(2007) identified genital traits in water striders that are known to be under directional
selection as hypoallmetric. A hypoallometric scaling relationship is a theoretical
outcome of directional selection on a trait with viability costs (Bonduriansky and Day

39	
  

2003) or when the strength of selection is stronger at smaller body sizes (Eberhard et
al. 2009), but to best test which form of selection is acting on male genitalia in vivo,
variation in male genital size alone would be required to see which size genitalia
females actually prefer.

Here, the genetic toolkit of Drosophila melanogaster was used to engineer male flies
with different sized genitalia while protecting the proportionality of somatic traits. The
males with variation in genital size were then exposed to females in three mating
regimes (single male, direct male-male competition, indirect male-male competition) to
identify how selection is acting on the male genitalia.

Methods
Allometric Engineering
All transgenes were first isogenized by backcrossing insertions used into the
wild-type Samarkand (Sam) line of Drosophila melanogaster for five generations. Male
flies of different genital sizes were made using the UAS-GAL4 system to regulate gene
expression in specific tissues (Duffy 2002). To limit expression of genes of interest to
the genital tissues only, a GAL4-driver was used (P{W6 Poxn-GAL4}14-1-7) which is
located on the third chromosome and drove expression in the male genital tissues that
develop into the posterior lobe of the genital arch and in some neuronal tissues (Boll
and Noll 2002) (kind gift of JP Masly). The Poxn driver was paired with a number of
transgenes to manipulate IIS in the developing tissues to create ‘large’, ‘wild-type’ or
‘small’ genital size classes of males. Large-genital males were made using a

40	
  

constitutively active version of the Inuslin Receptor (InR), UAS-InR.CA, genocopying
increased nutrition in the target tissues (Poxn > Inr.CA). Wild-type-genital control males
were generated using a green florescent protein, UAS-GFP.565T (Poxn>GFP) that had
no obvious additional phenotypic effects but did control for the effect of elevated
transcription in the target tissues (Poxn>GFP). Small-genital males were made using a
constitutively active version of the forkhead box transcription factor FOXO, UASdFOXO-3X (f3-9), that genocopies starvation (Poxn>UAS-FOXO-3x). All females used
in mating assays were wild-type (Poxn>GFP).

To determine which male mated with the female in direct and indirect competition
assays, each experimental male used was marked with a ubiquitously expressed GFP
(UAS-UbiGFP, Bloomington, IN). UbiGFP is expressed in all tissues of the male making
the entire body fluoresce when excited by a 405 nm Class II laser. Sperm from a male
labeled with UbiGFP will also fluoresce, as will the bodies of any offspring fertilized by
the labeled sperm. The UbiGFP insertion was on the X-chromosome, hence, only
female offspring were characterized by GFP expression.

While the Poxn driver was expressed primarily in the developing genital tissues, there
was some expression in non-target tissues including neuronal cells (Boll and Noll 2002).
Preliminary mating assay trials determined a significant effect on the courtship latency
of male flies expressing FOXO with the Poxn driver (Figure 3.1), therefore, a GAL80
with neuron-specific expression in the Elav domain (Koushika et al. 1996) was added to
the experimental flies to competitively inhibit transcription of any GAL4 in those tissues.

41	
  

Figure 3.1. Elav-GAL80 expressed with Poxn rescues wild-type mating behavior.
Courtship latency of IIS manipulated flies (Poxn > dFOXO.3x) had significantly longer
courtship latency than control male flies (Poxn > UAS-GFP). Coexpression of an ElavGAL80 that prevented expression of GAL4 in neuronal tissues recovered wild-type
courtship latency (Elav-GAL80:Poxn > FOXO.3x). The symbol (a) denotes significance
(Tukey’s HSD p < 0.0001).

42	
  

The Elav-GAL80 was the kind gift of Christen Mirth. Subsequent preliminary trials
confirmed that the simultaneous presence of Poxn and the GAL80 rescued wild-type
courtship behavior (Figure 3.1). To make the final experimental flies, the Elav-GAL80
was introgressed into all experimental flies (large: Elav-GAL80;Poxn > Inr.del, wild-type:
Elav-GAL80;Poxn > UAS-GFP, small: Elav-GAL80;Poxn > UAS-FOXO-3x ).

Fly Assays
All experimental flies were sorted while in the pupa stage as male or female and
transferred to individual 2mL microcentrifuge tubes for social isolation. It is necessary to
maintain social isolation until exposure to conspecifics for flies being used in behavioral
assays because flies are known to modify behaviors in response to the presence of
conspecifics (Siegel and Hall 1979, Gailey et al. 1982). Females are unreceptive to
mating for the first 48 h after eclosion, therefore all females used in mating experiments
were between 2 and 6 d old (O'Dell 2003). Males were also between 2 and 6 d old.

Flies were reared under a 12 h : 12 h light : dark circadian rhythm, at a constant
temperature of 25°C. Each mating assay consisted of 18 – 35 mm x 10 mm mating
chambers (Corning) with one 35mm disk of filter paper (Whatman) wetted with 70 µm of
distilled water (Appendix A1). For the single male regime, each mating pair consisted of
one female and one male. Six males of each genital size class were randomly assigned
one of the mating chambers and were mouth-aspirated from their individual tubes into
the mating chamber, with a female from her individual tube. Once the mating pair was in
the mating chamber, the mating chamber was placed into a climate-controlled box

43	
  

(24.8°C ± 0.5°C, 55% relative humidity ± 1.5%) and all activity was video recorded for
the duration of the experiment using a Sony Vixia HG21 video camera. The addition of
all flies took approximately 15 min from the time the first mating chamber was placed in
the climate-controlled box until the last mating chamber was in position. All experiments
were started within 1hr of the set ‘daylight’ time, and were allowed to run for 60 min from
the time the last mating chamber was placed into the climate-control box. If flies were
observed to be mating after 60 min, the assay continued until there were no flies
mating. After the trial, males were immediately stored in 75% EtOH with their puparial
case. Females were placed in individual tubes of cornmeal/molasses food and allowed
to oviposit for 2 - 24 h periods, and were then placed in 75% EtOH with their puparial
case. All adults that ecolosed from mating regime trials were counted and scored for
UbiGFP two weeks after their parental experiment date.

For the direct competition regime, two males of different genital size class were
randomly assigned a mating chamber and mouth-aspirated with a female from their
individual tubes into the mating chamber to create a full mating trial of 18 chambers.
After each direct competition trial, males were identified by UbiGFP and placed in 75%
EtOH with their puparial case. Females were placed in individual tubes of
cornmeal/molasses food and allowed to oviposit for 2 - 24 h periods, and were then
placed in 75% EtOH with their puparial case.

For the indirect competition regime, six males of each genital size class were randomly
assigned a mating chamber and mouth-aspirated with a female from their individual

44	
  

tubes into the mating chamber to create a full mating trial of 18 chambers. After the first
trial, males were placed in 75% EtOH with their puparial case. Females were placed in
individual tubes of cornmeal/molasses food and allowed to oviposit for 2 - 24 h periods.
Once the females had laid eggs for 48 h, six males of a different genital size class than
the first male were mouth-aspirated with a female into the mating chamber to create a
second full mating trial of 18 chambers. After the second trial, males were placed in
75% EtOH with their puparial case and females were placed into a second set of
cornmeal/molasses food vials to oviposit for an additional 2 – 24 h periods. Once the
second egg lay period was complete, females were placed in 75% EtOH with their
puparial cases.

Four behavioral parameters were scored using the video recording of each mating trial:
courtship latency, courtship duration, copulation latency and copulation duration.
Courtship latency was recorded as the length of time from the male(s) first being
introduced to the female until any of the typical courtship behaviors was exhibited
towards the female (O'Dell 2003). Copulation latency was the length of time from the
male(s) first introduced to the female until copulation was initiated. Courtship duration
was the time from initial introduction of male(s) to female until copulation was initiated,
or as long as the male(s) continued to court the female if no successful copulation
occurred. Copulation duration was the time from the initiation of copulation until the
male and female separated. These parameters could only be measured in the single
male and the indirect male-male competition assays because there was no visible way
to differentiate between the two males of direct male-male competition.

45	
  

In each of the three mating regimes (single male, direct male-male competition, indirect
male-male competition) 216 males were exposed to females for an overall total of 648
male flies. In the single male assays, 150 male flies were dissected, 50 from each of the
three genital size classes (‘large’, ‘wild-type’ or ‘small’). Males in the direct competition
assays were dissected if at least one of the pair successfully mated with the female (n =
50). Males in the indirect competition assays were only dissected if both males
successfully mated with the female (n = 48). For the direct and indirect regimes, only
males that successfully copulated were included in the analysis because female
preference for genital size based on UbiGFP expression could only be made if a male
sired offspring.

Morphology
Three traits were measured for each male fly: puparial case area, wing area and
area of the posterior lobe of the genital arch. Two organs were dissected from each
male fly: the wing and the posterior lobe of the genital arch. Organs were mounted in
dimethyl hydantoin formaldehyde for imaging. Organ measurements were taken as area
for the wing and the posterior lobe of the genital arch using a Leica DM6000B
compound microscope and Retiga 200R digital camera. Puparial case area was also
measured as an indication of overall body size. The size of the puparial case is directly
proportional to the size of the adult given that all growth for holometabolous insects is
determined by the nutritional intake of the larva (Chiang and Hodson 1950). All images
were analyzed using the program ImagePro. For an estimation of developmental

46	
  

stability, the amount of variation in trait size due to stochastic developmental errors,
fluctuating asymmetry (FA) was calculated for each organ of interest. The wing and
posterior lobe of the genital arch from both sides of each fly were dissected and
measured three times to correct for measurement error, and the analysis of FA was
done using the methods of Palmer and Strobeck (Palmer and Strobeck 2003).

Analysis
Static Allometry
The allometric coefficient for the static allometry was calculated using logtransformed trait measurements. These data were used to calculate the variancecovariance matrix for traits among lines, and the first eigenvector from this matrix was
extracted using the svd function in the base package of R (R-Development-Core-Team
2014). The allometric coefficient is reflected by the loadings of the first eigenvector.
Isometry occurs when all loadings of the vector equal 1/ n where n is the number of
traits measured. Multiplying the loadings by

n gives the bivariate allometric coefficient

€
for each trait against a measure of overall body size (Klingenberg 1996). We used a
random-variable bootstrap method €
to generate 95 percent confidence intervals for the
allometric coefficients for each trait (Shingleton et al. 2009).

Mating Assay – Behavioral Traits
To compare male mating behavior across groups, and to identify any
confounding effects of the IIS manipulations, data were fit to a linear mixed model with
the mating behavior as the response variable, genotype, relative humidity and

47	
  

temperature as fixed effects and mating position (1-18) and rearing vial as random
factors. Possible mixed models were generated using maximum likelihood estimates
and compared using log likelihood ratio tests. The final model was selected as the
simplest model where the addition of a factor did not significantly improve the fit of the
model. Individuals were only included in behavioral trait analysis if the trait was
applicable (e.g. copulation duration was only analyzed for flies that actually copulated).
Tukey’s Honest Significant Difference (HSD) tests were completed for each mating
regime to compare the behaviors between genital size classes within selection regimes.

Mating Assay – Female Preference
To calculate female preference for males of different genital sizes the probability of a
female successfully mating, and subsequently fertilizing eggs, was calculated.
Probability of mating was estimated using logistic regression on the bivariate data of
success or failure for mating or fertilization compared to genital size, wing size and
puparium size. Likelihood ratio tests were used to compare the regression model
parameters for significance. Parameters were only included if they significantly
improved the fit of the model. For each mating regime, males were only included in the
test for fertilization probability if they had mated with the female.

Mating Assay – Reproductive Success
Reproductive success, measured here as the number of offspring each male
sired, was compared across the genital size classes for each mating regime using
Ordinary Least Squares (OLS) regression. Linear regression was also completed using

48	
  

major axis and standard major axis methods with no significant slopes estimated for any
method (data not shown). Only males that were confirmed to have produced offspring
were included in the analysis. Tukey’s HSD tests were completed for each mating
regime to compare reproductive success between genital size classes within selection
regimes.

Developmental Stability
To calculate FA for the wing and posterior lobe of the genital arch the FA10b
index was used, which corrects for measurement error (Palmer and Strobeck 2003). To
calculate FA10b the repeated measurement of each trait were fit to the following model:

Y = u + S + I + SI + e

€ S is the effect of body side, left or right (fixed factor), I is
where u is the intercept term,
the effect of the individual (random factor), SI is the interaction between individual and
side and e is measurement error. I used the lmer function in the lme4 package in R to
estimate the variance components for SI ( σ2SI ), which is used to calculate FA10b:

€

FA10b = 0.798 2σ2SI

The MCMCglmm function€in the MCMCglmm package in R (Hadfield 2010) was used to
generate values of 95 percent support for each trait’s level of fluctuating asymmetry. A
prior equal to the variation in wing size measurements was used to generate parameter

49	
  

estimates that were compared the results using a non-informative prior. There was no
significant difference in parameter estimates. All traits were tested for antisymmetry and
directional asymmetry by assaying the distribution of trait size on the right (R) and left
(L) side of an individual. For antisymmetry we tested the (R – L) distribution for
normality and for directional asymmetry we compared the mean of the signed (R – L) to
zero.

Results
Allometric Engineering
The absolute size of the genitalia varied by genital size class, but not the size of
the wings or body size (puparial case area) (Figure 3.2). Within each size class there
was no significant difference in absolute size of either the wing or genitalia between the
genotypes labeled with the UbiGFP marker and those without (Tukey HSD p > 0.05),
therefore, data were pooled across size classes for the analysis of the mating assays.
Among size classes, there were significant differences between the small males and
both the large and wild-type males (

large

= 8.11 µm2,

wild-type

= 8.10 µm2,

Small

=

7.56µm2,Tukey’s HSD p < 0.001) but not between the large and wild-type genital sizes
(Tukey’s HSD p > 0.05) (Figure 3.2). Additionally, there was no significant difference in
size of the wing or puparial case size across any of the genital size classes (Tukey’s
HSD p > 0.05).

50	
  

Figure 3.2. IIS perturbation using Poxn changes size of the genitalia only. Size of the
posterior lobe of the genital arch was significantly smaller than the large or wild-type
genital size classes (Tukey HSD < 0.001, represented by symbol ‘a’). There were no
significant differences in size of the wings or puparial between the genital size classes
(Tukey HSD > 0.05). Data for genital size classes were pooled by the IIS transgene
present in the fly (large: InR.CA, wild-type: UAS-GFP, small: dFOXO.3x) because there
was no difference in size between flies with or without the Ubi-GFP (Tukey HSD > 0.05).
Error bars represent SEM.

51	
  

Static Allometry
Multivariate allometric coefficients were distinct between the two traits (wing and
genitalia) but not within trait by genotype (Figure A2). Genitalia across the three
genotypes had a hypoallometric coefficient (

α =0.512) confirming what has been

shown in other studies. The wings were slightly hyperallometric (

α = 1.32).

Mating Assay – Behavioral Traits
Male Drosophila mating behaviors were analyzed across the single male and
indirect competition mating regimes using a linear mixed-model. Mating behavior not
measured in the direct competition assays because males were indistinguishable during
the mating trial. There were no significant differences for the early stages of courtship in
either the single male or indirect competition regimes between genotypes (Appendix,
Table A1, Tukey HSD < 0.05). Differences in behavior for mating latency and mating
duration were attributed to variation in factors beyond our control, as there were no
differences in the initiation of courtship.

Mating Assay – Female Preference
Female flies preferred mating with wild-type-sized genitalia in all mating regimes.
Successful and failed copulations and fertilizations by males were coded as a 1 or 0,
respectively, and were analyzed using logistic regression for each mating regime.

Females were more likely to copulate as genital size increased in the single male
regime (p < 0.05 based on 123 matings, Figure 3.3) and more likely to fertilize eggs

52	
  

Figure 3.3. Males with larger genitalia are more likely to mate in single male regime.
Logistic regression of male mating success (probability = 1.0) or failure (probability = 0)
by log-transformed genital size of males. Females were more likely to mate with males
that had larger genitalia (p = 0.009). Males with small genitalia were able to successfully
mate with females, but did so at a lower frequency than males with larger genital sizes.
Histogram bars in grey (top) show frequency of successfully copulation and black bars
(bottom) show frequency of failure to copulate by genital size. Line represents logistic
regression probability curve.

53	
  

Figure 3.4. Males with larger genitalia are more likely to sire offspring in single mating
regime. Logistic regression of a female fertilizing (probability = 1.0) or not fertilizing
(probability = 0) eggs by log-transformed genital size. Females were more likely to
fertilize eggs using sperm from males with larger genitalia (p = 0.001). Females fertilized
eggs using sperm form males with small genitalia, but at a lower frequency than sperm
from males with larger genital sizes. Histogram bars in grey (top) show frequency of
successful fertilization and black bars (bottom) show frequency of failure to fertilize by
genital size. Line represents logistic regression probability curve.

54	
  

using sperm from males with larger genitalia (p < 0.05 based on N = 3435 progeny from
90 female, Figure 3.4). In the direct competition regime, females were more likely to
fertilize eggs using sperm from males with larger genitalia (p < 0.05 based on N = 505
progeny from 25 females that could be assigned a mate, Figure 3.5). Only males that
sired offspring were included in the analysis as the identity of the male the female chose
could only be determined by the presence/absence of GFP in offspring. Therefore, the
probability of fertilization was 1.0 for all males recorded in the direct competition regime.

In the indirect regime only male pairs that both successfully mated with the female were
included in the analysis. The competition of male sperm, and the cryptic female choice
of which male’s sperm to use, could only be assessed if a female had mated with both
males she was exposed to. Therefore, the probability of mating was 1.0 for all males in
the indirect competition regime. Females were more likely to fertilize offspring using
sperm from males with larger genitalia (p = 0.031 based on N = 333 progeny that could
be assigned sired from 23 females , Figure 3.6). Females in the indirect competition had
mated twice and therefore had sperm from two males available in their seminal
receptacle and spermathecae. There was no evidence of sperm precedence as
evidenced by no significant effect of the interaction between male genotype and mating
order on the number of progeny sired (all pairwise comparisons p > 0.05 Tukey’s HSD).

For the mating regimes that measured male mating probability, males of all genital sizes
were physically able to mate with the female, but did so at lower frequencies than males
with larger genitalia (Figures 3.3, 3.5). For each of the mating regimes, a linear

55	
  

Figure 3.5. Males with larger genitalia were more likely to mate in the direct competition
regime. Logistic regression of a male mating success (probability = 1.0) or failure
(probability = 0) by log-transformed genital siz. Females were more likely to mate with
males that had larger genitalia (p = 0.002). Males with small genitalia were able to
successfully mate with females, but did so at a much lower frequency than males with
larger genital sizes. Histogram bars in grey (top) show frequency of successful
copulation and black bars (bottom) show frequency of failure to copulate by genital size.
Line represents logistic regression probability curve.

56	
  

Figure 3.6. Males with larger genitalia were more likely to sire progeny in the indirect
competition regime. Logistic regression of female fertilizing (probability = 1.0) or not
fertilizing (probability = 0) eggs by log-transformed absolute genital size. There was a
significant positive relationship between genital size and fertilization probability (p =
0.031). Sperm from males with small genitalia was used to fertilize eggs, but at a much
lower frequency than sperm from males with larger genital sizes. Histogram bars in gray
(top) show frequency of fertilization and black bars (bottom) show frequency of failure to
fertilize by genital size. Line represents logistic regression probability curve.

57	
  

regression was used to compare genital size and the total number of progeny a male
sired if he mated and his sperm was used to fertilize eggs. Across each regime, there
was no significant relationship or correlation between number of offspring and logtransformed male genital size (psingle = 0.094, pdirect = 0.537, pindirect = 0.608)(Appendix,
Table A2).

Developmental Stability
Fluctuating asymmetry (FA) was calculated for all males that were exposed to
females in any of the mating contexts. FA for both the wing and genital arch were
estimated. Males were compared by genotype within a mating regime. There was a
substantial difference in FA values between the wings and the genitalia in each of the
mating regimes, with the FA much higher in the genitalia than in the wings (Figure 3.7).
There was no significant difference in FA across the three genotypes within a trait
indicating females were likely unable to derive any information on a male based on the
FA of his genitalia. Because there was no significant difference across the size classes
of males in the single male regime, males were pooled based on if they were selected
by the female in direct competition or not, and by if the male produced offspring or not in
the indirect competition regime. There was no difference between the FA of males that
were selected by females vs. those that were not (Figure 3.8) or of males that sired
offspring vs. those that did not (Figure 3.9).

58	
  

	
  

Figure 3.7. Fluctuating asymmetry (FA) is higher in genital traits than the wings in the
single male regime. There was no difference in FA values by trait across the genital size
classes, but there was a difference between the traits. FA was calculated using the
FA10b value which accounts for measurement error in log-transformed traits. Error bars
are 95% confidence intervals.

59	
  

Figure 3.8. No difference in FA between flies that mated and did not mate in the direct
competition regime. Error bars are 95% confidence intervals.

60	
  

Figure 3.9. No difference in FA between flies that sired progeny and those that did not
in the indirect competition regime. Error bars are 95% confidence intervals.

61	
  

Discussion
Female Drosophila melanogaster preferred males with larger genital sizes than
those with smaller genital sizes. Using the genetic tools available for D. melanogaster
the only detectable trait that varied in size across the treatment groups were the
genitalia, eliminating any confounding effect of covariation in other traits. This was
especially important with regards to the wings as they are known to play a critical role in
the courtship display of males (Spieth 1974).

Female Drosophila have ample opportunity to prevent a male from forcing copulation
(Spieth 1995, Lasbleiz et al. 2006 2006). That a female can physically assess a male
and his genitalia prior to mating with him indicates that genital size may be an important
factor in female mate choice. The significance of genital morphology is further
highlighted by the increased levels of diversity in the primary sexual traits across taxa
(Eberhard 1985, Hosken and Stockley 2004). It is well documented that male genitalia
across the phylum Arthropoda, and beyond, exhibit hypoallometry of their genitalia in
relation to overall body size (Eberhard 2009). The hypoallometry of male genitalia is
also commonly attributed to stabilizing selection acting on the genitalia (Eberhard et al.
2009). The ‘one-size-fits-all’ hypothesis states that the intermediate size genitalia would
be selected because they are compatible with the larges number of potential female
mates (Eberhard et al. 1998). Theoretically, stabilizing selection on one of two
correlated traits could result in a hypoallometric relationship where one trait has very
little variation in size across a range of sizes in the associated trait. However, the

62	
  

paucity of empirical data supporting this hypothesis has made it difficult to reach a
definitive answer.
Using a range of metrics, the preference of female Drosophila for genital size was
directly addressed. Females had a clear preference of copulating with males that had
wild-type-sized genitalia when compared to males with smaller genitalia, both when the
females had only one male to mate with and when they had a choice (Figures 3.3, 3.5).
The increased likelihood of a female choosing to mate with a male that had larger
genital size when exposed to only one male suggests preference for larger genital size
may have a genetic correlation with larger genital size. Males with smaller genital sizes
were able to successfully mate with females, suggesting that the lock-and-key
hypothesis is inconsistent with the observed genital morphology of male Drosophila.

Female Drosophila are able to control mating via behavioral modifications during
courtship (Lasbleiz et al. 2006) and can also control fertilization. There was likely a
significant effect of cryptic female choice as females preferentially fertilized progeny
using sperm from males with larger genital size. This outcome supports the hypothesis
that some form of sexual selection is occurring on the male genitalia in Drosophila as
cryptic female choice is one of the primary mechanisms of sexual selection (Eberhard
1996). There was no difference in the final number of progeny sired by a male across
any of the mating regimes, however. Eggs are a costly investment for females, and the
data suggests that once a female decides to copulate and fertilize eggs using sperm
from a male, she no longer differentiates between male genital sizes.

63	
  

Sexual selection acting on male genitalia could be based on an assessment of male
quality based on their genitalia (Eberhard 1985). One measure females could use to
determine male quality is an appraisal of developmental stability. Developmental
stability refers to the ability of an organism to resist stochastic developmental errors
resulting in small levels of variation in phenotype, particularly across planes of
symmetry. Males that are of high quality would be expected to have very symmetrical
traits due to increased levels of nutrition and general developmental robustness. While
there are some traits that are known to be asymmetrical (e.g. antisymmetry of male
fiddler crab claws), male Drosophila genitalia have no bias in their symmetry (Dreyer
and Shingleton 2011). No evidence of was found to support the hypothesis that females
use developmental stability of male genitalia to guide mate choice (Figures 3.7 – 3.9).
There was no difference in FA between the different size classes or between males that
mated or did not mate in the direct and indirect competition regimes.

The genitalia of male Drosophila are hypoallometric, but the type of selection acting on
them is unclear. Specifically, there is evidence for sexual selection, but the distinction
between directional or stabilizing selection is equivocal. The hypoallometry of male
Drosophila genitalia has been shown to be a result of reduced sensitivity to IIS signaling
(Tang et al. 2011). In this study, modification of the IIS pathway in the developing
genital tissues was only able to decrease the size of male genitalia, not increase it
(Figure 3.10). IIS signaling was depressed in the developing genitalia by expression of a
constitutively active form of the negative growth regulator, FOXO, and enhanced by
expression of a constitutively active form of the positive growth regulator, InR. The

64	
  

unidirectional effect of IIS perturbation in the genitalia stands in contrast to the response
of Drosophila wings which significantly increased and decreased in size with up- and

Figure 3.10. Response in genitalia and wings to IIS manipulation to make large and
small traits. Wing significantly increased and decreased in size in response to upregulating (large: NP6333>InR.CA) and down-regulating (small: NP6333>InR.DN25) IIS
in those tissues (Tukey’s HSD < 0.05). Posterior lobe of the genital arch only decreased
in size in response to down-regulating IIS (small: Poxn>dFOXO.3x) but did not increase
in size when IIS was up-regulated (large: Poxn>InR.CA) suggesting the genitalia are
insensitive to positive nutritional growth regulators or are at a maximum size in

65	
  

response to nutrition. Error bars are 95% confidence intervals. Control flies were > UASGFP for both traits.

66	
  

down-regulation of the same pathway via InR mutants (Figure 3.10). The lack of size
increase in genitalia in response to an increase in perceived nutrition suggests that they
may be insensitive to increases in InR or may be at a maximum size as determined by
IIS signaling.

The most parsimonious explanation for why Drosophila genital tissue would respond to
a downstream IIS growth regulator that decreases size but not an upstream IIS growth
regulator that increases size is that the trait is insensitive to changes in nutrient
signaling or it can simply not get any larger through insulin-signaling. Directional
selection is a likely type of selection that would result in a trait being as large as it can
be in response to any input, environmental or genetic, but it is possible that strong
stabilizing selection on a trait could isolate it from nutrient signaling as well. It appears
that the genitalia of Drosophila, at least, are insensitive to an increase in InR due to a
reduction in sensitivity to the negative growth regulator, FOXO (Tang et al. 2011). Even
as InR levels increase in the genital tissues, if there is low or no FOXO remaining in the
tissues to shut off, the tissues will not increase in size. There is at least one example of
a hypoallometric genital trait that is the result of strong directional selection: the male
genital structures of the water strider Aquarius remigis (Bertin and Fairbairn 2007).
Directional selection has been proposed as a hypotheses for the hypoallometry of
genitalia but only in the specific conditions of directional selection with viability
increasing with body size but decreasing with trait size (Bonduriansky and Day 2003) or
when selection is acting disproportionately on small males as opposed to large
(Eberhard et al. 2009). More work needs to be conducted to investigate how choosey

67	
  

females are being, and if females would preferentially copulate with and/or fertilize
offspring with males that have absolute larger genitalia. Collectively, the results of these
experiments suggest that the observed hypoallometry of male Drosophila, while
perpetuated by a female preference for larger genital sizes compared to smaller genital
sizes, are also the result of a more complicated underlying developmental size control
structure.

68	
  

APPENDIX

69	
  

Figure A1. Complete mating assay consisting of 18 – 35mm x 10mm mating chambers.

70	
  

Figure A2. There was a difference in allometric coefficient between, but not within,
traits. The allometric coefficient was lower in the genitalia thatn the wing, as expected.
he allometric coefficients do not vary across size class, but are different between traits.
Error bars are 95% bootstrap confidence intervals from 1000 replicates.

71	
  

Table A1. Pairwise comparisions by genotype of mating behaviors. No significant
difference between genotypes for early stages of courtship. Significant differences in
behaviors are indicated by (*). Direct competition regime was not analyzed because
males could not be differentiated in video recording of mating trial. All p-values
calculated by Tukey’s HSD tests.

Courtship Latency
Courtship Duration
Copulation Latency
Copulation Duration

No Competition
UASUASInR.CA &
GFP &
GFP &
FOXO.3x
InR.CA FOXO.3x
0.644
0.908
0.389
0.526
0.996
0.513
0.090
0.895
0.043*
0.008*
0.071
0.766

Indirect Competition
UASUASInR.CA &
GFP &
GFP &
FOXO.3x
InR.CA FOXO.3x
0.439
0.843
0.781
0.206
0.438
0.193
0.355
0.712
0.553
0.587
0.802
0.899

Table A2. No relationship between the number of progeny sired and genital size. Linear
regression of total number of offspring compared to genital size for each mating regime.
No correlation or significant relationship between genital size and reproductive success
across all males that sired offspring.

OLS
slope
R2
p-value

Single
Male

Direct
Competition

15.0

-10.9

Indirect
Competitio
n
9.63

0.031
0.094

0.016
0.537

0.016
0.608

72	
  

CHAPTER 4:
Trait development and the evolution of allometries in silico

73	
  

Introduction
Morphological diversity is broadly characterized by variation in body proportion
(Brown and West 2000). The variation in size among traits can largely be attributed to
adaptations that are a response to the universal requirements for survival. All organisms
must overcome the biological necessities such as securing nutrition, out-competing
other organisms and reproducing while balancing the physical and physiological
limitations of life on Earth. Natural selection has given rise to extreme biological
diversity across orders of magnitude from the smallest to the largest organisms. In an
effort to focus on tractable systems of study on the incredible range of size variation,
scientists have developed methods to describe patterns of phenotypic divergence. One
approach that has received great interest over the past century is the study of scaling
relationships between biological traits (Gayon 2000). Research on the relationships
between correlated traits in animals often compares some measure of body size to a
correlated trait, for example, metabolic rate (White and Seymour 2005) exaggerated
secondary sexual characteristics (Emlen and Nijhout 2000, Kodric-Brown et al. 2006,
Eberhard et al. 2009, Nijhout and German 2012), and brain size (Lande 1979). The
correlation between traits within and between species serves as a way to measure
morphological variation quantitatively, which can help address the question of why we
observe the phenotypic diversity that we do.

Research on scaling relationships focuses on how the relative size of traits varies within
and among species (Schmidt-Nielsen 1984, Brown and West 2000).The differences
between many closely related taxa can be summarized by the scaling relationships

74	
  

between different traits (Cheverud 1982). In the class insecta, for example, diversity in
relative size among the fixed set of characteristic insect features (three pairs of legs,
head, thorax, abdomen, two pairs of wings) represents their wide-ranging morphological
differences. Applying this principle to other groups of organisms, it is readily apparent
how important body proportion is to quantifying morphological diversification.

Traditionally, body proportion has been described using the scaling relationship
between two morphological traits, or between trait size and body size. How traits scale
in relation to one another is commonly referred to as an allometry (Cock 1966, Gould
1966). The comparison of scaling between traits can be done at the level of taxanomic
groups (evolutionary allometry), within a species at different developmental stages
(ontogenetic allometry) and within a species at the same developmental stage (static
allometry) (Cheverud 1982). Static allometries can be further subdivided based on the
causes of size variation. Variation due to genetic factors generate genetic static
allometries while variation due to environmental factors, such as nutrition or
temperature during development, generate environmental static allometries (Shingleton
et al. 2007). Traditionally, the scaling relationships of static allometries are modeled
α

mathematically using the allometric equation, y = bx where y and x are measurements
of morphological traits (Huxley and Teissier 1936). When log-transformed, the allometric
equation produces a linear relationship of log(y) = α * log(x) + log(b) with a slope of α,
the allometric coefficient, and an allometric intercept term, log(b) (Gould 1966,
Klingeberg 1998). Changes in body proportion are easily described by changes in the
slope, α, and the allometric intercept, log(b), resulting in widespread application of the

75	
  

allometric equation to myriad examples of scaling relationships including those in
vertebrates (Maina et al. 1989, Dixon 1993, Lemaitre et al. 2014) invertebrates
(Eberhard 2009, Arendt and Fairbairn 2012) and plants (Niklas 1994, Chave et al.
2005).

Complementing the extensive empirical work on allometry, there have been many
hypotheses of how allometries are expected to change in response to selection (Green
1992, Eberhard et al. 1998, Bonduriansky and Day 2003, Kodric-Brown et al. 2006,
Bertin and Fairbairn 2007, Nijhout and German 2012, Pelabon et al. 2014). However,
these hypotheses have been unable to provide a unifying framework to explain the
breadth of empirical research. For example, the hypoallometry of male genitalia in
arthropods is typically believed to be the result of stabilizing selection on intermediate
genital size (Eberhard et al. 1998, Eberhard 2009), but it has been shown that genital
traits under strong directional selection also exhibit hypoallometry (Bertin and Fairbairn
2007). In practice, it has also proved difficult to alter the slope of an allometry through
artificial selection on the allometric coefficient alone to test the theories (Egset et al.
2012). The difficulty of connecting predictions to observed patterns may result from
hidden assumptions about the structure of population-level scaling relationships and
suggests the need for an updated framework for allometry evolution.

Scaling relationships are the result of covariation between traits in response to factors
that regulate trait growth (Stern and Emlen 1999, Shingleton et al. 2008). Therefore,
incorporating our understanding of these mechanisms into predictions of allometry

76	
  

evolution is a necessity. Some of the most influential pathways controlling the
development of trait size are nutrient-sensing pathways (Brogiolo et al. 2001, Oldham
and Hafen 2003). Coordination of growth in response to nutritional access during
development has been well studied in holometabolous insects with a focus on the
highly-conserved insulin/insulin-growth factor signaling pathway (IIS) (Claeys et al.
2002, Dupont and Holzenberger 2003). In IIS signaling, the level of nutrition that is
taken in from the environment determines the amount of insulin-like peptides (ILPs)
produced, which initiates a signaling cascade that ultimately controls the size of
individual organs and overall body size of organisms (Mirth and Riddiford 2007, Mirth
and Shingleton 2012, Nijhout et al. 2013). This means that individuals with high levels of
ILPs will typically have large body/trait sizes whereas individuals with lower ILP levels
will exhibit smaller body/trait sizes. The two are not necessarily linked, however. The
relative sensitivity among traits to the ILPs present in a developing organism are one
determinant of relative size between traits, and consequently, of an individual’s static
allometry (Emlen et al. 2006, Shingleton 2012). This relationship between ILPs and
allometry has been supported for both hyperallometric and hypoallometric traits. In male
rhinoceros beetles for example, exaggerated traits have been shown to be more insulinsensitive than other body structures (Emlen et al. 2012). Similar research in Drosophila
has demonstrated that the hypoallometry of male genitalia is due to relative insensitivity
to circulating Drosophila Insulin-Like Peptides (dILPs) in the developing genital tissues
resulting in adult genitalia that remain a consistent size across a range of nutritional
levels (Tang et al. 2011).

77	
  

Empirical and theoretical work has focused on population-level scaling relationships.
However, a key insight is that population-level static allometries (populationSA) reflect
responses at the individual-level to the genetic and environmental factors that regulate
trait size. Individual-level responses are the output of a fixed system (the individual’s
genotype) responding to the environmental factors that affect trait size. If there is
genetic variation among individuals in their responses to the environment, each
individual has an unexpressed genotype-specific scaling relationship of all possible trait
sizes that individual could posses if reared in different environments (Figure 4.1).
Because each individual develops in a single environment, only one of the possible
genotype-specific scaling relationships is ever expressed, but the collection of possible
scaling relationships represents a hidden individual-level static allometry (individualSA).
As will become clear, the patterns of individualSA in a population are extremely
important in the prediction of allometry evolution.

Here we present a simple stochastic mathematical model of population-level covariation
between two traits using current understanding of the developmental mechanisms
underlying trait size within individuals (Stern and Emlen 1999, Shingleton et al. 2008,
Nijhout and German 2012, Nijhout et al. 2013). The model uses the fundamental
developmental parameters listed above to generate a population of individuals, each
with a unique individualSA. These theoretical populations can then be subject to
different forms of selection to determine how the populations as a whole, and the
developmental parameters within each individual, evolve. Establishing the how of

78	
  

proportional trait evolution will be a tremendous step forward in understanding the
broader question of the origins of morphological diversity.

Figure 4.1. Population static allometry and the individual static allometry. (A) A
hypothetical populationSA demonstrating the relationship between covarying traits T1
and T2 comprised of 25 individuals, represented by unique colors. (B) Individuals
express only one phenotype (colored circle) of the possible individualSA possible for
their genotype (colored lines) based on the environmental conditions they are exposed
to during development.

79	
  

Methods
Developmental Basis
Broadly, there are at least two classes of size regulation mechanisms in developing
tissues: those related to systemic factors, and organ autonomous factors (Stern and
Emlen 1999, Emlen and Allen 2003, Shingleton et al. 2007, Tang et al. 2011). Systemic
factors are expressed throughout the entire developing organism. Examples of systemic
factors include circulating hormones, developmental temperature and nutrition levels.
There are also positive and negative regulatory responses to the systemic growth
factors. The organ-specific response controls how sensitive organs are to the systemic
growth factor. Because the organ-specific response of multiple organs is to a single
systemic signal, it follows that the degree of covariation between traits will be dependent
on the ratio of their responses to the shared systemic factor.

The second class of growth regulators are the organ-autonomous growth factors. These
factors are intrinsic to each developing tissue and act independently of any external
influences. Organ-autonomous factors include the following: specific signals that
determine growth rate and duration in individual tissues, patterning genes, and tissuespecific expression of genes such as chico or Akt in developing organs (Bohni et al.
1999, Verdu et al. 1999)

Model
The model of morphological trait development described here was built around growth
as a geometric process. All morphological traits start as a single cell that undergoes

80	
  

many rounds of division to give rise to a final adult trait. Therefore, a modified geometric
growth model was used:
T = pert

[1]

where T is the final trait value, p is the starting number of cells, here set to one, e is the
exponential constant, r is the growth rate and t is time. Time in the model is
standardized to 1 as the model is a simulation of the effects of changing growth rate,
while the growth duration of two tissues within a single organism is assumed to be
equivalent. Thus, the modified geometric growth equation used in the model is:
T = er

[2]

where T is the final trait value, e is the exponential constant and r is the growth rate that
is the result of the developmental parameters controlling trait size.
The mathematical model is constructed by combining the geometric nature of
morphology with the factors listed above; systemic factors, organ-specific sensitivity to
the systemic factors and organ-autonomous factors. The three factors are included in
the model as: G, the systemic growth factor that affects all tissues, k, the organ-specific
response to the systemic growth factor and i, the organ-autonomous growth factor for
traits. Substituting these growth parameters into equation 2 gives the equation for
generic trait growth, where growth rate is determined by the multiplicative interaction of
the systemic growth factor and the tissue specific response with the additive interaction
of the tissue autonomous growth factor:
T = eGk+i

[3]

81	
  

To produce two traits, T1 and T2, that have covariation in size, trait specific parameters
are substituted into equation 3 to generate an equation for each trait:
T1 = eGk1 + i1
T2 = eGk2 + i2

[4]
[5]

Equations 4 and 5 can be log-transformed to create linear equations for the size of each
trait:
log(T1) = Gk1 + i1
log(T2) = Gk2 + i2

[6]
[7]

and the relationship between the two traits can be summarized as:
log(T2) = k2/k1*log(T1) - k2/k1*i1 + i2

[8]

Equation 8 demonstrates that the slope of the scaling relationship between T1 and T2 is
equivalent to the value of k2/k1, the interaction of the trait specific responses to the
systemic growth factor. Equation 8 also mathematically describes the range of possible
genotype-specific scaling relationships, or the individualSA. As a result of the
importance of the k2/k1 value, the degree to which any form of selection is able to target
the k2/k1 ratio in individuals will likely determine how effective the selection will be in
changing the populationSA.

To create a populationSA, unique parameter values for k1, k2, i1, and i2 were assigned to
each individual. The final value for each parameter was the sum of a diallelic system of
additive co-dominance for each parameter. Each individual has two values (alleles) for
each parameter (k1a and k1b) that combine to form the total parameter value. Parameter
values are all sampled from normal distributions with known means and standard

82	
  

deviations, and from the same distributions within a parameter type (k1, k2, i1, and i2).
The collection of parameter values defines a population of individuals that have no
phenotype, but each have an individualSA. Individuals are then assigned a value of G
which determines where on their individualSA their final phenotype will be expressed.

IndividualSA vs PopulationSA
An important aspect of the model is that different patterns of individualSA can
generate the same populationSA. For example, when µG = 0, the individualSA rotate
around the bivariate mean of the populationSA (Figure 4.2, A, A’). When µG > 0, the
individualSA rotate around a point that is less than the bivariatie mean of the
populationSA, even though the populationSA of the two populations is ostensibly
identical (Figure 4.2 B, B’). When µG < 0, the individualSA rotate around a point that is
greater than the bivariate mean of the population SA, but the populationSA is
indistinguishable from either of the other populationSA (Figure 4.2 C, C’). The
distribution of individualSA patterns will be referred to as ‘bowtie’ (µG = 0),
‘speedometer’ (µG > 0) and ‘broomstick’ (µG < 0), reflecting the general pattern of the
individualSA in each (Figure 4.2, A’, B’, C’).
Selection
Once an initial population of individuals was created, selection was carried out.
Selection is defined in the model as a set number of individuals being removed from the
population (truncation selection). Removing the same proportion of individuals for each
type of selection allows for a direct comparison of the outcome of selection across
generations and forms of selection. The type of selection performed was dependent on

83	
  

what part of the distribution of trait values individuals were removed. After individuals
and their associated parameter values were removed from the population, parameter
values from surviving individuals were recombined to form new individuals and the
population was returned to its original size by assigning new values of G for each new
individual.

84	
  

Figure 4.2.
85	
  

Figure 4.2 (con’t). Patterns of individualSA. (A, A’) When µG = 0, the populationSA
appears normally distributed (A) and the individualSA have equal levels of variation as
the size of T1 increases or decreases (A’) creating a bowtie shaped pattern. (B, B’)
When µG > 0, the populationSA appears similar in distribution to populations with µG = 0,
including similar slope and covariation between traits T1 and T2 (B), but the underlying
individual-level static allometries exhibit more variation as the size of T1 increases (B’)
producing a speedometer shaped pattern. (C, C’) When µG < 0, the population-level
static allometry has the same appearance and statistical properties as populations with
µG > 0 or µG = 0, but the individual-level static allometries exhibit more variation as the
size of T1 decreases (C’) creating a broomstick shaped pattern.

86	
  

Each pattern of individualSA (bowtie, speedometer and broomstick) was subjected to
selection. The parameter values for each population type were selected to maintain
population structure such that each population was essentially identical in terms of
slope, intercept and trait variance. There were four selection regimes that each
population type was exposed to were as follows:
1. Stabilizing selection – individuals were removed based on both the maximum
and minimum extreme values of absolute T2 size (Figure 4.3, A),
2. Directional selection – individuals were removed from either the maximum
(negative directional selection) or minimum (positive directional selection) values
of absolute T2 size (Figure 4.3, B),
3. Proportional selection – individuals were removed based on the relative size of
T2 being larger than T1 (negative proportional selection) or smaller than T1
(positive proportional selection) (Figure 4.3, C) and
4. Correlational selection – individuals were removed based on their having a large
T1 and a proportionally small T2 or small T1 and proportionally large T2 values
(positive correlational selection) or on their having large T1 and a proportionally
large T2 or a small T2 and a proportionally small T2 (negative correlational
selection) (Figure 4.3, D).

Population Structure
Population size for each trial was 1000 individuals, and each population was
allowed to evolve for ten generations. A total of eleven populations were recorded, the
initial population and the subsequent ten generations that were the result of selection.

87	
  

Figure 4.3. Four selection regimes: Stabilizing selection (A), Directional selection (B),
Proportional selection (C) and Correlational selection (D). Grey points indicate selected
individuals red points indicate individuals removed from the population prior to
reproduction. The positive form of selection is shown in B-D with negative selection for
those types removing individuals from the opposite orientation in the population.

88	
  

For each generation, the mean parameter values across the total population were
recorded as were the selection differential and selection intensity for each selection
regime, and the allometric coefficient and intercept of the static allometry for each
population. The allometric coefficient and intercept were calculated using ordinary least
squares (OLS), major axis (MA) and standardized major axis (SMA) linear regression
methods. Only the MA slope and intercept are included here as it represents the closest
estimation of the mathematical allometric coefficient between the two traits, defined
here as k2/k1 that is the interaction of the trait specific responses to the shared systemic
growth factor, G. Selection was replicated 1000 times for each regime and population
type to generate confidence intervals for the population parameter statistics. The
founding populations for each replicate were generated using the same initial parameter
distributions within each population type.
All of the computer modeling was implemented using R (R-Development-Core-Team
2014) and executed using the High Performance Computing Cluster at Michigan State
University.

Results
Visually and statistically indistinguishable populationSA with different underlying
individualSA responded differently to the selection regimes. The general structure of the
populationSA (speedometer (G > 0), bowtie (G = 0), broomstick (G < 0)) was the same;
there were no distinguishing characteristics that clearly delineate which pattern of
individualSA was underlying a given populationSA (Figure 4.2 A-C). Within a population
type, the positive and negative forms of each selection regime had nearly equal and

89	
  

opposite effects on the allometric coefficient and intercept of the population. These
results indicate that predictions of how an allometry will evolve cannot be made based
on the populatoinSA alone.

Stabilizing and correlational selection were both weak forms of selection. Across all
individualSA population types, stabilizing selection had only a minimal effect on the
allometric coefficient and intercept of the populationSA (Figures 4.4 – 4.6: B).
Correlational selection also had a weak effect on the allometric slope in each
individualSA population type (Figures 4.4 – 4.6: E,E’).

For selection types that had a large effect, the pattern of individualSA in the population
determined how the allometry responded to selection. Both directional and proportional
selection readily changed the allometric slope of the populationSA, but only in
speedometer (µG > 0) and broomstick (µG < 0) populations (Figures 4.4 – 4.6: C/C’,
D/D’). Bowtie populations (µG = 0) did not have large changes in allometric slope in
response to any selection type (Figure 4.5).

Figures 4.7 – 4.9 display an initial and final population for all population and selection
types and how the populationSA changed after 10 generations of selection. For each
population type, positive and negative proportional selection increased the correlation
between T1 and T2, while the remaining selection regimes largely preserved the initial
trait correlations (Figure 4.7-4.9: D,D’)

90	
  

Figure 4.4. Evolution of populationSA for ten generations of selection on bowtie populations (µG = 0). Inset plots display
the change in the allometric coefficient (α) through time (generation). (A) The individualSA pattern when µG = 0.
Generations 0 – 10 represented by colored lines blue – red. The effect on the populationSA from stabilizing selection (B),
positive and negative directional selection (C,C’), positive and negative proportional selection (D,D’) and positive and
negative correlational selection (E,E’).
91	
  

Figure 4.5. Evolution of populationSA for ten generations of selection on speedometer populations (µG > 0). Inset plots
display the change in the allometric coefficient (α) through time (generation). (A) The individualSA pattern when µG > 0.
Generations 0 – 10 represented by colored lines blue – red. The effect on the populationSA from stabilizing selection (B),
positive and negative directional selection (C,C’), positive and negative proportional selection (D,D’) and positive and
negative correlational selection (E,E’).

92	
  

Figure 4.6. Evolution of populationSA for ten generations of selection on broomstick populations (µG < 0). Inset plots
display the change in the allometric coefficient (α) through time (generation). (A) The individualSA pattern when µG < 0.
Generations 0 – 10 represented by colored lines blue – red. The effect on the populationSA from stabilizing selection (B),
positive and negative directional selection (C,C’), positive and negative proportional selection (D,D’) and positive and
negative correlational selection (E,E’).

93	
  

Figure 4.7. Initial (blue) and final (red) bowtie populations (µG = 0) after 10 generations of selection. All lines are of the
corresponding colored populationSA calculated using major axis regression. Represented are stabilizing selection (B)
positive and negative directional selection (C, C’) positive and negative proportional selection (D, D’), positive and
negative correlational selection (E, E’).

94	
  

Figure 4.8. Initial (blue) and final (red) speedometer populations (µG > 0) after 10 generations of selection. All lines are of
the corresponding colored populationSA calculated using major axis regression. Represented are stabilizing selection (B)
positive and negative directional selection (C, C’) positive and negative proportional selection (D, D’), positive and
negative correlational selection (E, E’).

95	
  

Figure 4.9. Initial (blue) and final (red) broomstick populations (µG < 0) after 10 generations of selection. All lines are of
the corresponding colored populationSA calculated using major axis regression. Represented are stabilizing selection (B)
positive and negative directional selection (C, C’) positive and negative proportional selection (D, D’), positive and
negative correlational selection (E, E’).

96	
  

Discussion
How scaling relationships evolve is largely dependent on the underlying
pattern of individualSA in a population. The individualSA patterns are determined by the

µG. G represents the environmental condition input as a systemic growth factor, able to
enhance and repress growth. The value of µG represents the degree of control between
positive and negative growth regulators acting on individuals. In populations with a µG >
0, positive growth factors are primarily regulating growth, i.e., growth is being enhanced
and not suppressed. Alternatively, in populations where µG < 0, negative growth factors
are primarily regulating growth and trait growth is being suppressed and not enhanced.
When µG = 0, there is a balance between the control of positive and negative growth
regulators. While we do not know what specific regulators are acting on tissues in this
manner, some systemic positive and negative regulators of growth in the IIS pathway
have been indentified. InR is a positive regulator of growth that responds to increased
levels of nutrition (ILPs) by initiating a signaling cascade and stimulating growth (Chen
et al. 1996, Goberdhan and Wilson 2002). A known negative regulator of growth is the
forkhead transcription factor, FOXO, which represses growth in response to low levels
of nutrition (Kramer et al. 2003). Genetic variation for a trait’s relative sensitivity to
nutrient signals such as InR and FOXO are thought to contribute to the evolution of traitspecific disproportionate growth (Lavine et al. 2015). An additional consideration for the
outcome of selection is that while individuals were selected based on their realized
allometry, their complete individualSA was also being selected. Therefore, possible
outcomes of selection were dependent on much more than the observed populationSA.

96

Stabilizing selection is often invoked as the ultimate cause of hypoallometric traits in
nature, particularly of the male genitalia in arthropods (Eberhard et al. 1998, Eberhard
2009). This prediction is a possible outcome of selection that restricts phenotypic
variation in one trait when compared to a covarying trait. The model does not support
this hypothesis as stabilizing selection had the weakest effect across the three
population types, regardless of the underlying pattern of individualSA (Figures 4.4 – 4.6,
D). What combination(s) of selection and population type did predict a decrease in
allometric coefficient? Both directional and proportional selection had the potential to
decrease the allometric coefficient. In speedometer populations (µG > 0), negative
directional and proportional selection decreased the allometric coefficient, while in
broomstick populations (µG < 0), positive directional and proportional selection
decreased the allometric coefficient. Therefore, one mechanism predicted to result in
hypoallometric relationships is positive directional selection on a trait that is controlled
by a negative growth factor.

Is there evidence for positive directional selection generating hypoallometry in
organisms? Bertin and Fairbairn (2007) describe genital traits that are known to be
under strong positive directional selection in water striders scaling hypoallometrically to
body size. Addtionallly, the hypoallometry of male genitalia in Drosophila has been
shown to be the result of relative insensitivity to FOXO levels in the developing genital
tissues (Tang et al. 2011). Based on the model, we know that positive directional
selection in a broomstick population is predicted to result in hypoallometry. We also
know that broomstick populations are the result of growth control by negative regulators

97

of growth, such as FOXO. Taken together, these two pieces of empirical evidence
support the evolution of hypoallometry as presented by our model.

Our model also predicts that the weakest forms of selection in relation to the allometric
scaling relationship are correlational and stabilizing selection (Figures 4.4 – 4.6).
Artificial selection targeting the allometric slope of scaling relationships has proven
difficult empirically (Egset et al. 2012), although given the type of artificial selection
being performed, our model predicts a poor response. The type of selection used by
Egset et al. (2012) most closely resembled correlational selection in our model, which
we found to be a weak form of selection. Intriguingly, other instances of artificial
selection targeting trait size rather that the slope of the scaling relationship resulted in
changes in the allometric coefficient (Wilkinson 1993, Emlen 1996, Nijhout and German
2012). The ability of directional selection to change the allometric coefficients in these
studies is also in congruence with our model in that the model predicts directional
selection to change the slope of an allometry rapidly as compared to either stabilizing or
correlational selection.

As shown in Figure 2 (A – C) the initial population types between the patterns of
individualSA are nearly indistinguishable. How then are the underlying individualSA of a
population to be indentified to for accurate predictions of how the population will
respond to selection? To predict how allometries will evolve, it is clear that some
understanding of the individualSA is necessary. One method of determining what the
pattern of indivdualSA is would be rearing genetically identical individuals at various

98

environmental conditions to map out the static allometry for a single genotype
empirically. Mapping of individualSA in this way is only theoretically possible in a select
few model systems such as Drosophila where isogenic lines are available. For
genetically distinct individuals, rearing a population under a series of tightly-controlled
environmental conditions would create a distribution of genetic static allometries for
those conditions. By comparing the variance of the sample distributions at each
environmental condition, the type of individualSA underlying the populationSA could be
estimated. The type of growth factor that is controlling trait growth is also an important
component of the individualSA pattern. Elucidating how trait growth is actively being
regulated in organisms should also help generate predictions of how traits will respond
to a given form of selection.

The number of hypotheses for how allometries have evolved and are expected to
evolve is a reflection of the level of interest in understanding the importance of relative
trait size on morphological diversification. Our model predicted responses to selection
based on the novel recognition that individuals in a population contribute the entire
range of static allometries their genotype could express to a population, not just their
realized phenotype. As a result, the patterns of variation in the hidden individualSA
determine how a populationSA will evolve. The underlying individualSA pattern that is
masked in the expression of the populationSA also explains how populations that
appear to have the same scaling relationships respond differently to the same forms of
selection. The effects of size regulation mechanisms such as IIS in determining how
traits respond to environmental variation reiterates how important research on such

99

mechanisms of trait size control will continue to be for allometry research, and for the
study of all morphological diversity.

100

CHAPTER 5:
Conclusion and future directions

101

Summary of Chapters
Chapter 1
!

Morphological variation is largely the result of changes in the relative size of traits
to one another

!

Scaling relationships between traits, called allometries, are a measure of the covariation between traits in response to factors regulating trait size.

!

The field of allometry research is focused on describing the proportional size of
traits across taxa (evolutionary allometries), within species at different
developmental stages (ontogenetic allometries) and within species at the same
developmental stage (static allometries).

!

Sources of variation that contribute to changes in allometric relationships include
genetic and environmental factors.

!

How allometries evolve is an ongoing question and the inclusion of trait
development in these predictions is an important step forward for allometry
research.

Chapter 2
!

Across many taxa, male genital size scales hypoallometrically to body size,
meaning, the genitalia remain a constant size across a range of body sizes. The
genitalia of male Drosophila melanogaster follow this pattern as well.

!

The most frequent hypothesis to explain the hypoallometry of genitalia is
stabilizing selection acting on genital size.

102

!

Size variation may be the result of environmental variation, genetic variation and
developmental instability (variation due to stochastic developmental perturbations
within an individual).

!

The genitalia of male Drosophila show relatively low responses to variation in
environmental and genetic factors that affect all trait size. The genitalia do not
show low levels of variation in genetic factors that act independently of other
traits or of developmental instability.

!

Increased levels of developmental instability of the genitalia in conjunction with
decreased responses to genetic size control, suggest that stabilizing selection is
not acting on the genitalia in Drosophila. If stabilizing selection were acting on
genital size, low levels of response to organ autonomous size control
mechanisms and of developmental instability would be expected.

!

To determine how scaling relationships evolve, the focus needs to be on how the
traits of interest interact with one another and not on one of the pair.

Chapter 3
!

Testing for selection on genitalia in vivo, there must be necessary levels of size
variation in the trait being studied.

!

Using the genetic tools in Drosophila melanogaster, variation in genital size that
extends beyond the norm was created by perturbations of the insulin/insulin
growth factor pathway (IIS). Three genital size classes of males were
engineered: small, wild-type and large, but only the small class of flies were
significantly different in size.

103

!

Males of different size classes were exposed to females in three mating contexts:
no competition, direct male-male competition and indirect male-male competition.

!

Females across all three mating contexts preferentially copulated with, and
fertilized progeny from males that had larger genital sizes. Females did not
differentiate between male genital size in the number of progeny produced.

!

Despite being able to manufacture Drosophila with larger and smaller wings by
increasing and decreasing IIS signaling in the developing wing tissues, genitalia
only decreased in size.

!

The lack of a response to an increase in IIS signal suggests that the genitalia
may be less sensitive to increased nutrition or at a maximum size in response to
nutritional variation.

Chapter 4
! To further explore allometry evolution, a mathematical model was created to
incorporate trait development into the co-variation between two traits, T1 and T2,
and then used to construct a population-level static allometry.
! The model used three classes of growth factors as parameters to create two
traits: a systemic growth factor that affects all tissues (G), a trait-specific
response to the systemic factor (k1 and k2) and a trait autonomous growth factor
(i1 and i2).
! Parameter values to create each individual were randomly selected from given
distributions to create individuals.

104

! Based on the parameter values that define an individual (k1, k2, i1, i2), each
individual had a unique range of possible responses to a range of G values,
called the individual static allometry (individualSA) of which only one point was
actually expressed per individual.
! Three patterns of individualSA were identified based on the balance of regulatory
control over trait size between positive and negative growth regulators.
! Populations of individuals with a T1 and T2 were generated and exposed to four
selection regimes: direct, proportional, correlational and stabilizing selection.
! The effects of selection on a populationSA are largely dependent on the
underlying pattern of the individualSA. To predict how allometries will evolve,
the mechanisms controlling trait size, and consequently the individualSA, should
be identified.

Conclusion
Genital Hypoallometry in Drosophila
The male genitalia in Drosophila are an ideal system in which to study the
evolution of hypoallometry. Genetic manipulations, short generation times and the ability
to run experiments in a controlled laboratory setting have each contributed a thorough
test of the causes of hypoallometry in Drosophila. With over a decade of general
support for stabilizing selection driving hypoalloetric scaling relationships (Eberhard et
al. 1998 2009, Eberhard 2009) recent evidence has suggested that stabilizing selection
may not be the de-facto cause (Bonduriansky and Day 2003). Results of experiments
on the male genitalia of Drosophila melanogaster have also failed to support the

105

hypothesis of stabilizing selection. Male genitalia are more developmentally unstable
than somatic traits and are equal response to variation in genetic factors that affect trait
size as somatic traits (Dreyer and Shingleton 2011). Additionally, in Chapter 3 of this
dissertation, female flies were demonstrated to prefer absolute larger genital sizes of
males, both to copulate with and to fertilize progeny with. An inability to increase genital
size in response to a simulated increase in nutrient signaling in the genitalia suggested
genital tissues may be insensitive to positive growth factors in addition to being
insensitive to negative growth regulators (Tang et al. 2011).

The evidence of these experiments remained equivocal as to the nature of the
hypoallometry of the genitalia, but indicated that directional selection may be the
ultimate cause rather than stabilizing selection. To further test this hypothesis, a
mathematical model of population scaling relationship evolution was created. The
results of the model simulations indicate that positive directional selection on a trait that
is regulated primarily by a negative growth regulator will result in a hypoallometric trait.
Collectively, the evidence presented here clearly points to positive directional selection
as the ultimate cause of genital hypoallometry in male Drosophila.

Future Directions
The results of the mathematical model on the possible responses in allometries
to different forms of selection outline a compelling new focus of allometry research.
Effects of the hidden individual-level scaling relationships (how a genotype encoding
many possible outcomes interacts with its environment to become a single phenotype)

106

present how much we have yet to learn about the interplay of trait regulation and
morphological evolution. Specifically, the recognition that organisms are but one
iteration of all possible ‘versions’ their genotype allows. Selection on a population of
individuals, each with their own individualSA, is also selecting on the range of possible
phenotypes each individual could have had. While this is not a revolutionary concept in
and of itself, the degree to which the collective patterns of indivdualSA determine the
evolution of covarying traits is a novel addition to what we currently know about
allometry evolution. Future studies should focus on the identifying specific types of
control mechanisms underlying a trait’s development. How trait size is regulated, be it
predominantly through positive or negative control factors, determines the pattern of
individualSA and subsequently how the population will respond to selection.

Research to identify growth control mechanisms is already rapidly expanding (Mirth and
Shingleton 2012, Shingleton 2012, Nijhout et al. 2013, Lavine et al. 2015). While insects
tend to be the study systems most amenable to research on growth, there are many asyet unexplored opportunities to investigate morphological evolution. If the evolution of
morphology is largely the evolution of allometry, it is a very exciting time to be invested
in the study of scaling relationships.

107

REFERENCES

108

REFERENCES
Arendt, J., and D. Fairbairn. 2012. Reproductive allometry does not explain the
temperature-size rule in water striders (Aquarius remigis). Evolutionary Ecology
26:745-757.
Arnold, E. 1973. Relationships of the palaearctic lizards assigned to the genera Lacerta,
Algyroides and Psammordromus (Reptilia: Lacertidae). Bulletin of the British
Museum of Natural History (Zoology) 25:291-366.
Arnqvist, G. 1997. The evolution of animal genitalia: distinguishing between hypotheses
by single species studies. Biological Journal of the Linnean Society 60:365-379.
Arnqvist, G., and L. Rowe. 2005. Sexual Conflict. Princeton University Press, Princeton,
New Jersey, USA.
Arnqvist, G., and R. Thornhill. 1998. Evolution of animal genitalia: patterns of
phenotypic and genotypic variation and condition dependence of genital and nongenital morphology in water strider (Heteroptera : Gerridae : Insecta). Genetics
Research 71:193-212.
Berger, D., M. Olofsson, M. Friberg, B. Karlsson, C. Wiklund, and K. Gotthard. 2012.
Intraspecific variation in body size and the rate of reproduction in female insectsadaptive allometry or biophysical constraint? Journal Of Animal Ecology 81:12441258.
Bernstein, S., and R. Bernstein. 2002. Allometry of male genitalia in a species of soldier
beetle: support for the one-size-fits-all hypothesis. Evolution 56:1707-1710.
Bertin, A., and D. J. Fairbairn. 2007. The form of sexual selection on male genitalia
cannot be inferred from within-population variance and allometry - a case study
in Aquarius remigis. Evolution 61:825-837.
Birkhead, T. R. 1998. Cryptic female choice: criteria for establishing female sperm
choice. Evolution 52:1212-1218.
Birkhead, T. R., and T. Pizzari. 2002. Postcopulatory sexual selection. Nature Reviews
Genetics 3:262-273.

109

Bohni, R., J. Riesgo-Escovar, S. Oldham, W. Brogiolo, H. Stocker, B. F. Andruss, K.
Beckingham, and E. Hafen. 1999. Autonomous control of cell and organ size by
CHICO, a Drosophila homolog of vertebrate IRS1-4. Cell 97:865-875.
Boll, W., and M. Noll. 2002. The Drosophila Pox neuro gene: control of male courtship
behavior and fertility as revealed by a complete dissection of all enhancers.
Development 129:5667.
Bonduriansky, R. 2007. Sexual selection and allometry: a critical reappraisal of the
evidence and ideas. Evolution 61:838-849.
Bonduriansky, R., and T. Day. 2003. The evolution of static allometry in sexually
selected traits. Evolution 57:2450-2458.
Brogiolo, W., H. Stocker, T. Ikeya, F. Rintelen, R. Fernandez, and E. Hafen. 2001. An
evolutionarily conserved function of the Drosophila insulin receptor and insulinlike peptides in growth control. Curr Biol 11:213-221.
Brown, J. H., and G. B. West, editors. 2000. Scaling in Biology: Patterns and
Processes, Causes and Consequences. Oxford University Press US, New York,
New York, USA.
Bulmer, M. G. 1976. The effect of selection on genetic variability: a simulation study.
Genetical Research 28:101-117.
Calder, W. A. 1984. Size, Function and Life History. Harvard University Press,
Cambridge, Massachusetts, USA.
Cayetano, L., A. Maklakov, R. Brookcs, and R. Bonduriansky. 2011. Evolution of male
and female genitalia following release from sexual selection. Evolution 65:21712183.
Chari, S., and I. Dworkin. 2013. The conditional nature of genetic interactions: the
consequences of wild-type backgrounds on mutational interactions in a genomewide modifier screen. PLoS Genetics 9:e1003661.
Chave, J., C. Andalo, S. Brown, M. Cairns, J. Chambers, D. Eamus, H. Folster, F.
Fromard, N. Higuchi, T. Kira, J.-P. Lescure, B. Nelson, H. Ogawa, H. Puig, B.

110

Riera, and T. Yamakura. 2005. Tree allometry and improved estimation of carbon
stocks and balance in tropical forests. Oecologia 145:87-99.
Chen, C., J. Jack, and R. S. Garofalo. 1996. The Drosophila insulin receptor is required
for normal growth. Endocrinology 137:846-856.
Cheverud, J. M. 1982. Relationships among ontogenetic, static, and evolutionary
allometry. Am J Phys Anthropol 59:139-149.
Chiang, H. C., and A. C. Hodson. 1950. An analytical study of population growth in
Drosophila melanogaster. Ecol Monogr 20:173-206.
Claeys, I., G. Simonet, J. Poels, T. Van Loy, L. Vercammen, A. De Loof, and J. Vanden
Broeck. 2002. Insulin-related peptides and their conserved signal transduction
pathway. Peptides 23:807-816.
Cock, A. G. 1966. Genetical Aspects of Metrical Growth and Form in Animals. Q Rev
Biol 41:131-190.
Cook, R. M. 1977. Behavioral Role of Sexcombs in Drosophila-Melanogaster and
Drosophila-Simulans. Behavior Genetics 7:349-357.
Crnokrak, P., and D. A. Roff. 1995. Dominance variance: associations with selection
and fitness. Heredity 75:530-540.
Debat, V. 2001. Mapping phenotypes: canalization, plasticity and developmental
stability. Trends In Ecology &amp; Evolution.
Dixon, D. 1993. Dougal Dixon's Dinosaurs. Bell Books, Spain.
Dreyer, A., and A. W. Shingleton. 2011. The effect of genetic and environmental
variation on genital size in male Drosophila: canalized but developmentally
unstable. PLoS One 6:e28278.
Duffy, J. B. 2002. GAL4 system in Drosophila: a fly geneticist's Swiss army knife.
Genesis 34:1-15.

111

Dufour, L. 1844. Anatomie Generale des Dipteres. Annales des Sciences Naturelles
1:244-264.
Dufour, L. 1936. Anatomie générale des Diptéres. Annuaire de Science Naturelle 1:244264.
Dupont, J., and M. Holzenberger. 2003. Biology of insulin-like growth factors in
development. Birth Defects Res C Embryo Today 69:257-271.
Eberhard, W., R. L. Rodriguez, and M. Polihronakis. 2009. Pitfalls in understanding the
functional significance of genital allometry. Journal of Evolutionary Biology
22:435-445.
Eberhard, W. G. 1985. Sexual Selection and Animal Genitalia. Harvard University
Press, Cambridge.
Eberhard, W. G. 1996. Female Control: Sexual Selection by Cryptic Female Choice.
Princeton University Press, Princeton, New Jersey, USA.
Eberhard, W. G. 2009. Static allometry and animal genitalia Evolution 63:48-66.
Eberhard, W. G., B. A. Huber, and R. L. Rodriguez. 1999. Don't forget the biology: a
reply to Green. Evolution 53:1624-1627.
Eberhard, W. G., B. A. Huber, R. L. Rodriguez, R. D. Briceno, I. Salas, and V.
Rodriquez. 1998. One size fits all? Relationships between the size and degree of
variation in genitalia and other body parts in twenty species of insects and
spiders. Evolution 52:415-431.
Egset, C., T. Hansen, A. Le Rouzic, G. Bolstad, G. Rosenqvist, and C. Pelabon. 2012.
Artificial selection on allometry: change in elevation but not slope. Journal of
Evolutionary Biology 25:938-948.
Elliot, D., and A. Brand. 2008. The GAL4 system: a versitile system for the expression
of genes.in C. Dahmann, editor. Drosophila: Methods and Protocols. Humana
Press, Totowa, New Jersey, USA.

112

Emlen, D. J. 1996. Artificial selection on horn length-body size allometry in the horned
beetle Onthophagus acuminatus. Evolution 50:1219-1230.
Emlen, D. J., and C. E. Allen. 2003. Genotype to phenotype: Physiological control of
trait size and scaling in insects. Integrative and Comparative Biology 43:617-634.
Emlen, D. J., and H. F. Nijhout. 2000. The development and evolution of exaggerated
morphologies in insects. Annual Review Of Entomology 45:661-708.
Emlen, D. J., Q. Szafran, L. S. Corley, and I. Dworkin. 2006. Insulin signaling and limbpatterning: candidate pathways for the origin and evolutionary diversification of
beetle 'horns'. Heredity 97:179-191.
Emlen, D. J., I. A. Warren, A. Johns, I. Dworkin, and L. Lavine. 2012. A mechanism of
extreme growth and reliable signaling in sexually selected ornaments and
weapons. Science 337:860-864.
Eshel, I., and C. Matessi. 1998. Canalization, genetic assimilation and preadaptation: A
quantitative genetic model. Genetics 149:2119-2133.
Falconer, D. S., and T. F. Mackay. 1996. Introduction to Quantitave Genetics. 3 edition.
The Ronald Press Company, New York.
Fromhage, L., and H. Kokko. 2014. Sexually selected traits evolve positive allometry
when some matings occur irrespective of the trait. Evolution 68:1332-1338.
Funke, S., and B. A. Huber. 2005. Allometry of genitalia and fighting structures in
Linyphia triangularis (Araneae, Linyphiidae). Journal of Arachnology 33:870-872.
Gailey, D., F. Jackson, and R. Siegel. 1982. Male courtship in Drosophila: the
conditiond response to immature males and its genetic control. Genetics
102:771-782.
Gavrilets, S., and A. Hastings. 1994. A Quantitative-Genetic Model for Selection on
Developmental Noise. Evolution 48:1478-1486.
Gayon, J. 2000. History of the concept of allometry. American Zoologist 40:748-758.

113

Gibson, G., and G. Wagner. 2000. Canalization in evolutionary genetics: a stabilizing
theory? BioEssays 22:372-380.
Goberdhan, D. C., and C. Wilson. 2002. Insulin receptor-mediated organ overgrowth in
Drosophila is not restricted by body size. Dev Genes Evol 212:196-202.
Gotoh, H., H. Miyakawa, A. Ishikawa, Y. Ishikawa, Y. Sugime, D. Emlen, L. Lavine, and
T. Miura. 2014. Developmental link between sex and nutrition; doublesex
regulates sex-specific mandible growth via juvenile hormone signaling in stag
beetles. PLoS Genetics 10:e1004098.
Gould, S. J. 1966. Allometry and size in ontogeny and phylogeny. Biological Reviews
41:587-638.
Green, A. J. 1992. Positive Allometry Is Likely with Mate Choice, Competitive Display
and Other Functions. Animal Behaviour 43:170-172.
Green, A. J. 1999. Allometry of genitalia in insects and spiders: one size does not fit all.
Evolution 53:1621-1624.
Grewal, S. 2009. Insulin/TOR signaling in growth and homeostasis: a view from the fly
world. The International Journal of Biochemistry and Cell Biology 41:1006-1010.
Hadfield, J. 2010. MCMC Methods for Multi-Response Generalized Linear Mixed
Models: The MCMCglmm R Package. 33:1-22.
Harvey, P. H., and M. D. Pagel. 1991. The Comparative method in evolutionary biology.
Osford University Press, Oxford.
Hosken, D. J., A. M. Minder, and P. I. Ward. 2005. Male genital allometry in
Scathophagidae (Diptera). Evolutionary Ecology 19:501-505.
Hosken, D. J., and P. Stockley. 2004. Sexual selection and genital evolution. Trends in
Ecology & Evolution 19:87-93.
Huber, B. A. 2003. Rapid evolution and species-specificity of arthropod genitalia: fact or
artifact? Organisms Diversity and Evolution 3:63-71.

114

Huxley, J. S. 1932. Problems of Relative Growth. Methuen & Co. Ltd., London.
Huxley, J. S., and G. Teissier. 1936. Terminology of relative growth. Nature 137:780781.
Kaufman, P. K., F. D. Enfield, and R. E. Comstock. 1977. Stabilizing Selection for Pupa
Weight in Tribolium-Castaneum. Genetics 87:327-341.
Kawano, K. 2006. Sexual dimorphism and the making of oversized male characters in
beetles (Coleoptera). Annals of the Entomological Society of America 99:327341.
Kerkhoff, A. J., and B. J. Enquist. 2009. Multiplicative by nature: why logarithmic
transformatoin is necessary in allometry. Journal of Theoretical Biology 257:519521.
Kijimoto, T., A. P. Moczek, and J. Andrews. 2012. Diversification of doublesex function
underlies morph-, sex- and species-specific development of beetle horns.
Proceeding of the National Acadamy of Science, USA 109:20526-20531.
Klingeberg, C. P. 1998. Heterochrony and allometry: the analysis of evolutionary
change in ontogeny. Pages 79-123 Biological Reviews of the Cambridge
Philosophical Society.
Klingenberg, C. P. 1996. Multivariate allometry. Pages 23-48 in L. F. Marcus, editor.
Advances in Morphometrics. Plenum Press, New York.
Klingenberg, C. P., and J. Marugan-Lobon. 2013. Evolutionary covariation in geometric
morphometric data: analyzing integration, modularity, and allometry in a
phylogenetic context. Systematic Biology 62:591-610.
Klowden, M. 2007. Physiological Systems in Insects. Elsevier, Oxford, United Kingdom.
Kodric-Brown, A., R. Sibly, and J. h. Brown. 2006. The allometry of ornaments and
weapons. Proceeding of the National Acadamy of Science, USA 103:8733-8738.

115

Koushika, S., M. Lisbin, and K. White. 1996. ELAV, a Drosophila neuron-specific
protein, mediates the generation of an alternatively spliced neural protein
isoform. Current Biology 6:1634-1641.
Koyama, T., C. Mendes, and C. K. Mirth. 2013. Mechanisms regulating nutritiondependent developmental plasticity through organ-specific effects in insects.
Frontiers in Physiology 4:1-12.
Kraaijeveld, K., F. Kraaijeveld-Smit, and M. Mann. 2011. Sexual selection and
speciation: the comparative evidence revisited. Biological Reviews 86:367-677.
Kramer, J., J. Davidge, J. Lockyer, and B. Staveley. 2003. Expression of Drosophila
FOXO regulates growth and can phenocopy starvation. BMC Developmental
Biology 3:5.
Lande, R. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain:
body size allometry. Evolution 33:402-416.
Lasbleiz, C., J. Ferveur, and C. Everaerts. 2006. Courtship behavior of Drosophila
melanogaster revisited. Animal Behavior 72:1001-1012.
Lavine, L., H. Gotoh, C. Brent, I. Dworkin, and D. Emlen. 2015. Exaggerated Trait
Growth in Insects. Annual Review Of Entomology 60.
Lemaitre, J., C. Vanpe, F. Plard, and J. Gaillard. 2014. The allometry between
secondary sexual traits and body size is nonlinear among cervids. Biology
Letters 10.
Maina, J., A. King, and G. Settle. 1989. An allometric study of morphometric parameters
in birds, with mammalian comparisons. Philosophical Transactions of the Royal
Society of London Series B - Biological Sciences 326:1-57.
Mayr, E. 1963. Animal Species and Evolution. Harvard University Press, Cambridge.
McMahon, T., and J. Bonner. 1983. On Size and LIfe. Scientific American LIbrary, New
York, New York, USA.

116

Merila, J., and B. C. Sheldon. 1999. Genetic architecture of fitness and nonfitness traits:
empirical patterns and development of ideas. Heredity 83:103-109.
Mirth, C. K., and L. M. Riddiford. 2007. Size assessment and growth control: how adult
size is determined in insects. BioEssays 29:344-355.
Mirth, C. K., and A. W. Shingleton. 2012. Integrating body and organ size in Drosophila:
recent advances and outstanding problems. Frontiers in Endocrinology 3:1-13.
Mutanen, M., A. Kaitala, and M. Monkkonen. 2006. Genital variation within and between
three closely related Euxoa moth species: testing the lock-and-key hypothesis.
Journal of Zoology 268:109-119.
Ng, C. S., and A. Kopp. 2008. Sex combs are important for male mating success in
Drosophila melanogaster. Behavior Genetics 38:195-201.
Nijhout, H. F. 2003. The control of body size in insects. Dev Biol 261:1-9.
Nijhout, H. F., and R. German. 2012. Developmental causes of allometry: new models
and implications for phenotypic plasticity and evolution. Integrative and
Comparative Biology 52:43-52.
Nijhout, H. F., L. M. Riddiford, C. K. Mirth, A. W. Shingleton, Y. Suzuki, and V. Callier.
2013. The developmental control of size in insects. WIREs Developmental
Biology 3:113-134.
Niklas, K. J. 1994. Plant Allometry. University of Chicago, Chicago.
O'Dell, K. 2003. The voyeurs' guide to Drosophila melanogaster courtship. Behavioural
Processes 64:211-223.
Oldham, S., and E. Hafen. 2003. Insulin/IGF and target of rapamycin signaling: a TOR
de force in growth control. Trends in Cell Biology 13:79-85.
Oosthuizen, W. H., and E. H. Miller. 2000. Bacular and testicular growth and allometry
in the cape fur seal Arcocephalus p. pusillus (Otariidae). Marine Mammal
Science 16:124-140.

117

Palestrini, C., A. Rolando, and P. Laiolo. 2000. Allometric relationships and character
evolution in Onthophagus taurus (Coleoptera: Scarabaeidae). Canadian Journal
of Zoology 78:1199-1206.
Palmer, A. R., and C. Strobeck. 1986. Fluctuating asymmetry: measurement, analysis,
patterns. Annual Review of Ecology and Systematics 17:391-421.
Palmer, A. R., and C. Strobeck. 2003. Fluctuating asymmetry analysis revisited.in M.
Polak, editor. Devolpental Instability: Causes and Consequences. Oxford
University Press, Oxford.
Parsons, P. A. 1990. Fluctuating asymmetry: an epigenetic measure of stress.
Biological Reviews 65:131-145.
Patterson, B. 1982. The mammalian baculum: hypotheses on the nature of bacular
variability. Journal of Mammalogy 63:1-15.
Pelabon, C., C. Firmat, G. Bolstad, K. Voje, D. Houle, J. Cassara, A. Le Rouzic, and T.
Hansen. 2014. Evolution of morphological allometry. Annals of the New York
Academy of Sciences 1320:58-75.
Peters, R. 1983. The Ecological Impliactions of Body Size. Cambridge University Press,
New York, New York, USA.
R-Development-Core-Team. 2009. R: A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing, Vienna, Austria.
R-Development-Core-Team. 2014. R: A language and environment for statistical
computing. R Foundation for Statistical Computing.
Scharloo, W., M. S. Hoogmoed, and A. T. Kuile. 1967. Stabilizing and disruptive
selection on a mutant character in Drosophila. I. The phenotypic variance and its
components. Genetics 56:709-726.
Schmidt-Nielsen, K. 1984. Scaling: Why is Animal Size so Important? Cambridge
University Press, New York, New York, USA.

118

Shapiro, A. M., and A. Porter. 1989. The lock-and-key hypothesis: evolutionary and
biosystematic interpretation of insect genitalia. Annual Review Of Entomology
34:231-245.
Shingleton, A. W. 2010. The regulation of organ size in Drosophila: Physiology,
plasticity, patterning and physical force. Organogenesis 6:1-12.
Shingleton, A. W. 2012. New perspectives on the evolution of exaggerated traits.
BioEssays 35:100-107.
Shingleton, A. W., C.M. Estep, M. V. Driscoll, and I. Dworkin. 2009. Many ways to be
small: different environmental regulators of size generate distinct scaling
relationships in Drosophila melanogaster. Proceedings of the Royal Society B:
Biological Science 276:2625-2633.
Shingleton, A. W., J. Das, L. Vinicius, and D. L. Stern. 2005. The temporal requirements
for insulin signaling during development in Drosophila. PLoS Biology 3:e289.
Shingleton, A. W., W. A. Frankino, T. Flatt, H. F. Nijhout, and D. J. Emlen. 2007. Size
and shape: the developmental regulation of static allometry in insects. BioEssays
29:536-548.
Shingleton, A. W., C. K. Mirth, and P. W. Bates. 2008. Developmental model of static
allometry in holometabolous insects. Proceedings of the Royal Society BBiological Sciences 275.
Siegal, M. L., and A. Bergman. 2002. Waddington's canalization revisited:
developmental stability and evolution. Proceedings of the National Academy of
Sciences of the United States of America 99:10528-10532.
Siegel, R., and J. Hall. 1979. Conditioned responses in courship behavior of normal and
mutant Drosophila. Proceeding of the National Acadamy of Science, USA
76:3430-3434.
Simmons, L. W. 2001. Sperm Competition and its Evolutionary Consequences in the
Insects. Princeton University Press, Princeton, New Jersey, USA.
Snook, R. 2005. Sperm in competition: not playing by the numbers. Trends in Ecology &
Evolution 20:46-53.

119

Spieth, H. R. 1995. Change In Photoperiodic Sensitivity During Larval Development Of
Pieris-Brassicae. Journal Of Insect Physiology 41:77-83.
Stearns, S. C., M. Kaiser, and T. J. Kawecki. 1995. The differential genetic and
environmental canalization of fitness components in Drosophila melanogaster.
Journal of Evolutionary Biology 8:539-557.
Stern, D. L., and D. J. Emlen. 1999. The developmental basis for allometry in insects.
Development 126:1091-1101.
Stocker, H., and P. Gallant. 2008. An overview on raising and handling Drosophila.in C.
Dahmann, editor. Drosophila: Methods adn Protocols. Humana Press, Totowa,
New Jersey, USA.
Sutter, N. B., C. D. Bustamante, K. Chase, M. M. Gray, K. Zhao, L. Zhu, B.
Padhukasahasram, E. Karlins, S. Davis, P. G. Jones, P. Quignon, G. S. Johnson,
H. G. Parker, N. Fretwell, D. S. Mosher, D. F. Lawler, E. Satyaraj, M. Nordborg,
K. G. Lark, R. K. Wayne, and E. A. Ostrander. 2007. A single IGF1 allele is a
major determinant of small size in dogs. Science 316:112-115.
Tang, H., M. Smith-Caldas, M. V. Driscoll, S. Salhadar, and A. W. Shingleton. 2011.
FOXO regulates organ-specific phenotypic plasticity in Drosophila. PLoS
Genetics 7:e1002373.
Tatsuta, H., K. Fujimoto, K. Mizota, K. Reinhardt, and S. Akimoto. 2007. Distinctive
developmental variability of genital parts in the sexually dimorphic beetle,
Prosopocoilus inclinatus (Coleoptera: Lucanidae). Biological Journal of the
Linnean Society 90:573-581.
Thompson, D. 1942. On Growth and Form. Cambridge University Press, Cambridge,
United Kingdom.
Uhl, G., and F. Vollrath. 2000. Extreme body size variability in the golden silk spider
(Nephila edulis) does not extend to genitalia. Journal of Zoology 251:7-14.
Van Valen, L. 1962. A study of fluctuating asymmetry. Evolution 16:125-142.

120

Verdu, J., M. A. Buratovich, E. L. Wilder, and M. J. Birnbaum. 1999. Cell-autonomous
regulation of cell and organ growth in Drosophila by Akt/PKB. Nat Cell Biol 1:500506.
Waddington, C. H. 1942. Canalization of development and the inheritance of acquired
characters. Nature 150:563-565.
Waddington, C. H. 1961. Genetic Assimilation. Academic Press Inc., New York.
Wagner, G. P. 1996. Homologues, natural kinds and the evolution of modularity.
American Zoologist 36:36-43.
Wagner, G. P., and V. Lynch. 2010. Evolutionary Novelties. Current Biology 20:48-52.
White, C., and R. Seymour. 2005. Allometric scaling of mammalian metabolism. Journal
Of Experimental Biology 208:1611-1619.
Wilkinson, G. S. 1993. Artificial Sexual Selection Alters Allometry in the Stalk-Eyed Fly
Cyrtodiopsis-Dalmanni (Diptera, Diopsidae). Genet Res 62:213-222.
Wright, S. 1952. The genetics of quantitative variability.in C. R. Reeve and C. H.
Waddington, editors. Quanititative Inheritance. Her Majesty’s Stationery Office,
London

121