Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Genetic architectures of phenotypic capacitance
(USC Thesis Other)
Genetic architectures of phenotypic capacitance
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
GENETIC ARCHITECTURES OF PHENOTYPIC CAPACITANCE
by
Matthew Bryce Taylor
______________________________________________________________________
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(MOLECULAR BIOLOGY)
May 2016
Copyright 2016 Matthew Bryce Taylor
ii
Dedication
To my parents.
iii
Acknowledgements
I’ve been lucky to have some phenomenal support along my way to earning my PhD. My
Advisor Ian, my mentors, friends, and most importantly my parents Patricia and David
have helped push, prod, and pull me to the point I’m at right now. I can’t possibly list
everyone who has influenced and helped me, which only highlights the embarrassment of
riches in my personal life. Thanks everyone for guiding me to realize my potential as a
scientist and a human.
iv
Table of Contents
Dedication
Acknowledgements
List of Tables
List of Figures
Abstract
Chapter 1: Introduction
1.1 Cryptic genetic variation
1.2 Capacitance
1.3 Rough morphology system
1.4 Goals of this thesis
1.5 Summary of chapters
Chapter 2: Genetic interactions involving five or more genes
contribute to a complex trait in yeast
2.1 Overview
2.2 Introduction
2.3 Identification of a discrete complex trait
2.4 Genetic mapping of rough morphology
2.5 Cloning of causal genes
2.6 IRA2 contains a mutation required for rough
morphology
2.7 Candidate causal polymorphisms
2.8 Description of causal genes
2.9 Only two discrete genotypes can exhibit rough
morphology
2.10 Discussion
2.11 Materials and methods
2.11.1 Phenotyping of yeast colony morphology
2.11.2 Assessing potential effects of transient
heritable factors
2.11.3 Generation of backcross segregants
2.11.4 Genome sequencing of backcross
segregants
2.11.5 Genetic mapping
2.11.6 Generation and genotyping of dissected
tetrads
2.11.7 Fine-mapping using multi-locus
introgression strains and dissected backcross
segregants
2.11.8 Genetic engineering experiments
ix
viii
iii
ii
1
2
2
3
4
5
6
7
7
7
9
14
16
22
23
24
25
33
34
34
34
35
35
36
36
36
37
v
Table of Contents
2.11.9 Identification of loci that complement the
3S allele of END3
Chapter 3: Transcriptional derepression uncovers cryptic higher-
order genetic interactions
3.1 Overview
3.2 Introduction
3.3 END3
3S
and ira2∆2933 are involved in a six-way
genetic interaction
3.4 The two interactions fully account for rough
morphology in the presence of ira2∆2933
3.5 FLO11 expression is needed for rough morphology
3.6 ira2∆2933 and SFL1 deletion cause FLO11
expression in 3S
3.7 Cryptic genetic variation uncovered by SFL1
deletion
3.8 Discussion
3.9 Materials and Methods
3.9.1 Phenotyping of yeast colony morphology
3.9.2 Generation of backcross segregants
3.9.3 Generation of IRA2 wild type, ira2∆2933,
and sfl1∆ crosses
3.9.4 Bulk segregant mapping of rough
morphology in the backcrosses
3.9.5 Genetic engineering experiments
3.9.6 Genotyping of causal alleles in ira2∆2933
and sfl1∆ crosses
3.9.7 RT-PCRs
Chapter 4: Genetic architectures of phenotypic capacitance
4.1 Overview
4.2 Introduction
4.3 Experimental design
4.4 Diverse architectures underlying capacitance
4.5 Characterization of identified capacitating mutations
4.6 Instances of capacitance in a wild type background
4.7 Most cryptic variants are in Ras pathway
components
4.8 Discussion
4.9 Materials and methods
4.9.1 Phenotyping of yeast colony morphology
4.9.2 Generation of rough segregants
4.9.3 Generation of backcross segregants
4.9.4 Sequencing of mapping populations
4.9.5 Genetic mapping using Multipool
38
39
39
40
45
47
48
49
50
51
53
53
53
54
55
56
57
57
58
58
58
60
62
64
67
69
70
72
72
72
73
74
74
vi
Table of Contents
4.9.6 Identification of capacitating mutations
4.9.7 Genetic engineering experiments
Chapter 5: Concluding remarks
5.1 Identification of the most complex genetic
interactions to date
5.2 Cloning of interacting genetic variants demonstrates
Ras signaling plays an important role
5.3 Transcriptional derepression of Ras targets mediates
capacitance
5.4 Additional architectures of capacitance can lead to
the same phenotype
5.5 Impact of my work
5.6 Future directions
References
Appendix A: Higher-order genetic interactions and their
contribution to complex traits
A.1 Overview
A.2 HGIs merit deeper investigation as a source of
complex trait variation
A.3 Evidence for HGIs
A.3.1 Suggestive evidence
A.3.2 Direct evidence
A.4 Identifying loci and genes involved in HGIs
A.5 Mechanistic basis of HGIs
A.6 Outstanding Questions
A.6.1 What is the typical architecture, effect size
distribution, and prevalence of HGIs?
A.6.2 Are there general mechanisms that result
in HGIs?
A.6.3 What is the relation between evolution and
HGIs?
A.7 Concluding remarks
Appendix B: The genotypic landscape of environmental
robustness
B.1 Overview
B.2 Introduction
B.3 Generation of a mapping population for dissecting
thermal robustness
B.4 Bulk segregant mapping of different thermal
robustness phenotypes
75
75
76
77
77
77
78
79
81
83
98
98
98
101
102
105
114
114
116
116
117
117
118
119
119
119
125
128
vii
Table of Contents
B.5 Differences in genetic architectures of fragility in the
BY backcross depend on MSS11
B.6 Discussion
B.7 Materials and Methods
B.7.1 Generating Backcross Segregants.
B.7.2 Phenotyping.
B.7.3 Bulk segregant mapping across levels of
robustness.
B.7.4 Identification of pairwise interacting loci
in the fragile class.
B.7.5 Genetic Engineering.
133
138
143
143
143
143
144
145
viii
List of Tables
Table 2.1 | Phenotypes and genotypes of tetrad spores from the
backcross to BY
Table 2.2 | Phenotypes and genotypes of tetrad spores from the
backcross to 3S
Table 2.3 | Initial bounds of detected loci for the five-way
interaction
Table 2.4 | Genotyping of multi-locus introgressed lines in the
BY direction
Table 2.5 | Genotyping of multi-locus introgressed strains in the
3S direction
Table 2.6 | Further genotyping of tetrad spores from the 3S
backcross
Table 2.7 | Genotypes within each phenotypic class among tetrad
spores from the backcross to BY
Table 2.8 | Genotypes within each phenotypic class among tetrad
spores from the backcross to 3S
Table 2.9 | Bounds of fixed loci among rough individuals with
END3
3S
Table 2.10 | Genotyping of additional rough individuals
possessing END3
3S
Table 2.11 | Genotyping of rough END3
3S
x 3S second-
generation backcross segregants
Table 2.12 | Genes within additional loci tested by deletion
Table 4.1 | Information regarding capacitating mutations
Table B.1 | Rough morphology is more common at low
temperature
Table B.2 | Rough segregants at 21° show various phenotypes at
higher temperatures
12
13
17
18
19
19
27
28
30
30
31
33
66
125
126
ix
List of Figures
Fig. 2.1 | Colony morphologies of parents and cross progeny on
rich medium containing ethanol
Fig. 2.2 | Sequencing coverage for each chromosome of the
backcrossed rough segregant
Fig. 2.3 | Allele frequency plots for control populations of BY
and 3S backcross segregants
Fig. 2.4 | Genetic dissection of the five-way interaction
Fig. 2.5 | Generation of multi-locus introgression strains
Fig. 2.6 | Evidence for functional variation at identified genes
Fig. 2.7 | Segregation analysis of causal genes and identification
of loci that complement END3
3S
Fig. 2.8 | Segregating phenotypes observed in cross between BY
and 3S
Fig. 2.9 | Smooth phenotype of end3
3S
∆ individual
Fig. 2.10 | Allele frequencies among rough individuals with
END3
3S
Fig. 3.1 | Capacitance, higher-order genetic interactions, and
genetic background effects might be related phenomena that
involve interactions among capacitating mutations and cryptic
variants
Fig. 3.2 | Colony morphology phenotypes that occur in the
BYx3S cross in the presence of ira2∆2933
Fig. 3.3 | Characterization of the six-way genetic interaction
Fig. 3.4 | Allele replacement results for FLO8
BY
, MGA1
3S
,
MSS11
3S
, and SFL1
3S
in the six-way genetic interaction
Fig. 3.5 | FLO11 is required for rough morphology and shows
differential expression across genetic backgrounds
Fig. 3.6 | Deletion of SFL1 reveals interacting cryptic variants
Fig. 4.1 | Phenotypes of 17 rough segregants used in this study.
Fig. 4.2 | Summary of genetic mapping data
Fig. 4.3 | Cloning of capacitating mutations in 8 rough
segregants carrying de novo mutations
Fig. 4.4 | Individuals 2 and 3 possess the same SSN8 lesion, but
are from different matings of BY and 3S
Fig. 4.5 | Deletion of mutant alleles from cognate rough
segregants harboring de novo lesions
Fig. 4.6 | Impact of temperature on rough segregants
Fig. 4.7 | Location of recombinations within the FLO11 locus
Fig. 4.8 | Replacement of cryptic allele suggests a background-
dependence
Fig. 4.9 | Capacitors and cryptic variants largely regulate
signaling and transcriptional control by the Ras pathway
Fig. A.1 | The types of genetic effects discussed in this
manuscript
Fig. A.2 | Examples of HGIs from model organism research
10
14
15
16
20
22
26
27
29
29
42
43
46
47
49
51
61
63
64
65
65
68
69
70
72
100
102
x
List of Figures
Fig. A.3 | Frequency of HGI progeny depends on parental
genotypes
Fig. A.4 | HGIs can cause differences in the phenotypic variance
of the two alleles at a locus
Fig. A.5 | Cloning genes underlying HGIs requires using
appropriate genetic backgrounds for engineering or
complementation testing
Fig. B.1 | Individuals can express the same phenotype due to
distinct genetic architectures that differ in environmental
robustness
Fig. B.2 | Ira2∆2933 segregants vary in their ability to express
the rough phenotype at different temperatures
Fig. B.3 | Bulk segregant mapping results for different robustness
classes
Fig. B.4 | A previously identified higher-order genetic interaction
leads to a moderate phenotype
Fig. B.5 | FLO11
3S
and MGA1
BY
confer robustness
Fig. B.6 | Multiple sets of genetic interactions cause the fragile
phenotype
Fig. B.7 | Multiple alleles contribute to rough morphology at
21°C
Fig. B.8 | Higher-order genetic interactions cause varying levels
of robustness
109
111
113
124
127
129
131
132
135
136
139
1
Abstract
Cryptic genetic variants, which typically do not show an impact on phenotype, can be
revealed by environmental or genetic perturbation. This phenomenon, known as
capacitance, may play a role in evolution and disease, as it allows for biological systems
to accumulate variation free of selection that can later be revealed. Despite a recognition
of its potential importance, very little is known about the genetic architectures or
molecular mechanisms underlying capacitance.
In this thesis, I describe the development of a system uniquely capable of providing
insights into capacitance at the genetic and molecular levels. In a cross of the yeast
isolates BY4716 and 322134S, both parents and the vast majority of progeny form
smooth colonies. However, a cryptic rough morphology phenotype can be revealed by
specific combinations of cryptic variants and capacitating mutations. The work described
in chapters 2-4 demonstrates how this phenotype change can function as a reporter for
capacitance, and thus allow for precise genetic dissection of this phenomenon.
In chapter 2, I demonstrate how a mutation in a negative Ras regulator can participate in
cryptic, higher-order genetic interactions far more complex than those that had previously
been demonstrated in the literature. In chapter 3, I explore the role this mutation plays at
the molecular level and find that transcriptional derepression can mediate phenotypic
capacitance. In chapter 4, I dissect the genetic architecture of several additional instances
of phenotypic capacitance, and show that many combinations of capacitors and cryptic
variants in the Ras pathway can reveal the same phenotype.
2
Chapter 1: Introduction
1.1 Cryptic genetic variation
Cryptic genetic variation refers to polymorphisms that do not show a phenotypic effect
under typical conditions., but that are revealed when an organism is exposed to an
alternate environment or a genetic perturbation (Gibson; Gibson & Dworkin; Masel &
Siegal; Paaby & Rockman). Since this variation is neutral under typical conditions, it is
free from selection and can accumulate by genetic drift within a population. The presence
of cryptic genetic variation may then provide a mechanism for complex combinations of
variants to potentiate rapid adaptation to a novel condition. It may also play a role in
disease, as many modern human diseases are influenced by ancestral alleles that may
represent cryptic variants exposed by our current environment (Gibson).
Many of the examples of cryptic variation in the literature demonstrate variability in
developmental pathways following environmental or genetic perturbation. The first
model of cryptic genetic variation showed that a heat shock during development could
reveal variation in wing formation within a population of fruit flies (Waddington).
Further, this variation could be selected upon, demonstrating that it had a genetic basis.
More recent work points to a mechanistic basis for this phenomenon. In a Caenorhabditis
elegans model, cryptic variation was shown to influence vulva cell fate, and researchers
found that this variation is likely caused by differences in Ras signaling (Milloz, Duveau,
Nuez, & Felix). However, few studies have demonstrated the genetic basis of cryptic
variation (e.g. (Chandler, Chari, & Dworkin; Gibson & Hogness; Jarosz & Lindquist)),
limiting our ability to broadly determine the mechanism and importance of this
phenomenon.
3
1.2 Capacitance
Capacitance occurs when cryptic genetic variants are revealed by some perturbation (i.e.
environment change, mutation). Just as an electrical capacitor stores charge that can be
released, a phenotypic capacitor buffers adaptive potential in the form of crpytic
variation. When a capacitor is perturbed by a mutation or an environmental stimulus, this
cryptic variation is revealed and can cause dramatic changes in phenotype.
The vast majority of work on capacitance comes from the study of Hsp90 (Rutherford &
Lindquist), a highly conserved chaperone found in bacteria and eukaryotes. Perturbation
of this protein by genetic or chemical methods can release cryptic genetic variation in
many model systems, including Arabidopsis (Queitsch, Sangster, & Lindquist), yeast
(Jarosz & Lindquist), and fruit flies (Rutherford & Lindquist). Discovery of this capacitor
instantly provided a mechanistic basis for capacitance: since chaperones help with the
folding of proteins, the presence of a chaperone allows for some flexibility in the coding
sequencing of a particular peptide. Non-synonymous SNPs that cause a slight reduction
in protein folding efficiency can thus be masked by Hsp90. However, when Hsp90 is
removed from the system, these changes may have a more dramatic effect.
The volume of work on capacitance by Hsp90 implies some exclusivity in its role has a
capacitor. However, systems modeling work has suggested that capacitance may be a
general feature of gene regulatory networks, and thus many genes may be able to act as
capacitors (Bergman & Siegal). This work is backed up by gene deletion experiments in
yeast that demonstrate a high variability in expression casued by knocking out “hub”
genes within networks (Bergman & Siegal), and by evidence in Drosophila that many
4
capacitors may be capable of revealing cryptic variation influencing wing shape
(Takahashi).
1.3 Rough morphology system
During my PhD, I have developed a powerful system for genetically dissecting instances
of capacitance. In a cross between the yeast isolates BY4716 (BY, a derivative of the lab
strain S288c) and 322134S (3S, a clinical isolate taken from an AIDS patient), both
parents and the vast majority of progeny show a smooth colony morphology. However,
some segregants, when perturbed by mutation or environmental change, can display a
rough morphology phenotype. Thus, each rough segregant from this cross represents an
instance of capacitance.
This system provides several key advantages for the study of capacitance. Key among
these is the fact that rough morphology is a discrete trait that can be easily screened for in
large populations of BY x 3S segregants. This allows for the use of standard genetic
mapping approaches to identify loci containing cryptic genetic variants and capacitating
events simultaneously. I take advantage of this in chapters 2 and 4 to assess the most
common genetic architectures underlying several instances of capacitance.
Further, the use of Saccharomyces cerevisiae as a model system allows for rapid and
precise genetic dissection of capacitance. Yeast has a very high recombination rate and is
very tractable for genetic engineering experiments. Once I cloned genes harboring cryptic
variation or capacitating mutations, I had access to decades of molecular data about the
role of these genes. This allowed me in chapter 3 to form hypotheses about the role of
5
transcriptional derepression in the Ras signaling pathway in phenotypic capacitance. I
then was able to strengthen the claims made in that chapter when I cloned additional
cryptic variants and capacitating events that almost all influence the Ras pathway or
transcription.
1.4 Goals of this Thesis
When I began my PhD, several important questions about capacitance had yet to be
answered. First, what is the genetic architecture of capacitance? Many studies had
characterized individual cryptic alleles and capacitors, but the complexity of interactions
between capacitors and cryptic alleles was unknown.
Second, what types of genes can act as capacitors? As mentioned above, the best-
characterized example of a capacitor is the chaperone Hsp90. The wealth of literature on
this protein has led to a view that it is an exceptional case, and that capacitance may be an
exclusive property of chaperones. However, theoretical work has demonstrated that
capacitance may be a general feature of gene regulatory networks.
Third, what is the molecular basis of capacitance beyond that demonstrated by
perturbation of chaperones? Again, despite theory on the role of gene regulatory
networks in capacitance, few empirical examples have been demonstrated. Characterizing
the mechanism underlying examples of capacitance would strengthen this argument and
provide a model for future work on the topic.
6
1.5 Summary of Chapters
In Chapter 2, I characterize the most complex genetic interactions to date. My findings
demonstrate how mutations can interact with specific combinations of loci to reveal
cryptic genetic interactions underlying capacitance.
In Chapter 3, I demonstrate how a capacitating mutation reveals cryptic genetic variation
by relieving transcriptional repression of Ras targets. Genes interacting with this mutation
are largely transcription factors in the Ras pathway, indicating that gene regulation plays
an important role in capacitance.
In Chapter 4, I use the rough morphology system to dissect the genetic architecture of 17
instances of capacitance. In doing so, I find that many unique combinations of cryptic
variants and capacitating events can cause the same trait. Capacitating events included
mutations, an intragenic recombination in a cell-cell adhesion molecule, and highly
complex interactions between specific cryptic variants. Cloning of genes harboring
capacitating mutations and cryptic variants suggests that combinatorial perturbation of
the Ras signaling pathway is responsible for the rough phenotype. This work is the first
demonstration of the range of architectures underlying capacitance, and suggests that
gene regulatory network perturbation is important in this phenomenon.
In Chapter 5 I tie together each of these findings, discuss the impact of my work, and
propose additional questions that could be answered using the colony morphology system
I developed.
7
Chapter 2: Genetic interactions involving five or more genes contribute to a
complex trait in yeast
This work appears essentially as published in 2014 in
PLoS Genetics, 10(5): e1004324
2.1 Overview
Recent research suggests that genetic interactions involving more than two loci may
impact a number of complex traits. How these ‘higher-order’ interactions arise at the
genetic and molecular levels remains an open question. To provide insights into this
problem, we dissected a colony morphology phenotype that segregates in a yeast cross
and is influenced by synthetic higher-order interactions. Using backcrossing and selective
sequencing of progeny, we found five loci that collectively produce the trait. We fine-
mapped these loci to 22 genes in total and identified a single gene at each locus that
caused loss of the phenotype when deleted. Complementation tests or allele replacements
provided support for functional variation in these genes, and revealed that combinations
of pre-existing genetic variants and spontaneous mutations underlie the phenotype. The
causal genes have diverse functions in endocytosis (END3), oxidative stress response
(TRR1), RAS-cAMP signalling (IRA2), and transcriptional regulation of multicellular
growth (FLO8 and MSS11), and for the most part have not previously been shown to
have functional relationships. Further efforts revealed two additional loci that together
can complement the non-causal allele of END3. Our work sheds light on the complex
genetic and molecular architecture of higher-order interactions, and raises questions
about the broader contribution of such interactions to phenotypic variation.
2.2 Introduction
Understanding the genetic basis of complex traits is critical for advancing medicine,
evolutionary biology, and agriculture (Falconer DS, 1996; Lynch M, 1998). A challenge
8
to progress in this area is that genetic variants can interact, resulting in unexpected
phenotypic consequences (Lehner, 2011; Mackay; Mackay et al., 2012; Nelson,
Pettersson, & Carlborg, 2013; Phillips, 2008). Most of our knowledge about these genetic
interactions in natural systems comes from studies focused on two-locus interactions
where at least one of the loci exhibits a measurable effect on its own (e.g., (Brem, Storey,
Whittle, & Kruglyak, 2005)). However, evidence suggests that genetic interactions
involving three or more loci also occur (Dowell et al., 2010; Pettersson, Besnier, Siegel,
& Carlborg, 2011), and that loci participating in such interactions may not individually
have detectable effects (Bloom, Ehrenreich, Loo, Lite, & Kruglyak, 2013). Determining
how these higher-order interactions arise and influence phenotypic variation could help
solve the ‘missing heritability’ problem faced by geneticists studying humans and model
species (Manolio et al., 2009).
In this paper, we describe the genetic basis of a complex trait in the budding yeast
Saccharomyces cerevisiae that is influenced by higher-order genetic interactions. We
identified this phenotype, a dramatic change in the morphology of yeast colonies, in a
cross of haploid derivatives of the lab strain BY4716 and the clinical isolate 322134S
(hereafter ‘BY’ and ‘3S’, respectively). The colony morphology trait in the BYx3S cross
is similar to phenotypes described in other yeast isolates and crosses (e.g., (Granek &
Magwene, 2010; Granek, Murray, Kayrkci, & Magwene, 2013; Halme, Bumgarner,
Styles, & Fink, 2004; Holmes, Lancaster, Lindquist, & Halfmann, 2013; Ryan et al.,
2012; Tan et al., 2013; Wilkening et al., 2013)). Thus, by comprehensively determining
the genetic basis of colony morphology variation among BYx3S offspring, we not only
generate novel insights into how higher-order interactions contribute to phenotypic
9
variation, but also provide new information regarding the genetic basis of a frequently
studied model complex trait.
2.3 Identification of a discrete complex trait
Although both BY and 3S, as well as most of their haploid offspring, form smooth
colonies (Fig. 2.1A-1C), ~2% of their progeny exhibited rough colonies when we
examined 250 segregants (Fig. 2.1D). Previous work has shown that such heritable
variation in colony morphology in S. cerevisiae can arise due to naturally occurring
polymorphisms or spontaneous mutations at chromosomal loci (Granek & Magwene,
2010; Granek et al., 2013; Halme et al., 2004; Wilkening et al., 2013), aneuploidies (Tan
et al., 2013), and prions (Holmes et al., 2013). Unlike chromosomal loci, which should
show stable inheritance across generations, aneuploidies and prions can be gained or lost,
resulting in phenotypic switching.
10
Fig. 2.1 | Colony morphologies of parents and cross progeny on rich medium
containing ethanol. 3S (A) and BY (B) each form smooth colonies. When these two
strains are crossed, most offspring also form smooth colonies, but a small fraction
form rough colonies. An example smooth segregant is shown in C and an example
rough segregant is shown in D.
Multiple lines of evidence suggest that chromosomal loci are the primary cause of rough
morphology in the BYx3S cross. Neither BY nor 3S exhibits rough morphology,
indicating that the phenotype likely requires a combination of alleles from both of these
strains. Consistent with this statement, we found that the frequency of rough morphology
increased to 12.5% and 21.2% among recombinant haploid progeny obtained by
backcrossing a rough segregant to BY and 3S, respectively (Methods; Tables 2.1 and 2.2).
The higher frequency of rough segregants in backcrosses is expected if alleles from both
parents contribute to the trait, as fewer causative alleles should segregate in the
backcrosses than in the original cross. Further supporting the argument that our
observations of rough morphology were due to chromosomal loci instead of transient
11
factors, we found no evidence for chromosome-scale aneuploidies or phenotypic
switching in the backcrossed segregant (Fig. 2.2; Methods).
12
Spore
# chrIV chrV chrXV mat phenotype
1A 1 1 0 x s
1B 1 1 1 a b
1C 0 0 0 a s
1D 0 0 1 x s
2A 0 0 1 a s
2B 0 0 0 a s
2C 1 1 1 x r
2D 1 1 0 x s
3A 0 0 1 a s
3B 1 1 0 x s
3C 0 0 0 x s
3D 1 1 1 a r
4A 0 1 1 a b
4B 1 1 1 x r
4C 1 0 0 a s
4D 0 0 0 x s
5A 0 0 1 x s
5B 1 0 0 x s
5C 0 1 0 a s
5D 1 1 1 a s
6A 1 1 0 a s
6B 0 1 1 x b
6C 0 0 0 x s
6D 1 0 1 a s
7A 0 0 0 a s
7B 1 0 1 x s
7C 1 1 0 x s
7D 0 1 1 a b
8A 1 1 1 x r
8B 0 0 1 x s
8C 1 1 0 a s
8D 0 0 0 a s
9A 1 0 1 a s
9B 0 0 1 x s
9C 1 1 0 a s
9D 0 1 0 x s
10A 0 0 1 x s
10B 1 0 0 a s
10C 1 1 1 a b
10D 0 1 0 x s
11A 1 0 0 x s
11B 0 1 1 a s
11C 1 1 1 a s
11D 0 0 0 x s
12A 0 0 1 a s
12B 1 1 1 x r
12C 1 1 0 a s
12D 0 0 0 x s
13A 1 1 1 x r
13B 0 0 1 x s
13C 1 0 0 a s
13D 0 1 0 a s
14A 0 0 0 x s
14B 0 1 1 a b
14C 1 0 0 x s
14D 1 1 1 a r
Table 2.1 | Phenotypes and genotypes of tetrad spores from the backcross to BY.
Individuals from 14 dissected tetrads were phenotyped and genotyped at
segregating markers within causal loci. Phenotypes are recorded as smooth (s),
rough (r) or bumpy subphenotype (b). Genotypes at segregating markers within
each locus are denoted as 1 (3S) or 0 (BY). Under the column “Spore #”, the number
in the name represents the tetrad, while the letters signify different spores.
13
spore chrXIII chrXIV mat phenotype
1A 0 1 x s
1B 1 0 a s
1C 0 0 a s
1D 1 1 x s
2A 0 1 x s
2B 1 0 a s
2C 0 1 x s
2D 1 0 a r
3A 0 1 x s
3B 0 0 a s
3C 1 1 a r
3D 1 0 x s
4A 0 0 a s
4B 0 0 x s
4C 1 1 x s
4D 1 1 a r
5A 1 1 x r
5B 0 1 a s
5C 0 0 a s
5D 1 0 x s
6A 0 0 a s
6B 0 0 a s
6C 1 1 x r
6D 1 1 x r
7A 0 1 a s
7B 0 1 x s
7C 1 0 x s
7D 1 0 a s
8A 1 0 a s
8B 0 0 x s
8C 1 1 a s
8D 0 0 x s
9A 0 1 a s
9B 1 1 x r
9C 1 0 a s
9D 0 0 x s
10A 1 0 a s
10B 0 1 x s
10C 0 1 x s
10D 1 0 a s
11A 1 0 a r
11B 0 0 x s
11C 1 1 a s
11D 0 1 x s
12A 0 0 a s
12B 1 1 x r
12C 1 1 a s
12D 0 0 x s
13A 0 0 a s
13B 0 0 x s
13C 1 1 x r
13D 1 1 a r
Table 2.2 | Phenotypes and genotypes of tetrad spores from the backcross to 3S.
Individuals from 13 dissected tetrads were phenotyped and genotyped at
segregating markers within causal loci. Phenotypes are recorded as smooth (s) or
rough (r). Genotypes at segregating markers within each locus are denoted as 1
(BY) or 0 (3S). Under the column “Spore #”, the number in the name represents the
tetrad, while the letters signify different spores. 5 individuals possessed BY alleles at
MSS11 and END3, yet showed smooth morphology. Further genotyping revealed
that bolded individuals lacked the IRA2
3S
-Δ2933 allele.
14
Fig. 2.2 | Sequencing coverage for each chromosome of the backcrossed rough
segregant. The segregant used in backcrossing was sequenced to ~4.39x coverage.
We determined the average coverage of nucleotides on each chromosome (grey
bars) and across the genome (dotted line). None of the chromosomes exhibited a
significant excess or deficit of coverage, suggesting that the strain did not carry any
large aneuploidies.
2.4 Genetic mapping of rough morphology
To identify loci that contribute to rough morphology, we generated thousands of random
spores from the aforementioned backcrosses and used low-coverage whole genome
sequencing to selectively genotype individuals that showed the phenotype (Methods). We
obtained 92 and 88 rough segregants from the BY and 3S backcrosses, respectively.
Using these data, we detected five genomic loci that were strongly enriched among these
individuals but not among control segregants (Fig. 2.3): three on Chromosomes IV, V,
and XV inherited from 3S (Fig. 2.4A), and two on Chromosomes XIII and XIV inherited
from BY (Fig. 2.4B). All of these loci, except the one on Chromosome XIV, were fixed
among individuals with rough morphology.
15
Fig. 2.3 | Allele frequency plots for control populations of BY and 3S backcross
segregants. To account for unintentional selection in our mapping populations, we
sequenced a control population of segregants from each backcross. Genome-wide
allele frequency plots are shown for control populations of segregants from
backcrosses to BY (A) and 3S (B). The diploid parent of these backcrosses was
sporulated and plated at high density on selective medium to obtain recombinant
MATa backcross progeny (Methods). Thousands of segregants were pooled together
by scraping them off plates. DNA was extracted from the pools and used to generate
Illumina whole genome sequencing libraries. These libraries were then sequenced to
~200X coverage. We pulled out data for SNPs described in the Methods and used
these data to generate the above plots. The plot was generated by smoothing the
data for each chromosome using the filter() function in R and a window size of 50
SNPs. Causal loci for rough morphology are labeled with black arrows and selected
markers used to generate MATa progeny are labeled with grey arrows. We note that
there is a site on Chromosome XIV near the causal locus that shows enrichment, but
is distinct from the region involved in rough morphology.
16
Fig. 2.4 | Genetic dissection of the five-way interaction. Genome-wide allele
frequency plots are shown for mapping populations from backcrosses to BY (A) and
3S (B). Enriched loci are outlined in orange (causal allele from 3S parent), blue
(causal allele from BY parent), or grey (allele is a selectable marker engineered into
BY). Genes within the Chromosome IV (C), V (D), XV (E), XIII (F), and XIV (G)
loci are shown. Every gene within our minimum interval for a locus is included.
Grey boxes delimit regions of the loci that were subsequently excluded from
consideration based on fine-mapping experiments. All non-essential genes within
these windows were deleted in a multi-locus introgression strain that showed rough
morphology. Genes that caused loss of the phenotype when deleted are colored red.
2.5 Cloning of causal genes
We attempted to determine causal genes underlying each of the five loci. Our initial
resolution of the loci was between 4 and 14 genes (Figures. 2.4C-2.4G; Table 2.3;
Methods). To decrease the number of candidate genes, we performed targeted genotyping
on additional backcross segregants (16 rough and 3 smooth), as well as 8 multi-locus
introgression strains that had been subjected to 6 rounds of backcrossing with selection
for the rough phenotype (Fig. 2.5; Methods). This additional stage of genetic mapping
refined the loci to between 2 and 9 genes per locus, and 22 genes in total (Figures. 2.4C-
G; Table 2.4-2.6). We deleted each of the 20 remaining non-essential candidate genes
17
from one of the multi-locus introgression strains (Methods). Across these deletions, a
single gene at each locus showed an effect on the phenotype: TRR1 (Chromosome IV),
FLO8 (Chromosome V), MSS11 (Chromosome XIII), END3 (Chromosome XIV), and
IRA2 (Chromosome XV) (Figures. 2C-2G).
chromosome
start
position
stop
position
IV 1174395 1183454
V 371849 378140
XIII 571144 592735
XIV 447324 473648
XV 171973 184291
Table 2.3 | Initial bounds of detected loci for the five-way interaction. Causal loci
were defined as regions present in at least 95% of individuals. To identify the
intervals at these loci, we took all individuals with the causal allele and determined
the minimum region delimited by recombination breakpoints.
18
typed pools
examined marker 1 2 3 4
4_1 0 0 1 0
4_1.5 0 0 1 0
4_2 1 1 1 1
4_3 1 1
4_3.5 0 0 1 0
4_4 0 0 1 0
5_0a 1 1 1 0
5_0b 1 1 1 0
5_1a 1 1 1 1
5_1 1
5_2 1 1 1 1
5_3 1 1 1 1
5_4a 1 1 1 1
5_4_5 1 0 1 0
15_1 1 0 1 1
15_2 1 1 1 1
15_3 1 1 1 1
15_3.5 0 1 1
15_4 0 1 1 0
Table 2.4 | Genotyping of multi-locus introgressed lines in the BY direction. Multi-
locus introgressed lines were generated through six rounds of backcrossing to BY
with phenotypic selection (Methods; Fig. 2.5). Lines were genotyped across causal
loci by Sanger Sequencing of segregating markers. 3S alleles are denoted as ‘1’; BY
alleles are denoted as ‘0’.
19
typed pools
examined
marker
1 2 3 4
13_1 1 0 1 0
13_2 0 0
13_3 1 1 1
13_7 1 1 1 1
13_9 0 0 1 1
14_0 1 1 1
14_1 1 1 1 1
14_3 1 1 1
14_9 1 1 1 1
14_10 0 1 1
Table 2.5 | Genotyping of multi-locus introgressed strains in the 3S direction. Multi-
locus introgressed lines were generated through six rounds of backcrossing to 3S
with phenotypic selection (Methods; Fig. 2.5). Lines were genotyped across causal
loci by Sanger Sequencing of segregating markers. A BY alleles are denoted as ‘1’;
3S alleles are denoted as ‘0’.
Spore # phenotype 14_0 14_1 14_2 14_3 14_4 14_7 14_9 14_10
1D S 1 1 1 1 1 1
3C R 1 1 1 1 1 1
4D R 1 1 1 1 1 1
4C S 1 1 1 1 1 1
5A R 1 1 1 1 1 1 1 1
6C R 1 1 1 1 1 1
6D R 1 1 1 1 1 1
7D S 1 1 1 0 0 0 1
9B R 1 1 1 1 1 1
11C S 1 1 1 1 1 1
12B R 1 1 1 1 1 1
12C S 1 1 1 1 1 1
13C R 1 1 1 1 1 1
2D R 0 0 0 0 0 0 0 0
11A R 0 0 0 0 0 0 0 0
Table 2.6 | Further genotyping of tetrad spores from the 3S backcross. Select tetrad
spores from the 3S backcross were genotyped across causal loci by Sanger
Sequencing of segregating markers. A 1 indicates that genotyped individuals
possessed the 3S allele at a given marker and 0 indicates the BY allele.
20
Fig. 2.5 | Generation of multi-locus introgression strains. A rough segregant was
subjected to six rounds of backcrossing with selection for the rough phenotype to
reduce the genetic contribution of one parent strain and allow for finer resolution of
causal loci.
Because the two remaining candidate genes—AVO1 and TOP2—are essential for
viability, we could not delete these genes from a rough haploid individual. We used an
alternate approach to test the effects of these genes. Heterozygous diploids for the loci
containing these genes were generated by mating a rough multi-locus introgression strain
to an appropriate backcross segregant. Deletion of AVO1 and TOP2 in these diploid
21
heterozygotes generated mixed populations of hemizygotes, with approximately equal
numbers of individuals carrying the BY and 3S alleles of these genes. All hemizygotes
exhibited rough morphology, suggesting that these genes do not harbor functional
variation that affects the trait.
We used complementation tests to determine whether the five identified genes possess
functional variation (Methods). Each haploid deletion strain was mated to three rough
and three smooth haploid backcross progeny (Methods). These matings were designed to
produce diploids that were homozygous for the required alleles at four of the causal loci
and hemizygous for the fifth causal locus. For END3, FLO8, MSS11, and TRR1, the
experiments provided support that the parental alleles differ in their effects. All matings
of deletion strains to smooth backcross progeny produced smooth hemizygotes. Further,
either two (in the cases of FLO8 and MSS11) or three (in the cases of TRR1 and END3)
of the matings of deletion strains to rough backcross progeny produced rough
hemizygotes (Fig. 2.6A). However, for IRA2, the two possible hemizygotes showed no
phenotypic difference, with both exhibiting smooth morphology (Fig. 2.6A). IRA2 has
been reported to show haploinsufficiency in growth rate experiments (Pir et al., 2012),
and this haploinsufficiency may also explain some of our reciprocal hemizygosity results
for this gene.
22
Fig. 2.6 | Evidence for functional variation at identified genes. In A, we show
representative results from genetic engineering experiments in a multi-locus
introgression strain. Haploid deletion strains are shown in the top row, while the
second and third rows contain diploid hemizygotes. For each gene and allele, we
constructed multiple hemizygotes for each allele, with representative phenotypes
shown in the figure. The terms ‘causal’ and ‘non-causal’ refer to which allele was
detected in our initial genetic mapping experiment. In B, we used genetic
engineering in a smooth backcross segregant with the genotype END3
BY
FLO8
3S
IRA2
BY
MSS11
BY
TRR1
3S
to confirm the involvement of IRA2 in rough colony
morphology. The segregant subjected to transformations is shown on the left,
followed by representative knock-ins of IRA2
BY
, IRA2
3S
, and the IRA2 allele of the
rough segregant used in backcrossing (referred to as IRA2
3S
-∆2933). The allele
replacements have a kanMX tail, which was used to select for integration into the
chromosome.
2.6 IRA2 contains a mutation required for rough morphology
To provide stronger support for IRA2’s role in the trait, we performed allele replacements
of IRA2 in a smooth backcross segregant that carried the non-causal allele of IRA2, as
well as the causal alleles of END3, FLO8, MSS11, and TRR1 (Methods). While
transformations with the IRA2
3S
allele had no phenotypic effect, we found that
transformations with the IRA2 allele from the rough segregant that had been backcrossed
resulted in a change from smooth to rough morphology (Fig. 2.6B). Sequencing of IRA2
from 3S and the rough segregant revealed a single difference between the two alleles: a
23
frameshift mutation that truncates the protein by 117 amino acids (hereafter referred to as
IRA2
3S
-Δ2933). IRA2 is known to be hypermutable and spontaneous mutations in this
gene have been shown to influence a variety of multicellular growth phenotypes (Halme
et al.; Roop & Brem). However, our results demonstrate that the effects of spontaneous
mutations in IRA2 can depend on an individual’s genotype at a number of additional
genes. We also checked for IRA2
3S
-Δ2933 in the four other rough individuals that we
found in our original BYx3S mapping population. Three of these rough segregants
possessed the frameshift mutation, suggesting both that IRA2
3S
-Δ2933 probably arose
during the outgrowth of the BY/3S diploid prior to its sporulation and that a different
causal mutation may be carried by one BYx3S segregant.
2.7 Candidate causal polymorphisms
Previous work by other groups identified functional polymorphisms in END3 and FLO8
that also segregate in our cross (Liu, Styles, & Fink, 1996; Steinmetz et al., 2002). BY
has a premature stop mutation in FLO8 that prevents it from undergoing many forms of
multicellular growth (Liu et al., 1996). As for END3, a missense polymorphism in this
gene contributes to variability in high temperature growth in a cross of the clinical isolate
YJM789 and S288c, the progenitor of BY (Steinmetz et al., 2002). Of relevance to our
study, this variant in END3 has effects that are strongly dependent on genetic background
(Sinha, Nicholson, Steinmetz, & McCusker, 2006). With respect to TRR1, the
Saccharomyces Genome Resequencing Project (Liti et al., 2009) and our own sequencing
data indicate that the BY and 3S alleles of this gene differ by a single nucleotide, which is
a synonymous SNP in the 52
nd
codon of the gene: BY has an ATC codon and 3S has an
ATT codon. Although both of these codons are recognized by the same isoleucine tRNA,
24
the ATT codon is preferred by a nearly two-to-one ratio throughout the yeast genome,
suggesting that the SNP might have an effect on translational efficiency. Only lab-
derived S. cerevisiae strains carry the ATC allele that confers smooth morphology, while
all other sequenced S. cerevisiae and S. paradoxus strains harbor the ATT allele that is
likely involved in rough morphology. Work to determine the functional variant(s) in
MSS11, which possesses a number of coding and noncoding polymorphisms that could
have effects, is ongoing.
2.8 Description of causal genes
The causal genes encode proteins with diverse cellular functions: End3 plays a role in
clathrin-mediated endocytosis (Benedetti, Raths, Crausaz, & Riezman, 1994; Tang, Xu,
& Cai, 2000), Flo8 and Mss11 are transcription factors that regulate cell-cell adhesion
and multicellular phenotypes in S. cerevisiae (Gagiano et al., 2003; Kobayashi, Suda,
Ohtani, & Sone, 1996), Ira2 is a negative regulator of the RAS-cAMP pathway (Tanaka,
Nakafuku, Tamanoi, et al., 1990), and Trr1 is an enzyme involved in oxidative stress
response (Pedrajas et al., 1999; Ross, Findlay, Malakasi, & Morgan, 2000). Flo8 and
Mss11 physically interact (T. S. Kim, Kim, Yoon, & Kang, 2004), and IRA2 and MSS11
show a genetic interaction when both are knocked out (Hoppins et al., 2011). To our
knowledge, none of the other pairs of identified genes have been reported to interact at
the biochemical, genetic, physical, or regulatory levels. To assess whether Flo8 and
Mss11 might directly regulate the expression of the other genes, we examined existing
data from calling card analyses, a technique that identifies genomic sites bound by
transcription factors (Ryan et al., 2012). These results indicated that Flo8 and Mss11 are
25
unlikely to bind the promoters of END3, IRA2, and TRR1, although admittedly the study
involved a different strain than our cross parents.
2.9 Only two discrete genotypes can exhibit rough morphology
After identifying causal genes at the five loci, we analyzed the effects of these genes in
more detail by genotyping them in a panel of phenotyped segregants from dissected
backcross tetrads (Methods). Every individual with rough morphology possessed the 3S
allele of, FLO8 and TRR1, the BY allele of MSS11, and IRA2
3S
-Δ2933 (Figures. 2.7A,
2.7B, 2.8A, and 2.8B; Tables 2.7 and 2.8). Although most individuals with rough
morphology carried END3
BY
, a small fraction of individuals with END3
3S
also showed
the trait (Fig. 2.7B and 2.8C; Table 2.8), indicating that alleles at additional loci
complement END3
3S
. All but one of the other examined genotypes produced smooth
colonies and could not be phenotypically differentiated from the parent strains (Fig.
2.8D-2.8F). Individuals carrying the interacting alleles at END3, FLO8, IRA2, and
MSS11, but the non-interacting allele at TRR1 were the exception, as they formed
colonies with surfaces that were slightly bumpy (Fig. 2.8D).
26
Fig. 2.7 | Segregation analysis of causal genes and identification of loci that
complement END3
3S
. Spores from dissected tetrads were phenotyped and then
genotyped at END3, FLO8, IRA2, MSS11, and TRR1. Results from the BY and 3S
backcrosses are in A and B, respectively. In C, we show genotyping results for
candidate loci on Chromosomes VII, XI, XII, and XV in a population of END3
3S
segregants from a second-generation backcross. Lastly, in D, we show the different
genotypes in the BYx3S cross that specify rough morphology. In A-D, blue and
orange denote BY and 3S alleles, respectively, while grey in D indicates that either
allele can occur. In the case of IRA2, blue specifies the BY allele while orange refers
to IRA2
3S
-∆2933. Rough and smooth morphology are specified in A-C by black and
light grey, respectively.
27
Fig. 2.8 | Segregating phenotypes observed in cross between BY and 3S. The
phenotypes of representative recombinant genotypes are shown: (A) a rough
segregant obtained from the backcross to 3S, (B) a rough segregant obtained from
the backcross to BY, (C) an individual with the 3S allele at END3 that shows rough
morphology, (D) an individual with the BY allele at TRR1 but the interacting alleles
at END3, FLO8, IRA2, and MSS11 shows a bumpy surface, (E) a smooth segregant
from the backcross to 3S, and (F) a smooth segregant from the backcross to BY.
genotype rough smooth
TRRI
3S
/FLO8
3S
/IRA2
3S
7 4
TRRI
BY
/FLO8
3S
/IRA2
3S
0 5
TRRI
3S
/FLO8
BY
/IRA2
3S
0 3
TRRI
3S
/FLO8
3S
/IRA2
BY
0 8
TRRI
BY
/FLO8
BY
/IRA2
3S
0 9
TRRI
BY
/FLO8
3S
/IRA2
BY
0 4
TRRI
3S
/FLO8
BY
/IRA2
BY
0 6
TRRI
BY
/FLO8
BY
/IRA2
BY
0 10
Table 2.7 | Genotypes within each phenotypic class among tetrad spores from the
backcross to BY.
28
genotype rough smooth
MSS11
BY
/END3
BY
9 5
MSS11I
3S
/END3
BY
0 11
MSS11
BY
/END3
3S
2 10
MSS11
3S
/END3
3S
0 15
Table 2.8 | Genotypes within each phenotypic class among tetrad spores from the
backcross to 3S.
We more deeply investigated the genetic basis of rough morphology among individuals
with END3
3S
. First, we used a gene knockout strategy to check whether END3
3S
is
necessary for these individuals to exhibit rough morphology (Methods). end3
3S
∆ strains
were smooth, suggesting that the alternate genetic architecture for rough morphology
requires END3
3S
. Second, we tried to identify loci that complement END3
3S
(Fig. 2.9).
Four rough END3
3S
progeny were present in our sequenced mapping population from the
3S backcross. Among these segregants, we detected 11 previously unidentified genomic
regions where individuals shared the same genotype (Fig. 2.10; Table 2.9; Methods). We
were able to reduce this set to four candidate loci on Chromosomes VII, XI, XII, and XV
by genotyping additional backcross progeny (Table 2.10; Methods). To determine which
of the four loci have causal roles in rough morphology, we mated a relevant backcross
segregant to 3S and analyzed a panel of 51 second-generation backcross progeny (Table
2.11; Methods). The BY alleles at the Chromosome VII and XV loci were fixed among
the 39 individuals with rough morphology, while the other two loci showed no evidence
of playing a role in the trait (Fig. 2.7C; Table 2.11). Given that only individuals that
carried BY alleles at both the Chromosome VII and XV loci exhibited rough morphology,
it is likely that these loci genetically interact to complement END3
3S
.
29
Fig. 2.9 | Smooth phenotype of end3
3S
∆ individual. END3 was deleted from a
segregant that possessed the END3
3S
allele and the alleles that cause rough
morphology at TRR1, FLO8, MSS11, AVO1, and the additional loci on chromosomes
VII and XV.
Fig. 2.10 | Allele frequencies among rough individuals with END3
3S
. Allele
frequencies of individuals from the initial 3S mapping population that showed
rough morphology but lacked the BY allele of END3 are plotted. Additional 3S
backcross segregants were obtained, and analyzed as individuals using phenotyping
and genotyping at the new loci. The SGA markers used to select MATa haploid
segregants are labelled with a grey arrow, while the chromosome XIII locus
containing MSS11 is labelled with a green arrow. The 11 novel loci are shown with
black and red arrows, with the difference being black loci were sites that remained
as candidates after typing of additional segregants that possessed rough morphology
and the alternate causal genotype.
30
chromosome start stop
I 114628 227733
II 1687 19967
V 177250 254385
VII 949001 1009506
VIII 380512 489755
XI 85423 155453
XI 578824 625172
XII 210695 397605
XIII 823029 864754
XV 138746 143022
XV 586701 736524
Table 2.9 | Bounds of fixed loci among rough individuals with END3
3S
. Four
sequenced segregants with rough morphology possessed END3
3S
. These individuals
shared 11 previously undetected loci. Intervals we later identify as causal are in bold.
typed individual
typed loci A1 A2
chrI 0 0
chrII 1 0
chrV 1 0
chrVII 1 1
chrVIII 0 1
chrXI-1 1 1
chrXI-2 0 1
chrXII 1 1
chrXIII 0 0
chrXV-1 1 0
chrXV-2 1 1
Table 2.10 | Genotyping of additional rough individuals possessing END3
3S
. Two
rough individuals with END3
3S
were typed across loci identified in figure S6. A 1
indicates that all genotyped individuals possessed the BY allele at a given marker
and 0 indicates the 3S allele.
31
ndividual chrVII chrXI chrXII chrXV
1 1 1 0 1
2 1 1 1 1
3 1 0 0 1
4 1 0 0 1
5 1 1 1 1
6 1 0 1 1
7 1 0 1 1
8 1 1 0 1
9 1 0 0 1
10 1 1 1 1
11 1 0 1 1
12 1 1 0 1
13 1 0 0 1
14 1 0 0 1
15 1 1 1 1
16 1 1 0 1
17 1 1 1 1
18 1 1 0 1
19 1 0 1 1
20 1 0 0 1
21 1 0 1 1
22 1 1 0 1
23 1 0 0 1
24 1 1 1 1
25 1 0 0 1
26 1 1 1 1
27 1 0 1 1
28 1 1 0 1
29 1 0 1 1
30 1 0 0 1
31 1 1 0 1
32 1 0 1 1
33 1 0 1 1
34 1 1 0 1
35 1 1 1 1
36 1 0 0 1
37 1 0 0 1
38 1 0 0 1
39 1 1 1 1
c1 0 0 1 0
c2 0 1 1 0
c3 1 0 0 0
c4 1 0 0 1
c5 0 1 1 0
c6 1 0 1
c7 1 0 1 1
c8 0 1 0 0
c9 1 1 0 1
c10 1 0 1 1
c11 1 1 0 0
c12 1 1 1 1
Table 2.11 | Genotyping of rough END3
3S
x 3S second-generation backcross
segregants. 39 individuals with rough morphology (1-39) and 12 with smooth
morphology (c1-c12) were genotyped at segregating markers within four candidate
loci. A 1 indicates that all genotyped individuals possessed the 3S allele at a given
marker and 0 indicates the BY allele.
32
Our findings indicate that the segregant used for backcrossing carried more than one set
of interacting alleles that can specify rough morphology (Fig. 2.7D). Identifying the
causal genes and genetic variants underlying the Chromosome VII and XV loci can thus
shed light on how these different genotypes produce the same trait. However, our
ongoing efforts to clone the causal factors at these loci are limited by the crude resolution
of the present data (each locus is presently resolved to >60 kilobases; Table 2.9). We note
that initial gene deletion experiments focused on 18 candidates (Table 2.12), including
LAS17 and YAP1802, whose cognate proteins functionally interact with End3 (Howard,
Hutton, Olson, & Payne, 2002; Tarassov et al., 2008), have been unsuccessful. Moving
forward, we plan to determine the genes that underlie the Chromosome VII and XV loci,
and characterize their relationship with END3.
33
gene chromosome position
PHO81 VII 958210 to 954674
YAP1802 VII 976581 to 974875
SDA1 VII 982068 to 979765
SOL4 VII 985972 to 986739
MGA1 VII 988049 to 989419
SFL1 XV 586981 to 589281
ELG1 XV 605092 to 602717
PNO1 XV 606171 to 605347
MDM32 XV 606607 to 608475
SPP2 XV 609197 to 608640
SMP3 XV 611388 to 609838
MRPL23 XV 611999 to 612490
RPB2 XV 616671 to 612997
YOR152C XV 618288 to 617518
PDR5 XV 619840 to 624375
SLP1 XV 624729 to 626492
LAS17 XV 675939 to 677840
BFR1 XV 720065 to 718653
Table 2.12 | Genes within additional loci tested by deletion. Each of the above genes
were tested for involvement in rough morphology by deletion in a rough individual.
2.10 Discussion
In summary, we have demonstrated that sets of five or more genetic variants can
synthetically interact to produce major phenotypic effects. Alleles involved in these
higher-order interactions may either be polymorphisms that segregate in natural
populations or spontaneous mutations. Our work also shows that rather than functioning
in a single biochemical pathway, protein complex, or regulatory circuit, the genes
involved in higher-order interactions can play roles in a number of cellular processes.
This finding implies that characterizing higher-order interactions using data from screens
and annotations focused solely on reference genomes may be a challenge, and highlights
34
how genetic variation can serve as a tool for detecting previously unidentified functional
relationships among genes. Further, we have shown that different sets of alleles can
interact to produce the same phenotypic effect. Additional work is necessary to determine
how this latter finding is mediated at the molecular and systems levels. Overall, our study
suggests that characterizing the larger-scale contribution of higher-order interactions to
phenotypic variation is a necessary step in improving our basic understanding of the
genotype-phenotype map.
2.11 Materials and methods
2.11.1 Phenotyping of yeast colony morphology. All phenotyping experiments were
performed on agar plates containing yeast extract and peptone (YP) with 2% ethanol as
the carbon source (YPE). Prior to phenotyping, strains were grown up in liquid YP with
2% dextrose (YPD). Stationary-phase cultures were manually pinned onto YPE and
allowed to grow for five days at 30°C, and were then imaged using a standard digital
camera.
2.11.2 Assessing potential effects of transient heritable factors. Sequencing data from
the rough segregant used in backcross experiments was examined at the chromosome-
scale for evidence of aneuploidy. Average coverage of each chromosome was computed
in R and compared to the genome-wide average. This segregant was also plated at low
density on a large number of YPE plates. We screened tens of thousands of colonies for
instances of phenotypic switching and observed no cases where an individual converted
from rough to smooth morphology.
35
2.11.3 Generation of backcross segregants. Strains used in this paper contained the
Synthetic Genetic Array marker system (Tong et al., 2001), which allowed us to easily
generate large numbers of recombinant MATa progeny. All segregants discussed in the
paper were MATa can1∆::STE2pr-SpHIS5 his3∆ and all backcrosses involved mating
these individuals to either a BY or a 3S strain that was MATα his3∆. In these crosses,
strains with opposite mating types were mixed together on a YPD plate and incubated for
four hours at 30°C. Zygotes were then obtained by microdissection. To generate
segregants, diploids were sporulated at room temperature using the protocol described by
Guthrie and Fink (Guthrie, 1991). Once sporulation had completed, spore cultures were
digested with β-glucuronidase and then plated onto yeast nitrogen base (YNB) plates
containing canavanine, as described previously (Ehrenreich et al., 2010). Spores were
plated at a density of roughly 100 to 200 colonies per plate.
2.11.4 Genome sequencing of backcross segregants. Whole genome sequencing
libraries were prepared using the Illumina Nextera kit, with each of the backcross
segregants barcoded with a unique sequence tag. The libraries were mixed together in
equimolar fractions and sequenced on an Illumina HiSeq machine by the Beijing
Genomics Institute using 100 base pair (bp) x 100 bp reads. Sequencing reads were then
mapped to the S. cerevisiae reference genome using the Burrows-Wheeler Aligner
(BWA) (Li & Durbin, 2009). We used data from 36,756 high confidence SNPs that had
been identified based on comparison of Illumina sequence data for 3S to the BY genome.
Similar to Andolfatto et al. (Andolfatto et al., 2011), we employed Hidden Markov
Models (HMMs) to determine the haplotypes of the segregants based on the sequencing
data. We computed the fraction of reads at each SNP that came from BY and used the
36
vector of these fractions in HMMs that were implemented chromosome-by-chromosome
in the HMM() package of the R statistical programming environment. Any segregants
producing data that showed evidence of contamination, diploidy, or aneuploidy were
excluded from genetic mapping and downstream analyses. Four and eight such
individuals were left out of the BY and 3S mapping populations, respectively.
2.11.5 Genetic mapping. Genotypes inferred from the HMM were used in genetic
mapping analyses. At each position in the genome, we determined the fraction of
individuals that carried the allele from the parent not used in the backcross. We scanned
the genome for alleles from the non-backcross parent that were detected in a large
fraction of segregants. We report loci where these alleles were at 95% frequency or
higher. To determine intervals in which causal genes were located, we identified the
smallest region that was bounded by recombination breakpoints among individuals from
a backcross that shared the same allele at a peak.
2.11.6 Generation and genotyping of dissected tetrads. Backcross diploids were
sporulated and digested in β-glucuronidase to permit tetrad dissection. Standard
microdissection techniques were used to isolate tetrads and separate individual spores.
2.11.7 Fine-mapping using multi-locus introgression strains and dissected backcross
segregants. Haploid multi-locus introgression strains were generated through six rounds
of recurrent backcrossing with phenotypic selection, starting from the same segregant
used in our backcross mapping experiment. Eight of these strains were generated, with
four made by recurrently backcrossing to 3S and four made by recurrently backcrossing
37
to BY. We also used a subset of individuals from the tetrad dissections that showed rough
morphology. To conduct the fine-mapping, we typed these individuals at a number of
markers in each interval using PCR and restriction digestion, or Sanger sequencing.
2.11.8 Genetic engineering experiments. All genes within causal loci were deleted
using the CORE cassette, in the same manner described by Storici et al. (Storici, Lewis,
& Resnick, 2001). Homology tails matching the 60 bases immediately up- and
downstream of each gene were attached to the CORE cassette through PCR and
introduced into cells using the Lithium Acetate method (Gietz & Woods, 2002). Selection
for G418 resistance was used to screen for integration of the CORE cassette; correct
integration was then checked using PCR. All deletions were performed in a haploid
multi-locus introgression strain. To perform complementation tests, deletion strains were
mated to multiple dissected segregants that carried either the causal or non-causal allele
of the deleted gene, as well as the causal alleles at the four other involved genes. The
same phenotyping methods described above were employed for these strains. To generate
allele replacement strains for IRA2, a smooth segregant with the non-causal allele of IRA2
and the causal alleles at the other four loci was transformed using a modified form of
adaptamer mediated allele replacement (Erdeniz, Mortensen, & Rothstein, 1997).
Transformations were conducted with two partially overlapping PCR products—a full-
length amplicon of IRA2 that was tailed at the 3’ end with the 5’ portion of the kanMX
cassette and a copy of the kanMX cassette that was tailed on the 3’ end with part of the
intergenic region downstream of IRA2. Knock-ins were identified using selection on
G418 and verified by Sanger sequencing.
38
2.11.9 Identification of loci that complement the 3S allele of END3. Sequenced strains
from the backcross to 3S were partitioned based on their genotype at END3. We then
screened these individuals for sites where they all carried BY alleles. A group of
additional rough segregants with END3
3S
that had been obtained during tetrad dissections
were genotyped by PCR amplification and restriction digestion of markers across each of
the new loci. One of these additional backcross segregants was mated to 3S, and a panel
of rough progeny from this second-generation backcross were typed at the remaining
candidate loci.
39
Chapter 3: Transcriptional derepression uncovers cryptic higher-order
genetic interactions
This work appears essentially as published in 2015 in
PLoS Genetics, 11(10): e1005606
3.1 Overview
Disruption of certain genes can reveal cryptic genetic variants that do not typically show
phenotypic effects. Because this phenomenon, which is referred to as ‘phenotypic
capacitance’, is a potential source of trait variation and disease risk, it is important to
understand how it arises at the genetic and molecular levels. Here, we use a cryptic
colony morphology trait that segregates in a yeast cross to explore the mechanisms
underlying phenotypic capacitance. We find that the colony trait is expressed when a
mutation in IRA2, a negative regulator of the Ras pathway, co-occurs with specific
combinations of cryptic variants in six genes. Four of these genes encode transcription
factors that act downstream of the Ras pathway, indicating that the phenotype involves
genetically complex changes in the transcriptional regulation of Ras targets. We provide
evidence that the IRA2 mutation reveals the phenotypic effects of the cryptic variants by
disrupting the transcriptional silencing of one or more genes that contribute to the trait.
Supporting this role for the IRA2 mutation, deletion of SFL1, a repressor that acts
downstream of the Ras pathway, also reveals the phenotype, largely due to the same
cryptic variants that were detected in the IRA2 mutant cross. Our results illustrate how
higher-order genetic interactions among mutations and cryptic variants can result in
phenotypic capacitance in specific genetic backgrounds, and suggests these interactions
might reflect genetically complex changes in gene expression that are usually suppressed
by negative regulation.
40
3.2 Introduction
Cryptic genetic variants are standing polymorphisms that only exhibit phenotypic effects
under atypical conditions, such as when specific genes are compromised or the
environment dramatically changes (Gibson & Dworkin, 2004; Hermisson & Wagner,
2004; Paaby & Rockman, 2014). Work in Arabidopsis thaliana (e.g., (Queitsch et al.,
2002; Sangster, Salathia, Lee, et al.; Sangster, Salathia, Undurraga, et al.)),
Caenorhabditis elegans (e.g., (Duveau & Felix, 2012; Felix; Milloz et al.)), Drosophila
melanogaster (e.g., (Dworkin, Palsson, Birdsall, & Gibson; Gibson & Hogness, 1996;
Gibson, Wemple, & van Helden; Rutherford & Lindquist, 1998; Waddington)), multiple
budding yeasts (e.g., (Halfmann et al.; Jarosz & Lindquist, 2010; Tirosh, Reikhav, Sigal,
Assia, & Barkai; True, Berlin, & Lindquist; True & Lindquist, 2000)), and a number of
non-model organisms (e.g., (Berger, Bauerfeind, Blanckenhorn, & Schafer; Kienle &
Sommer, 2013; Lauter & Doebley, 2002; Ledon-Rettig, Pfennig, & Crespi; Rohner et al.;
Rosas, Barton, Copsey, Barbier de Reuille, & Coen; Suzuki & Nijhout, 2006)) has shown
that cryptic variation is abundant within and between species. Because it is so prevalent,
cryptic variation could plausibly contribute to adaptation and phenotypic novelty
(Ehrenreich & Pfennig, 2015; Le Rouzic & Carlborg, 2008; Moczek; Paaby & Rockman,
2014), as well as to disease susceptibility (Gibson). Yet due to their entirely conditional
phenotypic effects, cryptic variants have proven difficult to study and are not understood
as well as other classes of polymorphisms. In particular, the genetic and molecular
mechanisms that suppress and uncover cryptic variation have yet to be fully determined.
For the purposes of this paper, we focus on the mechanisms by which functional
disruption of specific ‘capacitor’ genes exposes the phenotypic effects of cryptic variants.
41
This phenomenon is often referred to as ‘phenotypic capacitance’ or ‘evolutionary
capacitance’, though for simplicity we refer to it as ‘capacitance’ (Rutherford &
Lindquist, 1998) (Bergman & Siegal, 2003). The first described capacitor was Hsp90, a
chaperone that assists in the folding and stabilization of other proteins (Rutherford &
Lindquist, 1998) (Sangster, Lindquist, & Queitsch). Early research on capacitance
suggested that Hsp90 might have distinct biochemical features that cause cryptic
variation to be uncovered when it is compromised (Queitsch et al., 2002; Rutherford &
Lindquist, 1998) (Sangster et al.). However, subsequent theoretical work showed that
capacitance most likely occurs as a general consequence of gene regulatory network
perturbation and that many genes might be able to act as capacitors (Bergman & Siegal,
2003). Supporting this finding, a number of genes involved in chromatin regulation have
also been shown to be capacitors of cryptic variation (Tirosh et al.) (Richardson,
Uppendahl, Traficante, Levy, & Siegal) (Sollars et al.) and to even phenocopy the effects
of Hsp90 perturbation (Sollars et al.).
More recent work suggests that capacitance depends not only on the perturbation of
capacitors but also on the specific cryptic variants that are present. This is because cryptic
variants themselves can play an important role in capacitance by genetically interacting
with and ‘potentiating’ the phenotypic effects of their capacitors (Hermisson & Wagner,
2004) (Blount, Borland, & Lenski; Cowen & Lindquist, 2005; Jarosz & Lindquist, 2010;
Richardson et al.). The genetic architecture of this potentiating cryptic variation has not
been characterized in detail (Dworkin), but may involve complex epistatic interactions
between multiple cryptic variants and capacitating mutations (i.e., higher-order genetic
interactions) (Taylor & Ehrenreich, 2015b). In such a scenario, the phenotypic effect of a
42
given capacitating mutation would depend on the cryptic variants with which it co-occurs,
with the mutation having an effect only in certain genetic backgrounds (Chandler et al.)
(Fig. 3.1). This possibility is not unfounded, as several recent studies suggest that genetic
background effects can involve higher-order genetic interactions among de novo or
induced mutations and sets of cryptic variants (Chandler, Chari, Tack, & Dworkin;
Dowell et al.; Taylor & Ehrenreich, 2014).
Fig. 3.1 | Capacitance, higher-order genetic interactions, and genetic background
effects might be related phenomena that involve interactions among capacitating
mutations and cryptic variants. ‘YFG’ and ‘yfg∆’ refer to the wild type and mutant
alleles of a gene that can genetically interact with cryptic variants. The green yeast
indicates the combination of a capacitating mutation and cryptic variants that
shows a phenotypic effect.
We recently described an experimental system that can be used to study how higher-order
genetic interactions among mutations and cryptic variants result in capacitance (Taylor &
Ehrenreich, 2014). In our previous paper, we showed that a de novo mutation in IRA2, a
43
negative regulator of the Ras-cAMP-PKA (Ras) pathway (Tanaka, Nakafuku, Tamanoi,
et al.), uncovers sets of interacting cryptic variants that influence colony morphology in
Saccharomyces cerevisiae. This mutation (ira2∆2933) occurred spontaneously while we
were generating a cross of the lab strain BY4716 (‘BY’) and a derivative of the clinical
isolate 322134S (‘3S’) (Liti et al.; Schacherer, Shapiro, Ruderfer, & Kruglyak), and
results in a truncated, partially functional Ira2 protein that lacks 117 amino acids relative
to its wild type form. When the ira2∆2933 lesion is present in specific haploid
recombinants in the BYx3S cross, it causes a change in colony morphology from ‘smooth’
to ‘rough’ (Fig. 3.2).
Fig. 3.2 | Colony morphology phenotypes that occur in the BYx3S cross in the
presence of ira2∆2933. BY, 3S, and most segregants show a smooth phenotype, while
a small fraction of segregants show a rough phenotype.
Through comprehensive genetic mapping experiments, we showed that ira2∆2933
induces the rough phenotype when it co-occurs with specific combinations of cryptic
variants at four or more genes (Taylor & Ehrenreich, 2014). To better understand these
44
higher-order genetic interactions, we cloned all of the genes involved in one of the
combinations. This resulted in the identification of two transcriptional activators that
heterodimerize and function downstream of the Ras pathway (FLO8 (Kobayashi et al.)
and MSS11 (Gagiano et al.)), a structural protein that plays a role in vesicle formation
(END3 (Benedetti et al.; Tang et al.)), and an enzyme that helps cells detoxify themselves
of endogenous redox stress (TRR1 (Pedrajas et al.)). Most of the rough individuals in our
past study had the genotype END3
BY
FLO8
3S
ira2∆2933 MSS11
BY
TRR1
3S
. However, we
also provided evidence for a more complex genotype involving END3
3S
that requires
specific alleles at two additional loci.
In this paper, we complete our efforts to determine the genetic basis of ira2∆2933-
dependent rough morphology in the BYx3S cross under our standard assay conditions.
We show that in addition to the previously identified five-way genetic interaction, a six-
way interaction can also cause the trait. Specifically, individuals with the genotype
END3
3S
FLO8
3S
ira2∆2933 MSS11
BY
exhibit the rough phenotype if they possess BY
alleles at two other transcription factors that are regulated by the Ras pathway (Pan &
Heitman, 2002; Robertson & Fink, 1998): the activator MGA1 (Lorenz & Heitman, 1998)
and the repressor SFL1 (Conlan & Tzamarias, 2001; Fujita et al.). This suggests that the
rough phenotype arises due to genetically complex changes in the regulation of Ras target
genes. We examine the role of ira2∆2933 in these regulatory changes and find that it
alleviates the silencing of FLO11, a gene that encodes a cell surface protein required for
rough morphology. We also show that this ability to disrupt FLO11 repression is not
unique to IRA2. These results illustrate how higher-order combinations of cryptic variants
can confer the potential for capacitance to specific genetic backgrounds and indicate that
45
capacitating mutations may reveal cryptic phenotypic potential by causing transcriptional
derepression.
3.3 END3
3S
and ira2∆2933 are involved in a six-way genetic interaction
To determine the specific combination of alleles involved in rough morphology in an
END3
3S
background, we generated new mapping populations by mating an END3
3S
rough segregant from a (BYx3S)x3S backcross to BY and 3S (Methods). Throughout the
paper, the term ‘backcross’ refers specifically to these ((BYx3S)x3S)xBY and
((BYx3S)x3S)x3S matings. Because END3
3S
segregated in the BY backcross, we
genotyped rough individuals recovered from this population to determine the allele of
END3 they carried (Methods). In total, we obtained 63 and 88 rough END3
3S
individuals
from the BY and 3S backcrosses, respectively. We then pooled cells from these rough
individuals by cross and performed bulk segregant mapping by sequencing (Ehrenreich et
al.; Michelmore, Paran, & Kesseli) (Methods). We found that the more complex genetic
interaction involves a specific combination of alleles at six loci, with individual loci
detected on Chromosomes V, VII, XIII, and XIV, and two loci identified on
Chromosome XV (Fig. 3.3A and B). The chromosome XIV locus corresponds to END3
3S
,
while allele replacements in a backcross segregant that carried the six-way interaction
confirmed that FLO8
3S
, MSS11
BY
, and ira2∆2933 underlie the Chromosome V, XIII, and
XV-1 loci, respectively (Fig. 3.3C and Fig. 3.4; Methods). The new mapping data also
allowed us to delimit the Chromosome VII and XV-2 loci, which we were unable to
clone in our prior study (Taylor & Ehrenreich, 2014), to a single gene (MGA1) and five
genes (SFL1, ARP8, LSC1, SUF5, THI80), respectively. We used allele swaps to show
that the BY alleles of MGA1 and SFL1, which respectively encode an activator and a
46
repressor that are regulated by the Ras pathway, are the causal alleles at these loci (Fig.
3.4). These results show the six-way interaction occurs in individuals with the genotype
END3
3S
FLO8
3S
ira2∆2933 MGA1
BY
MSS11
BY
SFL1
BY
(Fig. 3.3B). Thus, the differences
between the five- and six-way interactions involve which END3 allele is involved and
whether specific alleles of MGA1, SFL1, and TRR1 are required (Fig. 3.3B).
Fig. 3.3 | Characterization of the six-way genetic interaction. (A) Allele frequency
plots for BY and 3S second iteration backcross populations of END3
3S
rough strains.
Fixed loci are denoted with a blue, orange, or grey bars depending on whether the
BY, 3S, or mutant alleles, respectively, were detected at a locus. The allele
frequencies were estimated by averaging data in sliding windows containing 10
SNPs. (B) Cryptic variants involved in the five- and six-way interactions. (C)
Dependence of both genetic interactions on the ira2∆2933 mutation.
47
Fig. 3.4 | Allele replacement results for FLO8
BY
, MGA1
3S
, MSS11
3S
, and SFL1
3S
in
the six-way genetic interaction. The role of END3
3S
was verified in (Taylor &
Ehrenreich, 2014), while the effect of ira2∆2933 in this background is shown in Fig.
3.2C.
3.4 The two interactions fully account for rough morphology in the presence of
ira2∆2933
Based on our genetic mapping results in this paper and our past work (Taylor &
Ehrenreich, 2014), we have identified alleles of six genes (END3, FLO8, MGA1, MSS11,
SFL1, TRR1) that genetically interact in two different combinations with ira2∆2933 (Fig.
3.3B and C). We tested whether these two allele combinations fully explain rough
morphology in the BYx3S ira2∆2933 cross by generating a new BYx3S cross in which
3S carried ira2∆2933 (Methods). As our past work focused on matings of segregants to
BY or 3S, this population enabled us to test for the first time the effects of all possible
combinations of BY and 3S alleles in the presence of ira2∆2933. Among 42 rough
individuals that we recovered, 40 (95.2%) carried the five-way interaction, while two
(4.8%) carried the six-way interaction. The five-way interaction should occur twice as
often as the six-way interaction, yet the observed ratio was 20:1. This may be due to
linkage between END3 and a locus at which the BY allele confers a strong selective
advantage during random spore isolation (see Figure S2B from (Taylor & Ehrenreich,
48
2014)). Alternatively, the enrichment of rough individuals carrying the five-way
interaction could simply have occurred because the sample of rough individuals in this
experiment was small. Nevertheless, our observation that all the examined rough
individuals harbored either the five- or six-way interactions suggests that we have
completely determined the genetic basis of rough morphology in the BYx3S ira2∆2933
cross under our experimental conditions.
3.5 FLO11 expression is needed for rough morphology
Rough morphology in the BYx3S cross likely arises due to genetically complex changes
in the regulation of Ras target genes. Such a possibility is supported by the finding that
four Ras-regulated transcription factors (Robertson & Fink, 1998) harbor cryptic variants
involved in the rough phenotype, as well as by the fact that these cryptic variants are
revealed by a capacitating mutation in IRA2, a negative regulator of Ras signaling. A
gene that is likely influenced by these genetic factors is FLO11, which encodes a cell
surface glycoprotein that facilitates cell-cell adhesion and is thought to be regulated by
Flo8-Mss11, Mga1, and Sfl1 (Bruckner & Mosch, 2012; Lo & Dranginis, 1996). To
determine if expression of the rough phenotype due to the five- and six-way interactions
requires FLO11, we deleted the gene from a nearly isogenic line possessing the five-way
interaction and a backcross segregant carrying the six-way interaction (Methods). This
was sufficient to convert both of these strains from rough to smooth (Fig. 3.5A),
indicating that both genetic interactions are FLO11-dependent. RT-PCR showed that
FLO11 is expressed in individuals carrying the five- and six-way interactions, but not in
BY or 3S (Fig. 3.5B; Methods). These results suggest expression of the rough phenotype
requires active transcription of FLO11.
49
Fig. 3.5 | FLO11 is required for rough morphology and shows differential expression
across genetic backgrounds. (A) Deletion of FLO11 leads to smooth morphology in
both the five- and six-way genetic interaction backgrounds. (B) RT-PCR of FLO11
and the housekeeping gene ACT1 in multiple genetic backgrounds. FLO11 is not
expressed in BY or 3S, but is expressed in recombinants that carry the five- and six-
way genetic interactions. FLO11 is also expressed in 3S ira2∆2933 and 3S sfl1∆
strains.
3.6 ira2∆2933 and SFL1 deletion cause FLO11 expression in 3S
We tested whether ira2∆2933 influences FLO11 expression by introducing the lesion
into BY and 3S, and conducting RT-PCR (Methods). Each strain remained smooth after
this manipulation, which was expected because they both lack a complete set of alleles
that can give rise to rough morphology. Furthermore, BY ira2∆2933 did not express
FLO11, likely because this strain carries a nonsense allele of FLO8, the major
transcriptional activator of FLO11 (Liu et al.). However, introduction of ira2∆2933 into
3S, which possesses a functional allele of FLO8, converted FLO11 from a silenced to an
actively transcribed state (Fig. 3.5B). Given that ira2∆2933 alleviated repression of
FLO11 in 3S, we hypothesized that it might do so by indirectly inhibiting Sfl1, which is
thought to negatively regulate FLO11 and other targets of the Ras pathway when Ras
signaling is low by recruiting the Ssn6-Tup1 corepressor complex (Conlan & Tzamarias,
2001), which in turn recruits the histone deacetylase Hda1 (Halme et al.; Wu, Suka,
Carlson, & Grunstein). To test this possibility, we deleted SFL1 from 3S. This knockout
50
phenocopied the results of introducing ira2∆2933: 3S remained smooth, but expressed
FLO11 (Fig. 3.5B). This suggests that ira∆2933 disrupts Sfl1-mediated transcriptional
repression of Ras target genes.
3.7 Cryptic genetic variation uncovered by SFL1 deletion
To test whether loss of transcriptional repression by Sfl1 is sufficient to reveal the cryptic
higher-order genetic interactions that specify rough morphology, we generated new
BYx3S crosses. We first created a BYx3S cross that lacked the IRA2 mutation and
screened for rough morphology among thousands of recombinants (Methods). All
segregants in this cross were smooth. We then constructed a cross in which BY and 3S
carried wild type alleles of IRA2, but had SFL1 deleted (Methods). Rough morphology,
as well as a ‘bumpy’ intermediate phenotype that we previously reported (see Figure S4D
and Table S1 in (Taylor & Ehrenreich, 2014)), segregated in this sfl1∆ cross (Fig. 3.6A).
Genotyping of 44 rough sfl1∆ segregants showed that the rough phenotype is expressed
in the ira2∆2933 and sfl1∆ backgrounds largely due to the same cryptic variants
(Methods). 43 (98%) of the rough sfl1∆ segregants possessed the genotype END3
BY
FLO8
3S
MSS11
BY
TRR1
3S
, which also potentiates the five-way interaction involving
ira2∆2933 (Fig. 3.6B). The other rough sfl1∆ segregant had the genotype END3
BY
FLO8
3S
MSS11
BY
TRR1
BY
, which does not give rise to rough morphology in the presence
of ira2∆2933 (Fig. 3.6B). None of the rough sfl1∆ segregants had a genotype resembling
the six-way interaction involving ira2∆2933. This could have occurred because SFL1
BY
,
which is required for the six-way interaction, is missing from the sfl1∆ cross; our
sampling was biased due to the selectively advantageous locus that is linked to END3; or,
as the detection of a rough sfl1∆ segregant with the END3
BY
FLO8
3S
MSS11
BY
TRR1
BY
51
genotype also suggests, ira2∆2933 and sfl1∆ have similar but not identical molecular
effects. Despite these differences between the ira2∆2933 and sfl1∆ crosses, our results
clearly show that transcriptional repression normally suppresses rough morphology and
that multiple genes can act as capacitors by disrupting this negative regulation.
Fig. 3.6 | Deletion of SFL1 reveals interacting cryptic variants. (A) Three phenotypic
classes—smooth, bumpy, and rough—were observed among progeny from the
BYx3S sfl1∆ cross. The proportion of segregants observed in each phenotypic class
is shown below representative pictures for each class. (B) Genotypes observed
among rough progeny from the BYx3S sfl1∆ cross.
3.8 Discussion
Across this manuscript and our previous paper (Taylor & Ehrenreich, 2014), we have
cloned six genes that harbor cryptic variants that interact in two specific allele
combinations to determine the phenotypic effect of ira2∆2933. These two genetic
backgrounds can be viewed as potentiating genotypes that facilitate the expression of
rough morphology in the presence of a capacitating mutation, such as ira2∆2933. This
finding is important because it shows sets of cryptic variants can genetically interact with
52
each other and their capacitating mutation, and implies a conceptual link between
capacitance, higher-order genetic interactions, and genetic background effects (Fig. 3.1).
Given that four of the identified genes encode transcription factors, our work suggests
complex gene regulatory changes underlie the expression of rough morphology in the
BYx3S cross. This finding is consistent with theoretical results that have shown an
important role for gene regulatory network perturbation in capacitance (Bergman &
Siegal, 2003) and higher-order genetic interactions (Gjuvsland, Hayes, Omholt, &
Carlborg). In our specific case, the role of ira2∆2933 is likely to cause transcriptional
derepression, which may enable the involved cryptic variants to collectively alter the
gene regulatory network underlying colony morphology. Supporting such a role for
derepression in the rough phenotype, we have shown that IRA2 is not unique in its ability
to act as a capacitor. Rather, SFL1 can also serve as a capacitor of rough morphology,
presumably because its deletion also causes transcriptional derepression.
Moving forward, fully understanding capacitance in the BYx3S colony morphology
system will likely require defining the gene regulatory network underlying rough
morphology and determining how it changes across combinations of cryptic variants and
capacitating mutations. Such work can shed light on the individual and collective
contributions of the identified cryptic variants to the rough phenotype; may reveal why
MGA1
BY
, SFL1
BY
, and TRR1
3S
only have phenotypic effects in specific END3
backgrounds; and might further clarify how multiple genes can act as capacitors of the
same cryptic variants and trait. More generally, research along these lines has the
53
potential to provide basic insights into how genetically complex, cryptic phenotypes are
suppressed and uncovered.
Additionally, to our knowledge, the present study, when considered with (Taylor &
Ehrenreich, 2014), represents the first comprehensive genetic characterization of a
genetic background effect in any organism. Our work demonstrates how genetic
background effects can arise due to complex epistatic relationships between mutations
and cryptic variants at multiple modifier loci, as others have previously suggested
(Chandler et al.). Our findings also indicate that multiple epistatic configurations of
cryptic variants may enable a given mutation to show a phenotypic effect. Although these
results advance understanding of the causes of genetic background effects, determining
the generality of these findings will require dissecting other genetic background effects
that involve different mutations, species, and traits.
3.9 Materials and Methods
3.9.1 Phenotyping of yeast colony morphology. All phenotyping experiments were
performed on agar plates containing yeast extract and peptone (YP) with 2% ethanol as
the carbon source (YPE). Prior to phenotyping, strains were grown to stationary phase in
liquid YP with 2% dextrose (YPD). Cultures were manually pinned onto YPE and
allowed to grow for five days at 30°C, and were then imaged using a standard digital
camera.
3.9.2 Generation of backcross segregants. Strains with opposite mating types were
mixed together on a YPD plate and incubated for four hours at 30°C. A zygote from each
54
cross was obtained by microdissection. To generate segregants, diploids were sporulated
at room temperature using standard yeast sporulation procedures (Sherman). Once
sporulation had completed, spore cultures were digested with β-glucuronidase and then
plated onto YPE plates at a density of roughly 100 to 200 colonies per plate.
Approximately 10 plates were screened per backcross.148 (BY backcross) and 88 (3S
backcross) rough segregants were picked manually and streaked to obtain single cell
isolates. The mating type of each of these strains was checked to confirm that they were
indeed haploid. Segregants from the BY backcross could be either END3
BY
or END3
3S
. In
order to ensure sequenced strains possessed the END3
3S
allele, each segregant was
genotyped using a nearby restriction marker. 63 of the 148 BY backcross progeny
possessed the END3
3S
allele and were used for genetic mapping. We note that other
multicellularity phenotypes (e.g., flocculation) segregated in the backcrosses, but were
not strongly correlated with expression of the rough phenotype, implying they have
different genetic architectures.
3.9.3 Generation of IRA2 wild type, ira2∆2933, and sfl1∆ crosses. The BY and 3S
strains used in the ira2∆2933 and sfl1∆ crosses possessed the Synthetic Genetic Array
marker system (Tong et al.), which allowed for generation of large numbers of
recombinant MATa progeny. Regarding the IRA2 wild type cross, we re-mated BY and
3S to produce a different diploid than the one used in (Taylor & Ehrenreich, 2014). For
the ira2∆2933 cross, the lesion was introduced into 3S using allele replacement
techniques described below and then this 3S ira2∆2933 strain was mated to a wild type
BY strain. We designed the cross in this way because the ira2∆2933 mutation originally
occurred in the 3S allele of the gene. However, we note that we have never seen evidence
55
for a genetic interaction between ira2∆2933 and other genetic variants in IRA2
3S
. As for
the sfl1∆ cross, we constructed BY and 3S strains that lacked the entire coding region of
SFL1 using genetic engineering techniques described below. A BY/3S sfl1∆/sfl1∆ diploid
was then used to generate a population of BYx3S sfl1∆ recombinants. For each of the
three crosses described in this section, diploids were generated and sporulated as
described for the backcrosses, but sporulations were plated at low density onto YNB
plates containing canavanine to select for haploid progeny. These were then replica
plated on YPE to phenotype colony morphology. For each cross, around 20 plates
containing roughly 100 to 200 colonies were screened.
3.9.4 Bulk segregant mapping of rough morphology in the backcrosses. Each rough
END3
3S
segregant from the backcrosses was grown to stationary phase as an individual,
clonal culture. Cells from these stationary cultures were then mixed in equimolar
fractions by backcross and DNA was extracted from the two pools using Qiagen G-tip
columns. Whole genome sequencing libraries were prepared using the Illumina Nextera
kit, with each of the backcross segregant pools barcoded with a unique sequence tag. The
libraries were mixed together in equimolar fractions and sequenced on an Illumina MiSeq
machine by the company Laragen, Inc. using 250 base pair (bp) x 250 bp reads. These
sequencing reads, which have been uploaded to the NCBI Sequence Read Archive (Study
Accession SRP062432) under the accession numbers SAMN03956543 (BY backcross)
and SAMN03956544 (3S backcross), were then mapped to the S. cerevisiae S288c
reference and 322134S draft genomes (http://www.yeastgenome.org). S288c is the
progenitor of BY, and to ensure high quality read mapping, reads from the BY and 3S
backcrosses were mapped to S288c and 3S, respectively. Alignments were performed
56
using the Burrows-Wheeler Aligner (BWA) version 7 with options mem -t 20 (Li &
Durbin, 2009). Based on these alignments, we obtained 73- and 122-fold genomic
coverage, as determined by the average per site coverages, from the BY and 3S backcross
populations, respectively. A custom Python script was used to assess genome-wide allele
frequencies at 36,756 high confidence SNPs that had previously been identified by
mapping Illumina sequencing reads for 3S to the S288c genome (Taylor & Ehrenreich,
2014). Loci influencing colony morphology were called as regions enriched at 95%
frequency or higher when the data were averaged within running windows of 10 SNPs.
Intervals containing causal genes were identified in the R statistical programming
environment as the smallest regions that had mean allele frequencies above a threshold of
95%. Subsequent restriction typing experiments focused on individual segregants and the
selected loci showed that the detected loci were in fact fixed, and that deviations from
fixation occurred due to the presence of a small number of sequencing or read mapping
errors.
3.9.5 Genetic engineering experiments. To generate allele replacement strains for ARP8,
LSC1, MGA1, SFL1, SUF5, and THI80, a backcross segregant that expressed rough
morphology due to the six-way genetic interaction was transformed using a modified
form of adaptamer-mediated allele replacement (Erdeniz et al.). Also, adaptamer-
mediated allele replacement was used to introduce the ira2∆2933 lesion into 3S.
Transformations were conducted with two partially overlapping PCR products—a full-
length amplicon of the gene of interest that was tailed at the 3’ end with the 5’ portion of
the kanMX cassette and a copy of the kanMX cassette that was tailed on the 3’ end with
part of the intergenic region downstream of the gene (as shown in Figure S1 of (Matsui,
57
Linder, Phan, Seidl, & Ehrenreich)). Knock-ins were identified using selection on G418
and verified by Sanger sequencing. Deletions were constructed using the CORE cassette
(Storici et al.). Homology tails matching the 60 bases immediately up- and downstream
of each gene were attached to the CORE cassette through PCR and introduced into cells
using the Lithium Acetate method (Gietz & Woods, 2002). Selection for G418 resistance
was used to screen for integration of the CORE cassette; correct integration was then
checked using PCR. SFL1 was deleted from BY and 3S, while FLO11 was deleted from a
nearly isogenic line and a backcross segregant harboring the five- and six-way genetic
interactions, respectively.
3.9.6 Genotyping of causal alleles in ira2∆2933 and sfl1∆ crosses. Markers within
END3, FLO8, MGA1, MSS11, SFL1, and TRR1 were genotyped using PCR and
restriction digestion. These markers were identified from among the 36,756 high
confidence SNPs that differentiate BY and 3S.
3.9.7 RT-PCRs. Strains were grown to stationary phase in liquid YPD media at 30
o
C and
pinned on to YPE agar plates. After four days of growth at 30
o
C, total RNA was
extracted with the Qiagen RNeasy kit. cDNA was then generated with Superscript
reverse transcriptase from Life Technologies. ACT1, a well-known housekeeping gene,
was used as a control for our FLO11 RT-PCRs. Strains that were used in the RT-PCR
experiments are described in the main text. The specific primers that we used were taken
from (Fichtner, Schulze, & Braus).
58
Chapter 4: Genetic architectures of phenotypic capacitance
This work is presented as submitted for publication in 2015
The author list is as follows:
Matthew B. Taylor, Joann Phan, Jonathan T. Lee, Madelyn McCadden, and Ian M.
Ehrenreich
4.1 Overview
Cryptic genetic variants that do not typically influence traits can interact with each other
and mutations to unexpectedly show phenotypic effects(Gibson & Dworkin; Paaby &
Rockman). The genetic and molecular basis of this phenomenon, which is known as
‘phenotypic capacitance’ (or ‘capacitance’)(Bergman & Siegal; Rutherford), remains
incompletely understood. Here, we provide detailed new insights into capacitance by
genetically dissecting 17 independent instances of the same cryptic phenotype in a yeast
cross. Roughly half of the cases involve a spontaneous mutation in a ‘capacitor’
gene(Queitsch et al.; Rutherford) uncovering sets of cryptic variants. These capacitating
mutations differ in the total number and specific identities of cryptic variants they interact
with to induce the same phenotypic change. However, we also show that capacitance can
arise without functional disruption of capacitors due to higher-order genetic
interactions(Taylor & Ehrenreich, 2015a) that only involve cryptic variants. Our results
may be relevant to other species and disease(Eichler et al.), as most of the capacitating
mutations and cryptic variants identified in our study are components of the
evolutionarily conserved and oncogenic Ras signaling pathway(Cox & Der).
4.2 Introduction
Populations harbor a large amount of standing genetic variation that only shows
phenotypic effects under atypical genetic or environmental conditions(Gibson &
59
Dworkin; Paaby & Rockman). Theoretical and empirical evidence has shown these
cryptic variants can interact with each other, as well as with functional perturbations of
capacitors, to cause individuals to show genetically complex traits that are not normally
expressed(Hermisson & Wagner; Le Rouzic & Carlborg; Richardson et al.; Taylor &
Ehrenreich, 2014, 2015b). Despite increasing awareness of this potentially important
phenomenon(Gibson & Dworkin; Paaby & Rockman), the architectures of cryptic
variants and genetic perturbations that can collectively result in capacitance have not yet
been comprehensively explored(Dworkin).
Multiple pieces of information are needed to improve understanding of capacitance’s
genetic basis. First, the capacitors that can reveal cryptic phenotypes must be broadly
determined, as system-level modeling suggests that many genes should be capable of
uncovering cryptic variation when perturbed(Bergman & Siegal). Second, the
combinations of cryptic variants that enable these capacitors to show phenotypic effects
when compromised needs to be assessed, as distinct sets of cryptic variants might
potentiate the phenotypic effects of different capacitors(Jarosz & Lindquist; Taylor &
Ehrenreich, 2015b). Third, if multiple configurations of capacitating mutations and
cryptic variants result in capacitance, the differences among these genetic architectures
should be related to a phenotype’s underlying gene regulatory network(Bergman &
Siegal; Gjuvsland et al.) to determine how they exhibit the same phenotypic effect.
We have developed an experimental system in the budding yeast Saccharomyces
cerevisiae that can be used to screen for and genetically dissect examples of
capacitance(Lee, Taylor, Shen, & Ehrenreich, Submitted; Taylor & Ehrenreich, 2014,
60
2015b). Specifically, the lab strain BY4716 (‘BY’)(Liti et al., 2009), a haploid derivative
of the clinical isolate 322134S (‘3S’)(Liti et al., 2009), and their wild type recombinant
progeny show a ‘smooth’ colony phenotype(Taylor & Ehrenreich, 2014). However,
spontaneous and induced mutations can cause certain BYx3S segregants to show an
alternative, ‘rough’ morphology(Lee et al., Submitted; Taylor & Ehrenreich, 2014,
2015b) (Fig. 4.2A). In the best characterized example of capacitance in this system, we
demonstrated that a spontaneous mutation in the Ras negative regulator IRA2
(ira2∆2933) can interact with specific combinations of four or five cryptic variants to
cause the rough phenotype under standard lab conditions(Taylor & Ehrenreich, 2014,
2015b).
In more recent work, we have shown that the genetic architecture of capacitance
significantly varies in complexity across environments(Lee et al., Submitted). For
instance, a specific combination of seven or more cryptic variants is needed to enable
ira2∆2933 to cause the rough phenotype at 37°C. However, at room temperature, a broad
space of BYx3S ira2∆2933 genotypes show rough morphology and evidence suggests
that some wild type segregants might even express the phenotype in this condition. This
permissive nature of the low temperature environment provides an excellent opportunity
to generally examine the space of genetic architectures that can capacitate the same trait.
4.3 Experimental design
In this paper, we take advantage of the rough colony morphology system to conduct the
first large-scale examination of the genetic architectures underlying capacitance. We
began this effort by screening more than 100,000 haploid segregants derived from 106
61
different matings of BY and 3S at room temperature (Methods). Through this screen, we
obtained 17 rough segregants that were descended from different BY/3S diploids and
thus represented biologically independent examples of capacitance (Fig. 4.1; Methods).
To determine the genetic basis of these individuals’ phenotypes, we backcrossed each
rough segregant to both BY and 3S. Bulk segregant mapping by sequencing(Ehrenreich
et al., 2010; Michelmore et al.) was performed on pools containing between 46 and 95
rough F
2
B progeny (Fi.g 1b). Backcross pools were sequenced to an average genomic
coverage of 244X (Methods). Control pools were also generated for each backcross using
lawns of BYx3S segregants that had not been selected for colony morphology (Methods).
These control pools were sequenced to an average genomic coverage of 139X (Methods).
Fig. 4.1 | Phenotypes of 17 rough segregants used in this study.
62
4.4 Diverse architectures underlying capacitance
We separately analyzed the rough and control populations for each backcross using
MULTIPOOL(Edwards & Gifford) and excluded any loci detected in the control pools
from further consideration (Methods). Based on this procedure, we detected between two
and 10 loci per rough segregant, with an average number of 6.6 loci identified. These
detected genomic regions correspond to eight de novo mutations and 18 distinct genomic
regions that harbor cryptic variants, which are referred to as ‘loci’ hereafter (Fig. 4.2C;
Fig. 4.3; Methods). Only one of these cryptic alleles, which we previously identified as
the 3S version of the Ras-regulated transcriptional activator FLO8(Taylor & Ehrenreich),
was found to play a role in capacitance in all cases (Fig. 4.2D). Given that BY carries a
null allele of FLO8(Liu et al., 1996; Matsui et al.), this finding indicates that a functional
copy of FLO8
is necessary for expression of the rough phenotype. The remaining loci
were detected on average 4.1 times, with a range between 1 and 15.
63
Fig. 4.2 | Summary of genetic mapping data. A Representative images of rough and
smooth colonies in the BYx3S cross. B Genetic mapping strategy to determine genetic
architecture of capacitance. C Genetic mapping data from each rough segregant is shown
horizontally. Vertical bars represent detected loci, with blue and orange coloring
indicating cryptic variants from BY and 3S, respectively. Spontaneous mutations that are
required for individual instances of capacitance are shown in red, with the specific
mutated gene noted to the right. The mapping results for each rough segregant correspond
to data from backcrosses of each segregant to BY and 3S. The allele frequencies of
detected loci among rough individuals in the backcross in which they were detected is
shown in the color scale to the right of the figure. D The number of times each cryptic
variant was detected across the different mapping populations. Results corresponding to
capacitating mutations was excluded from this last panel.
64
Fig. 4.3 | Cloning of capacitating mutations in 8 rough segregants carrying de novo
mutations. Genes containing candidate capacitating mutations in segregants 1-8 were
replaced with the mutant (m), BY (BY), or 3S allele (3S).
4.5 Characterization of identified capacitating mutations
The eight de novo lesions that induced capacitance were comprised of six small
spontaneous deletions and two point mutations (Fig. 4.4; Table 4.1). The six genes
harboring these lesions fell into three functional classes (Table 4.1): negative regulators
of Ras signaling (GPB1, IRA1, and IRA2), non-essential components of RNA polymerase
II that act in the Mediator complex (SSN3 and SSN8), and a gene of unknown function
that localizes to bud tips during cell division (IRC8)(Cherry et al.). Based on gene
deletion experiments (Methods), seven of the mutations are complete loss-of-function
alleles (Fig. 4.5). The only partial loss-of-function mutation was a frameshift in GPB1
(Fig. 4.5; Table 4.1). This is consistent with our previous finding that the capacitating
mutation ira2∆2933 confers partial loss of function(Taylor & Ehrenreich, 2014), as Gpb1
and Ira2 physically interact to coregulate Ras signaling and both of the partial loss-of-
65
function lesions disrupt protein-protein interaction domains needed for Gpb1-Ira2
binding(Phan et al.) (Table 4.1).
Fig. 4.4 | Individuals 2 and 3 possess the same SSN8 lesion, but are from different
matings of BY and 3S. Genome-wide haplotypes are shown below mapping data from
Fig. 4.2C. Blue horizonal bars represent regions in which the segregant possessed the BY
allele, and orange bars represent regions in which the segregant possessed the 3S allele.
Fig. 4.5 | Deletion of mutant alleles from cognate rough segregants harboring de
novo lesions.
66
Mutant
gene
Function of gene Lesion Functional
consequence
Impact on gene
sequence
GPB1 Multistep
regulator of
cAMP-PKA
signaling
Single nucleotide
mutation
720G->T
Loss-of-
function
May disrupt Kelch
domain (Phan et al.)
IRA1 GTPase-activating
protein; negatively
regulates RAS
Single base
deletion
1160∆G
Null Nonsense; Loss of
Gpb2 binding site and
GAP-related domain
(Harashima, Anderson,
Yates, & Heitman);
Truncates 2694 amino
acids, 87% of protein
IRA1 GTPase-activating
protein; negatively
regulates RAS
Single base
deletion
7719∆G
Null Nonsense; Loss of
Gpb2 binding site and
disruption of GAP-
related domain
(Harashima et al.);
Truncates 503 amino
acids, 16% of protein
IRA2 GTPase-activating
protein; negatively
regulates RAS
Single nucleotide
mutation
-60C->T
Null Likely disrupts TBP
binding site (Venters et
al.)
IRA2* GTPase-activating
protein; negatively
regulates RAS
Single base
deletion
8801∆A
Loss-of-
function
Nonsense; Loss of
Gpb1 binding site
(Harashima et al.);
Truncates 117 amino
acids, 4% of protein
IRC8 Bud tip localized
protein
Single base
deletion
938∆A
Null Nonsense; Truncates
487 amino acids, 59%
of protein
SSN3 Cyclin-dependent
protein kinase;
component of
RNA pol II
holoenzyme
Single base
deletion
111∆T
Null Nonsense; Truncates
508 amino acids, 92%
of protein
SSN8 Cyclin-like
component of
RNA polymerase
II holoenzyme
22 base deletion
40∆22
Null Nonsense; Loss of
cyclin box domain
(Kuchin, Yeghiayan, &
Carlson); Truncates 291
amino acids, 90% of
protein
Table 4.1 | Information regarding capacitating mutations. Functional consequences
were assessed by deleting the mutant allele from the rough segregant carrying a given
lesion. We note the same SSN8 mutation occurred twice. These likely represent
independent occurrences of the same lesion or the presence of the mutation in cultures
used to generate diploids (Fig. 4.4).
67
When considered with our previous work in which IRA2 and the Ras-regulated
transcriptional repressor SFL1 were identified as capacitors(Taylor & Ehrenreich,
2015b), the current study brings the total number of identified capacitors of rough
morphology to seven. The identities of these genes, as well as the fact that the Mediator
complex is known to be regulated by Ras signaling(Chang, Howard, & Herman, 2004)
and more specifically has been shown to physically interact with Sfl1 to inhibit
transcription(Song & Carlson, 1998), suggests that most of these capacitors influence
transcriptional regulation by the Ras pathway. This finding supports our recent discovery
that capacitance in the BYx3S colony morphology system requires transcriptional
derepression of Ras target genes(Taylor & Ehrenreich, 2015b). However, even though
these capacitors each likely lead to derepression, they show significant differences in the
combinations of cryptic variants they require to exert their effects. Specifically, between
one and nine cryptic variants were detected in backcross populations derived from the
mutants (Fig. 4.2C). As we discuss later, these differences in genetic complexity among
segregants likely relate to the rough phenotype’s underlying gene regulatory network
architecture.
4.6 Instances of capacitance in a wild type background
The nine other rough segregants obtained did not harbor spontaneous mutations. This
supports a previous finding that some individuals in the BYx3S cross show rough
morphology at room temperature despite lacking the ira2∆2933 lesion(Lee et al.). This is
consistent with our current results, as rough segregants that lack capacitating mutations in
our current study show only weaker ‘bumpy’ phenotypes at higher temperatures (Fig.
4.6). Six or more loci were detected in each of these cases, implying that higher-order
68
genetic interactions among cryptic variants induced capacitance. All of the mapping
populations lacking mutations were fixed for the 3S allele of FLO8 and the BY allele of
MSS11, which encodes an activator that dimerizes with Flo8. Two-thirds of these
individuals also possessed intragenic recombinations within a ~1.3 kb region preceding
the transcription start site of the cell surface glycoprotein FLO11, a gene that must be
transcribed for the rough phenotype to be expressed(Taylor & Ehrenreich, 2015b) (Fig.
4.7A). Specifically, the BY promoter and the 3S coding region of FLO11 harbor cryptic
variants that together can facilitate capacitance in the presence of other cryptic variants
(Fig. 4.2C; Fig. 4.7B). The other third of the wild type rough segregants did not carry
particular FLO11 haplotypes, but instead showed selection for four additional cryptic
variants on Chromosomes V, VI, XIV, and XV, as well as enrichment at other loci that
differed among the mapping populations (Fig. 4.2C).
Fig. 4.6 | Impact of temperature on rough segregants.
69
Fig. 4.7 | Location of recombinations within the FLO11 locus. A Genotypes of rough
segregants that possess a FLO11 intragenic recombination event are shown, along with
annotations of this locus from the Saccharomyces Genome Database(Cherry et al.). B
Representative images of allele replacements in the coding (‘c’) and promoter (‘p’)
regions of FLO11, which verify the presence of two cryptic variants at this gene, are
provided.
4.7 Most cryptic variants are in Ras pathway components
We next sought to better define the cryptic variants that potentiate capacitance, as most of
the identified cryptic variants contributed to multiple instances of the rough phenotype.
Among the 16 loci not corresponding to the FLO11 cryptic variants, seven contain genes
that we previously showed harbor cryptic variation or identified as capacitors(Lee et al.,
Submitted; Taylor & Ehrenreich, 2014, 2015b). In addition to FLO8, IRA2, MSS11, and
SFL1, we detected loci overlapping the vesicle component END3, the activators MGA1
and MSS11, and the redox stress detoxifier TRR1 (Fig. 4.2C). Flo8-Mss11 heterodimer,
Mga1, and Sfl1 each act downstream of the Ras pathway and likely to co-regulate FLO11
and other genes that are important for yeast colony morphology traits (Bruckner &
Mosch). Furthermore, although End3 and Trr1 are not components of the Ras pathway,
functional relationships between these genes and Ras signaling likely exist(Charizanis,
Juhnke, Krems, & Entian; Gourlay & Ayscough).
In addition to previously identified genes supporting an important role for cryptic
variation in the Ras pathway in capacitance, we cloned novel genes that harbor cryptic
70
variation and influence Ras signaling. We focused on determining three of the remaining,
uncharacterized loci that were detected the greatest number of times in our genetic
mapping experiments. By performing allele replacements in multiple rough segregants,
we successfully resolved loci on Chromosomes V, VII, and XI to GPA2, MDS3, and
TPK3, respectively (Fig. 4.2C; Fig. 4.8; Methods). These genes include a G protein
subunit that is required for recruitment of Ras-GTP (GPA2)(Cherry et al.), a component
of the TOR pathway that has also been shown to influence Ras signaling
(MDS3)(McDonald et al., 2009), and a subunit of the Ras effector kinase Protein Kinase
A (TPK3)(Cherry et al.).
Fig. 4.8 | Replacement of cryptic allele suggests a background-dependence. Cryptic
alleles of GPA2, MDS3, and TPK3 were replaced with the non-causal allele in four rough
segregants.
4.8 Discussion
In summary, we have determined the genetic architectures underlying 17 independent
instances of capacitance. These cases vary significantly in the number of involved cryptic
variants and also in whether they require a capacitating mutation. Among cases involving
71
functional disruption of a capacitor, our work suggests that the genetic architecture of
capacitance depends both on the nature of a capacitating event and the identities of the
cryptic variants that are present. This relationship between capacitating mutation and
genetic complexity of capacitance is likely tied to Ras signaling and the transcriptional
control of Ras target genes (Fig. 4.9). For example, SSN3 and SSN8 mutations, which act
the most proximally to transcription, show the lowest genetic complexity in our study
(Fig. 4.2C; Fig. 4.9). In contrast, GPB1, IRA1, and IRA2 mutations, which act at the
beginning of the Ras cascade, show higher numbers of detected loci (Fig. 4.2C; Fig. 4.9).
Our study also illustrates how capacitance can arise without mutations due to higher-
order genetic interactions among cryptic variants, which presumably recapitulate the
molecular and systems level effects of capacitance involving mutations.
Given that most of the identified cryptic variants act in or are influenced by the Ras
pathway (Fig. 4.9), which is evolutionarily conserved(Cox & Der, 2010), our findings
might extend to other species and traits. In fact, cryptic variation in the Ras pathway is
known to impact development in Caenorhabditis elegans(Milloz et al.) and perturbation
of Ras components in humans can lead to cancer and other diseases(Krauthammer et al.).
Thus, further characterizing capacitance due to the Ras pathway in yeast might provide
valuable new insights into the mechanisms that give rise to genetically complex
phenotypes that are relevant to health and evolution.
72
Fig. 4.9 | Capacitors and cryptic variants largely regulate signaling and
transcriptional control by the Ras pathway. Most of the identified cryptic variants and
capacitating mutations from this work are involved in Ras signaling or in the Mediator
transcriptional complex.
4.9 Materials and methods
4.9.1 Phenotyping of yeast colony morphology. Strains were grown at 30
o
C overnight
in liquid media comprised of yeast extract and peptone (YP) with 2% dextrose as the
carbon source (YPD). Stationary phase cultures were pinned onto YP agar media with
2% ethanol as the carbon source (YPE) and grown for 7 days at 21
o
C, unless otherwise
noted.
4.9.2 Generation of rough segregants. The rough segregants examined in this work
come from a cross between the BY4716 (‘BY’) lab strain, a derivative of S288C, and
73
322134S (‘3S’), a clinical isolate. Strains used in this work possess the Synthetic Genetic
Array marker system(Tong et al.), which facilitated rapid generation of MATa haploid
progeny from diploid strains. This system allows for selection of MATa spores by plating
of sporulated cultures onto yeast nitrogen base (YNB) media containing canavanine. A
MATa BY strain and a MATα 3S strain were mixed on YPD agar media and incubated for
4 hours at 30
o
C. 106 zygotes were obtained from this cross by microdissection. Each of
these was sporulated as in(Taylor & Ehrenreich, 2014), after which segregants were
plated onto YNB containing canavanine to an average density of ~200 per plate. In total,
>500 such plates were produced, or >100,000 segregants. Once colonies were visible on
these YNB plates, they were replicated onto YPE plates. After 7 days of growth at room
temperature, segregants were screened for rough morphology. No more than one rough
segregant was gathered from each diploid to minimize instances where multiple
segregants were rough due to the same capacitating event.
4.9.3 Generation of backcross segregants. Each rough segregant was backcrossed to
both BY and 3S to generate diploids as described above. These were sporulated and
plated onto YNB media containing canavanine to select for MATa haploid progeny.
Between 46 and 95 segregants were obtained from each of these crosses to create a
mapping population. Segregants from each cross were grown to stationary phase in YPD
at 30
o
C and mixed together in equal volumes. DNA was extracted from this mixture
(hereafter referred to as a rough pool) using the Qiagen Genomic-tip 100/G kit. To
generate control populations for these crosses, sporulations were plated at high density
onto YNB with canavanine and grown at 21
o
C for two days to produce lawns. These
74
lawns were then scraped directly off the plates and DNA was extracted from these cell
populations.
4.9.4 Sequencing of mapping populations. Sequencing libraries were generated from
each backcross pool using the Illumina Nextera kit. These libraries were sequenced on an
Illumina NextSeq machine by the USC Epigenome Center using 75x75 base reads.
Rough pools were sequenced to an average of 244-fold and control pools were sequenced
to an average of 139-fold coverage. Reads were aligned to either a BY or 3S reference
genome using the Burrows-Wheeler Aligner (BWA) version 7 with options mem -t 20(Li
& Durbin).
4.9.5 Genetic mapping using Multipool. Genome-wide allele frequencies at 36,756 high
confidence SNPs were determined by a custom Python script(Taylor & Ehrenreich). Loci
were detected in each mapping population using MULTIPOOL(Edwards & Gifford).
Segregating genomic intervals in each population were independently analyzed, and
regions associated with the phenotype were identified as having a LOD score > 5
spanning a region greater than 30kb. Within each locus, a 90% confidence interval
around the point of maximum significance was used to delimit the location of the causal
variant. Loci that were co-detected in the corresponding control populations were
ignored. In a few instances, detected loci were broad and overlapped two crpytic variants
identified in this or our previous work (Taylor & Ehrenreich, 2014, 2015b). When
assessing the complexity of genetic architectures, these broad regions were counted as
two loci.
75
4.9.6 Identification of capacitating mutations. Identified loci were scanned for
mutations by standard computational methods. In short, mpileups were generated by
alignment of sequencing data to BY and 3S reference genomes. These were examined for
differences from both genomes. Identified lesions were then validated by Sanger
sequencing of the corresponding rough segregant. To determine if aneuploidies caused
rough morphology in any individuals, the average coverage of each chromosome was
examined in the rough and control backcross pools. Only mapping populations descended
from rough segregant 12 possessed an aneuploidy, which occurred on Chromosome I. 38
and 34% enrichment of this chromosome was observed relative to average genome-wide
coverage in the BY and 3S backcrosses, respectively. However, similar enrichment in
coverage for Chromosome I was seen in control pools descended from rough segregant
12, suggesting that the aneuploidy does not contribute to the rough phenotype.
4.9.7 Genetic engineering experiments. Gene alleles were converted from BY to 3S or
3S to BY using a modified form of adaptamer mediated allele replacement(Erdeniz et al.,
1997; Matsui et al.; Taylor & Ehrenreich, 2015b). Transformations were conducted with
two partially overlapping PCR products—a full-length amplicon of a gene of interest that
was tailed at the 3’ end with the 5’ portion of the kanMX cassette and a copy of the
kanMX cassette that was tailed on the 3’ end with part of the intragenic region
downstream of the gene. Knock-ins were identified using selection on G418 and verified
by Sanger sequencing.
76
Chapter 5: Concluding remarks
In this thesis, I describe the development of a powerful model system that allows for
inquiry into several understudied phenomenon in complex trait genetics. My work sheds
light on the genetic architecture of genetic interactions, capacitance, and genetic
background effects, all of which have a strong theoretical basis but little empirical data.
In this Chapter, I will tie together the major findings of Chapters 2-4 by reviewing the
conceptual link between each. I will then discuss the impact of this work on our
understanding of complex trait genetics. Finally, I will discuss future directions that could
be explored using my system.
5.1 Identification of the most complex genetic interactions to date
In Chapter 2, I characterize the genetic basis of a phenotype caused by higher-order
genetic interactions. I found that a discrete rough colony morphology trait segregates at a
low frequency among progeny of two yeast strains that form smooth colonies.
Backcrosses suggested that this phenotype was genetically complex, so I used a
combination of genome-wide and targeted genotyping approaches to map seven loci that
influence the trait. Examination of the genotypes of individual progeny demonstrated that
only two specific combinations of five or six loci could cause rough morphology. These
two genotypes represent the most complex genetic interactions ever described. This
finding highlights the role higher-order genetic interactions can play in specifying
complex traits, which will inform future work as most studies only account for
interactions between two loci.
77
5.2 Cloning of interacting genetic variants demonstrates Ras signaling plays an
important role
In Chapters 2 and 3, I use cloning strategies to resolve interacting loci to gene-level
resolution. One of the identified loci contains a mutation in the gene IRA2 (ira2∆2933), a
negative regulator of the Ras pathway. This gene is thus a capacitor, as the mutation
reveals cryptic variants that cause rough morphology. In addition to IRA2, I identified 4
transcription factors (FLO8, MGA1, MSS11, SFL1) that are regulated by the Ras
pathway, suggesting that changes in gene regulation likely play a role in these
interactions. These findings support systems modelling work which theorized that genetic
capacitors could be components of gene regulatory networks. Additionally, it suggests
that complex perturbation of gene regulatory networks may be an important mechanism
underlying capacitance.
5.3 Transcriptional derepression of Ras targets mediates capacitance
The transcription factors involved that I cloned in Chapters 2 and 3 all regulate the cell-
cell adhesion molecule FLO11, a target of the Ras pathway. In Chapter 3, I showed that
FLO11 is necessary for rough morphology, and that it is only expressed in certain genetic
backgrounds in the presence of the ira2∆2933 allele. This suggests that transcriptional
de-repression plays a role in revealing the cryptic interactions described above. This is
further supported by an additional BY x 3S cross, in which the FLO11 repressor SFL1
was deleted from both parents. This cross yielded rough progeny, demonstrating that
78
SFL1 is a capacitor in this system. Genotyping of these rough individuals demonstrated
that largely the same cryptic variants are revealed by perturbation of IRA2 and SFL1.
This finding demonstrates that some genes can both harbor cryptic variation and act as
capacitors (SFL1). Additionally, it suggests that the role of the ira2∆2933 allele is to
relieve transcriptional repressor of at least one key Ras target.
5.4 Additional architectures of capacitance can lead to the same phenotype
In Chapter 4, I described the use of this rough morphology system as a reporter for
additional instances of capacitance. I used a bulk-segregant mapping approach to dissect
the genetic architecture underlying 17 unique instances of the colony morphology
phenotype. Importantly, this work was carried out at a non-standard temperature (21oC),
which our group has shown to reveal additional genetic architectures of capacitance.
Surprisingly, every instance involved a required a unique combination of loci that
contributed to the rough morphology. This finding highlights the complexity of detecting
cryptic genetic variation.
Bioinformatic analysis and genetic engineering revealed 8 new mutations that can
capacitate rough morphology. The genetic architecture of rough morphology in these
mutants varied greatly in composition and complexity depending on the capacitating
event. The 8 mutations fell primarily in negative regulators of the Ras signaling pathway
or in components of the mediator complex, which is directed to Ras targets. Additionally,
cloning of three new loci revealed cryptic variants in genes that play a role in the Ras
79
pathway. These results support previous findings that complex perturbation of gene
regulatory network components can underlie capacitance. Further, the genetic complexity
of capacitance appears to be a function of the capacitating mutation involved, as negative
regulators of the Ras pathway require far more loci than components of the mediator
complex.
About half of the 17 rough segregants used for mapping experiments did not possess a
new mutation. 6 carried an intragenic recombination that combined the BY allele of the
FLO11 promoter with the 3S allele of its coding sequence. An additional 3 individuals
possessed highly-complex genetic interactions between 8 or more loci. We believe that in
these individuals, complex combinations of cryptic variants can themselves capacitate the
rough morphology phenotype. This demonstrates that higher-order genetic interactions
can to reveal cryptic variation within a system.
5.5 Impact of my work
The total body of work from my PhD tells in unprecedented detail how phenotypic
capacitance can occur. Much of the theory behind capacitance relies on simulations and a
few model systems, as studies have been limited in scope by the challenges associated
with studying this phenomenon. My research has led to the first demonstration of how
capacitance occurs at the genetic level, how it can be mediated at the molecular level, and
how different capacitors will reveal different combinations of cryptic variants.
Interestingly, much of this cryptic variation occurs within the same signaling pathway,
80
which is conserved throughout eukaryotes. That this pathway can harbor such a great
amount of genetic variation without showing an impact on phenotype in most instances
may be particularly relevant to understandings of human disease. Indeed, the Ras
pathway harbors many oncogenes, such as NF1, the human homolog of IRA2. Mutations
and anneuploidies involving this gene are responsible for neurofibromatosis in ~1/3,5000
children each year in the US ("Neurofibromatosis Fact Sheet,").
Additionally, my work highlights some of the challenges in studying capacitance. Each
unique instance of capacitance revealed unique combinations of capacitors, but caused
the same phenotype. Thus, a bulk segregant mapping approach based on F
2
progeny of
the BY x 3S cross would likely not give a true assessment of the complexity underlying
this phenotype due to genetic heterogeneity. Only the most commonly detected loci
would show up in mapping data, whereas less-commonly-involved loci would show a
non-significant skew in allele frequency. Capacitors would surely be missed in analysis
since seven unique genes were implicated in capacitance in our work, and thus none
would be significantly enriched in a mapping study. The finding that most of the
variation detected in this work falls within the Ras pathway suggests that a more fruitful
approach may be to explore individual pathways for cryptic variation, rather than focus
on population-level analysis.
81
5.6 Future directions
Moving forward, this system has the potential to yield additional insights into the
phenomenon of genetic capacitance. The work in Chapter 4 examined major architectures
in a population through a pooled sequencing approach. However, it is likely that genetic
heterogeneity masks the true complexity of rough morphology in these instances of
capacitance. Genotyping of individuals from each population could thus expand our
understanding of the architectures underlying capacitance. This would further allow us to
determine the role that enriched loci play in generating rough morphology, as it is
currently unclear whether their allele frequency skew is caused by genetic heterogeneity,
additive effects, or by a combination of the two.
The strains I generated in this work could also be used to explore the molecular basis of
capacitance. Reporter constructs attached to key genes in the Ras pathway (e.g. Ras2,
Flo8, Sfl1, and Flo11) could allow us to more finely examine the precise molecular role
of each cryptic variant and perturbation, as well as the role of combinations of these
factors. This would allow us to assess the exact mechanism underlying capacitance, and
thus whether a common mechanism is responsible in all cases examined.
Finally, dissecting even more instances of capacitance could provide a unique insight into
the general characterisitcs of capacitors. The current screen certainly did not reach
saturation, as at least one known capacitor (SFL1) and one likely capacitor (GPB2, which
interacts with IRA1) were not identified. My findings in Chapters 3 and 4 suggest that a
82
specific set of genes are primarily capable of capacitating rough morphology. However,
the identification of IRC8, which has an unknown function and is likely not associated
with Ras signaling or adhesion, suggesting that other factors may be capable of acting as
capacitors. Determining whether this is the case would provide critical insights in to the
characteristics of this important but understudied phenomenon.
83
References
(Ed.), J. B. W. (2000). Epistasis and the evolutionary process. Oxford University Press.
Abecasis, G. R., Auton, A., Brooks, L. D., DePristo, M. A., Durbin, R. M., Handsaker, R.
E., . . . McVean, G. A. (2012). An integrated map of genetic variation from 1,092
human genomes. Nature, 491(7422), 56-65. doi:10.1038/nature11632
Andersen, E. C., Gerke, J. P., Shapiro, J. A., Crissman, J. R., Ghosh, R., Bloom, J. S., . . .
Kruglyak, L. (2012). Chromosome-scale selective sweeps shape Caenorhabditis
elegans genomic diversity. Nature genetics, 44(3), 285-290. doi:10.1038/ng.1050
Andolfatto, P., Davison, D., Erezyilmaz, D., Hu, T. T., Mast, J., Sunayama-Morita, T., &
Stern, D. L. (2011). Multiplexed shotgun genotyping for rapid and efficient
genetic mapping. Genome research, 21(4), 610-617. doi:10.1101/gr.115402.110
Ansel, J., Bottin, H., Rodriguez-Beltran, C., Damon, C., Nagarajan, M., Fehrmann, S., . .
. Yvert, G. (2008). Cell-to-cell stochastic variation in gene expression is a
complex genetic trait. PLoS genetics, 4(4), e1000049.
doi:10.1371/journal.pgen.1000049
Atwell, S., Huang, Y. S., Vilhjalmsson, B. J., Willems, G., Horton, M., Li, Y., . . .
Nordborg, M. (2010). Genome-wide association study of 107 phenotypes in
Arabidopsis thaliana inbred lines. Nature, 465(7298), 627-631.
doi:10.1038/nature08800
Aylor, D. L., Valdar, W., Foulds-Mathes, W., Buus, R. J., Verdugo, R. A., Baric, R. S., . .
. Churchill, G. A. (2011). Genetic analysis of complex traits in the emerging
Collaborative Cross. Genome research, 21(8), 1213-1222.
doi:10.1101/gr.111310.110
Bagheri, H. C., Hermisson, J., Vaisnys, J. R., & Wagner, G. P. (2003). Effects of epistasis
on phenotypic robustness in metabolic pathways. Mathematical Biosciences,
184(1), 27-51. doi:10.1016/S0025-5564(03)00057-9
Barkoulas, M., van Zon, J. S., Milloz, J., van Oudenaarden, A., & Felix, M. A. (2013).
Robustness and epistasis in the C. elegans vulval signaling network revealed by
pathway dosage modulation. Dev Cell, 24(1), 64-75.
doi:10.1016/j.devcel.2012.12.001
Benedetti, H., Raths, S., Crausaz, F., & Riezman, H. (1994). The END3 gene encodes a
protein that is required for the internalization step of endocytosis and for actin
cytoskeleton organization in yeast. Mol Biol Cell, 5(9), 1023-1037. Retrieved
from http://www.ncbi.nlm.nih.gov/pubmed/7841519
Berger, D., Bauerfeind, S. S., Blanckenhorn, W. U., & Schafer, M. A. (2011). High
temperatures reveal cryptic genetic variation in a polymorphic female sperm
storage organ. Evolution, 65(10), 2830-2842. doi:10.1111/j.1558-
5646.2011.01392.x
Bergman, A., & Siegal, M. L. (2003). Evolutionary capacitance as a general feature of
complex gene networks. Nature, 424(6948), 549-552. doi:10.1038/nature01765
Bloom, J. S., Ehrenreich, I. M., Loo, W. T., Lite, T. L., & Kruglyak, L. (2013). Finding
the sources of missing heritability in a yeast cross. Nature, 494(7436), 234-237.
doi:10.1038/nature11867
Blount, Z. D., Borland, C. Z., & Lenski, R. E. (2008). Historical contingency and the
evolution of a key innovation in an experimental population of Escherichia coli.
Proceedings of the National Academy of Sciences of the United States of
America, 105(23), 7899-7906. doi:10.1073/pnas.0803151105
84
Boone, C., Bussey, H., & Andrews, B. J. (2007). Exploring genetic interactions and
networks with yeast. Nature reviews. Genetics, 8(6), 437-449.
doi:10.1038/nrg2085
Borneman, A. R., Leigh-Bell, J. A., Yu, H., Bertone, P., Gerstein, M., & Snyder, M.
(2006). Target hub proteins serve as master regulators of development in yeast.
Genes Dev, 20(4), 435-448. doi:10.1101/gad.1389306
Brem, R. B., Storey, J. D., Whittle, J., & Kruglyak, L. (2005). Genetic interactions
between polymorphisms that affect gene expression in yeast. Nature, 436(7051),
701-703. doi:10.1038/nature03865
Brem, R. B., Yvert, G., Clinton, R., & Kruglyak, L. (2002). Genetic dissection of
transcriptional regulation in budding yeast. Science, 296(5568), 752-755.
doi:10.1126/science.1069516
Bruckner, S., & Mosch, H. U. (2012). Choosing the right lifestyle: adhesion and
development in Saccharomyces cerevisiae. FEMS Microbiol Rev, 36(1), 25-58.
doi:10.1111/j.1574-6976.2011.00275.x
Caicedo, A. L., Stinchcombe, J. R., Olsen, K. M., Schmitt, J., & Purugganan, M. D.
(2004). Epistatic interaction between Arabidopsis FRI and FLC flowering time
genes generates a latitudinal cline in a life history trait. Proceedings of the
National Academy of Sciences of the United States of America, 101(44), 15670-
15675. doi:10.1073/pnas.0406232101
Carlborg, O., Andersson, L., & Kinghorn, B. (2000). The use of a genetic algorithm for
simultaneous mapping of multiple interacting quantitative trait loci. Genetics,
155(4), 2003-2010. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10924492
Carlborg, O., & Haley, C. S. (2004). Epistasis: too often neglected in complex trait
studies? Nature reviews. Genetics, 5(8), 618-625. doi:10.1038/nrg1407
Carlborg, O., Jacobsson, L., Ahgren, P., Siegel, P., & Andersson, L. (2006). Epistasis and
the release of genetic variation during long-term selection. Nature genetics,
38(4), 418-420. doi:10.1038/ng1761
Chandler, C. H., Chari, S., & Dworkin, I. (2013). Does your gene need a background
check? How genetic background impacts the analysis of mutations, genes, and
evolution. Trends in genetics : TIG, 29(6), 358-366.
doi:10.1016/j.tig.2013.01.009
Chandler, C. H., Chari, S., Tack, D., & Dworkin, I. (2014). Causes and consequences of
genetic background effects illuminated by integrative genomic analysis.
Genetics, 196(4), 1321-1336. doi:10.1534/genetics.113.159426
Chang, Y. W., Howard, S. C., & Herman, P. K. (2004). The Ras/PKA signaling pathway
directly targets the Srb9 protein, a component of the general RNA polymerase II
transcription apparatus. Mol Cell, 15(1), 107-116.
doi:10.1016/j.molcel.2004.05.021
Chari, S., & Dworkin, I. (2013). The conditional nature of genetic interactions: the
consequences of wild-type backgrounds on mutational interactions in a genome-
wide modifier screen. PLoS genetics, 9(8), e1003661.
doi:10.1371/journal.pgen.1003661
Charizanis, C., Juhnke, H., Krems, B., & Entian, K. D. (1999). The oxidative stress
response mediated via Pos9/Skn7 is negatively regulated by the Ras/PKA
pathway in Saccharomyces cerevisiae. Mol Gen Genet, 261(4-5), 740-752.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10394911
85
Cherry, J. M., Hong, E. L., Amundsen, C., Balakrishnan, R., Binkley, G., Chan, E. T., . . .
Wong, E. D. (2012). Saccharomyces Genome Database: the genomics resource
of budding yeast. Nucleic acids research, 40(Database issue), D700-705.
doi:10.1093/nar/gkr1029
Clark, R. M., Schweikert, G., Toomajian, C., Ossowski, S., Zeller, G., Shinn, P., . . .
Weigel, D. (2007). Common sequence polymorphisms shaping genetic diversity
in Arabidopsis thaliana. Science, 317(5836), 338-342.
doi:10.1126/science.1138632
Collins, R. L., Hu, T., Wejse, C., Sirugo, G., Williams, S. M., & Moore, J. H. (2013).
Multifactor dimensionality reduction reveals a three-locus epistatic interaction
associated with susceptibility to pulmonary tuberculosis. BioData Min, 6(1), 4.
doi:10.1186/1756-0381-6-4
Conlan, R. S., & Tzamarias, D. (2001). Sfl1 functions via the co-repressor Ssn6-Tup1
and the cAMP-dependent protein kinase Tpk2. J Mol Biol, 309(5), 1007-1015.
doi:10.1006/jmbi.2001.4742
Corbett-Detig, R. B., Zhou, J., Clark, A. G., Hartl, D. L., & Ayroles, J. F. (2013). Genetic
incompatibilities are widespread within species. Nature, 504(7478), 135-137.
doi:10.1038/nature12678
Cordell, H. J. (2009). Detecting gene-gene interactions that underlie human diseases.
Nature reviews. Genetics, 10(6), 392-404. doi:10.1038/nrg2579
Costanzo, M., Baryshnikova, A., Bellay, J., Kim, Y., Spear, E. D., Sevier, C. S., . . .
Boone, C. (2010). The genetic landscape of a cell. Science, 327(5964), 425-431.
doi:10.1126/science.1180823
Cowen, L. E., & Lindquist, S. (2005). Hsp90 potentiates the rapid evolution of new traits:
drug resistance in diverse fungi. Science, 309(5744), 2185-2189.
doi:10.1126/science.1118370
Cox, A. D., & Der, C. J. (2010). Ras history: The saga continues. Small GTPases, 1(1), 2-
27. doi:10.4161/sgtp.1.1.12178
D.L. Hartl, A. G. C. (2006). Principles of population genetics. Sinauer Associates.
de Visser, J. A., Hermisson, J., Wagner, G. P., Ancel Meyers, L., Bagheri-Chaichian, H.,
Blanchard, J. L., . . . Whitlock, M. C. (2003). Perspective: Evolution and
detection of genetic robustness. Evolution, 57(9), 1959-1972. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/14575319
Dowell, R. D., Ryan, O., Jansen, A., Cheung, D., Agarwala, S., Danford, T., . . . Boone,
C. (2010). Genotype to phenotype: a complex problem. Science, 328(5977), 469.
doi:10.1126/science.1189015
Duveau, F., & Felix, M. A. (2012). Role of pleiotropy in the evolution of a cryptic
developmental variation in Caenorhabditis elegans. PLoS biology, 10(1),
e1001230. doi:10.1371/journal.pbio.1001230
Dworkin, I. (2005). Towards a genetic architecture of cryptic genetic variation and
genetic assimilation: the contribution of K. G. Bateman. J Genet, 84(3), 223-226.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/16385156
Dworkin, I., Palsson, A., Birdsall, K., & Gibson, G. (2003). Evidence that Egfr
contributes to cryptic genetic variation for photoreceptor determination in natural
populations of Drosophila melanogaster. Curr Biol, 13(21), 1888-1893.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/14588245
86
Earley, E. J., & Jones, C. D. (2011). Next-generation mapping of complex traits with
phenotype-based selection and introgression. Genetics, 189(4), 1203-1209.
doi:10.1534/genetics.111.129445
Edwards, M. D., & Gifford, D. K. (2012). High-resolution genetic mapping with pooled
sequencing. BMC Bioinformatics, 13 Suppl 6, S8. doi:10.1186/1471-2105-13-S6-
S8
Edwards, M. D., Symbor-Nagrabska, A., Dollard, L., Gifford, D. K., & Fink, G. R.
(2014). Interactions between chromosomal and nonchromosomal elements reveal
missing heritability. Proceedings of the National Academy of Sciences of the
United States of America, 111(21), 7719-7722. doi:10.1073/pnas.1407126111
Ehrenreich, I. M., & Pfennig, D. W. (2015). Genetic assimilation: a review of its potential
proximate causes and evolutionary consequences. Ann Bot.
doi:10.1093/aob/mcv130
Ehrenreich, I. M., Torabi, N., Jia, Y., Kent, J., Martis, S., Shapiro, J. A., . . . Kruglyak, L.
(2010). Dissection of genetically complex traits with extremely large pools of
yeast segregants. Nature, 464(7291), 1039-1042. doi:10.1038/nature08923
Eichler, E. E., Flint, J., Gibson, G., Kong, A., Leal, S. M., Moore, J. H., & Nadeau, J. H.
(2010). Missing heritability and strategies for finding the underlying causes of
complex disease. Nature reviews. Genetics, 11(6), 446-450. doi:10.1038/nrg2809
Erdeniz, N., Mortensen, U. H., & Rothstein, R. (1997). Cloning-free PCR-based allele
replacement methods. Genome research, 7(12), 1174-1183. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/9414323
Falconer DS, M. T. (1996). Introduction to quantitative genetics (4th edition). Harlow,
England: Pearson Educated Limited.
Felix, M. A. (2007). Cryptic quantitative evolution of the vulva intercellular signaling
network in Caenorhabditis. Curr Biol, 17(2), 103-114.
doi:10.1016/j.cub.2006.12.024
Felix, M. A., & Barkoulas, M. (2015). Pervasive robustness in biological systems. Nature
reviews. Genetics, 16(8), 483-496. doi:10.1038/nrg3949
Felix, M. A., & Wagner, A. (2008). Robustness and evolution: concepts, insights and
challenges from a developmental model system. Heredity, 100(2), 132-140.
doi:10.1038/sj.hdy.6800915
Fichtner, L., Schulze, F., & Braus, G. H. (2007). Differential Flo8p-dependent regulation
of FLO1 and FLO11 for cell-cell and cell-substrate adherence of S. cerevisiae
S288c. Mol Microbiol, 66(5), 1276-1289. doi:10.1111/j.1365-2958.2007.06014.x
Flatt, T. (2005). The evolutionary genetics of canalization. Q Rev Biol, 80(3), 287-316.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/16250465
Frankel, W. N., & Schork, N. J. (1996). Who's afraid of epistasis? Nature genetics, 14(4),
371-373. doi:10.1038/ng1296-371
Fujita, A., Kikuchi, Y., Kuhara, S., Misumi, Y., Matsumoto, S., & Kobayashi, H. (1989).
Domains of the SFL1 protein of yeasts are homologous to Myc oncoproteins or
yeast heat-shock transcription factor. Gene, 85(2), 321-328. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/2697640
Gaertner, B. E., Parmenter, M. D., Rockman, M. V., Kruglyak, L., & Phillips, P. C.
(2012). More than the sum of its parts: a complex epistatic network underlies
natural variation in thermal preference behavior in Caenorhabditis elegans.
Genetics, 192(4), 1533-1542. doi:10.1534/genetics.112.142877
87
Gagiano, M., Bester, M., van Dyk, D., Franken, J., Bauer, F. F., & Pretorius, I. S. (2003).
Mss11p is a transcription factor regulating pseudohyphal differentiation, invasive
growth and starch metabolism in Saccharomyces cerevisiae in response to
nutrient availability. Mol Microbiol, 47(1), 119-134. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/12492858
Gavin, A. C., Aloy, P., Grandi, P., Krause, R., Boesche, M., Marzioch, M., . . . Superti-
Furga, G. (2006). Proteome survey reveals modularity of the yeast cell
machinery. Nature, 440(7084), 631-636. doi:10.1038/nature04532
Gerke, J., Lorenz, K., & Cohen, B. (2009). Genetic interactions between transcription
factors cause natural variation in yeast. Science, 323(5913), 498-501.
doi:10.1126/science.1166426
Gerke, J., Lorenz, K., Ramnarine, S., & Cohen, B. (2010). Gene-environment interactions
at nucleotide resolution. PLoS genetics, 6(9), e1001144.
doi:10.1371/journal.pgen.1001144
Gibson, G. (2009). Decanalization and the origin of complex disease. Nature reviews.
Genetics, 10(2), 134-140. doi:10.1038/nrg2502
Gibson, G., & Dworkin, I. (2004). Uncovering cryptic genetic variation. Nature reviews.
Genetics, 5(9), 681-690. doi:10.1038/nrg1426
Gibson, G., & Hogness, D. S. (1996). Effect of polymorphism in the Drosophila
regulatory gene Ultrabithorax on homeotic stability. Science, 271(5246), 200-
203. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8539619
Gibson, G., & Wagner, G. (2000). Canalization in evolutionary genetics: a stabilizing
theory? BioEssays : news and reviews in molecular, cellular and developmental
biology, 22(4), 372-380. doi:10.1002/(SICI)1521-1878(200004)22:4<372::AID-
BIES7>3.0.CO;2-J
Gibson, G., Wemple, M., & van Helden, S. (1999). Potential variance affecting homeotic
Ultrabithorax and Antennapedia phenotypes in Drosophila melanogaster.
Genetics, 151(3), 1081-1091. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10049924
Gietz, R. D., & Woods, R. A. (2002). Transformation of yeast by lithium acetate/single-
stranded carrier DNA/polyethylene glycol method. Methods in enzymology, 350,
87-96. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12073338
Gjuvsland, A. B., Hayes, B. J., Omholt, S. W., & Carlborg, O. (2007). Statistical epistasis
is a generic feature of gene regulatory networks. Genetics, 175(1), 411-420.
doi:10.1534/genetics.106.058859
Gourlay, C. W., & Ayscough, K. R. (2006). Actin-induced hyperactivation of the Ras
signaling pathway leads to apoptosis in Saccharomyces cerevisiae. Mol Cell Biol,
26(17), 6487-6501. doi:10.1128/MCB.00117-06
Granek, J. A., & Magwene, P. M. (2010). Environmental and genetic determinants of
colony morphology in yeast. PLoS genetics, 6(1), e1000823.
doi:10.1371/journal.pgen.1000823
Granek, J. A., Murray, D., Kayrkci, O., & Magwene, P. M. (2013). The genetic
architecture of biofilm formation in a clinical isolate of Saccharomyces
cerevisiae. Genetics, 193(2), 587-600. doi:10.1534/genetics.112.142067
Guthrie, C. F., GR. (1991). Guide to Yeast Genetics and Molecular Biology. (Vol. 194
).
88
Halfmann, R., Jarosz, D. F., Jones, S. K., Chang, A., Lancaster, A. K., & Lindquist, S.
(2012). Prions are a common mechanism for phenotypic inheritance in wild
yeasts. Nature, 482(7385), 363-368. doi:10.1038/nature10875
Halme, A., Bumgarner, S., Styles, C., & Fink, G. R. (2004). Genetic and epigenetic
regulation of the FLO gene family generates cell-surface variation in yeast. Cell,
116(3), 405-415. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/15016375
Hanlon, P., Lorenz, W. A., Shao, Z., Harper, J. M., Galecki, A. T., Miller, R. A., &
Burke, D. T. (2006). Three-locus and four-locus QTL interactions influence
mouse insulin-like growth factor-I. Physiol Genomics, 26(1), 46-54.
doi:10.1152/physiolgenomics.00247.2005
Hansen, T. F. (2013). Why epistasis is important for selection and adaptation. Evolution,
67(12), 3501-3511. doi:10.1111/evo.12214
Hansen, T. F., Alvarez-Castro, J. M., Carter, A. J., Hermisson, J., & Wagner, G. P.
(2006). Evolution of genetic architecture under directional selection. Evolution,
60(8), 1523-1536. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/17017054
Harashima, T., Anderson, S., Yates, J. R., 3rd, & Heitman, J. (2006). The kelch proteins
Gpb1 and Gpb2 inhibit Ras activity via association with the yeast RasGAP
neurofibromin homologs Ira1 and Ira2. Mol Cell, 22(6), 819-830.
doi:10.1016/j.molcel.2006.05.011
Hartman, J. L. t., Garvik, B., & Hartwell, L. (2001). Principles for the buffering of
genetic variation. Science, 291(5506), 1001-1004. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/11232561
Hermisson, J., Hansen, T. F., & Wagner, G. P. (2003). Epistasis in polygenic traits and
the evolution of genetic architecture under stabilizing selection. American
Naturalist, 161(5), 708-734. doi:Doi 10.1086/374204
Hermisson, J., & Wagner, G. P. (2004). The population genetic theory of hidden variation
and genetic robustness. Genetics, 168(4), 2271-2284.
doi:10.1534/genetics.104.029173
Holmes, D. L., Lancaster, A. K., Lindquist, S., & Halfmann, R. (2013). Heritable
remodeling of yeast multicellularity by an environmentally responsive prion.
Cell, 153(1), 153-165. doi:10.1016/j.cell.2013.02.026
Hoppins, S., Collins, S. R., Cassidy-Stone, A., Hummel, E., Devay, R. M., Lackner, L.
L., . . . Nunnari, J. (2011). A mitochondrial-focused genetic interaction map
reveals a scaffold-like complex required for inner membrane organization in
mitochondria. The Journal of cell biology, 195(2), 323-340.
doi:10.1083/jcb.201107053
Howard, J. P., Hutton, J. L., Olson, J. M., & Payne, G. S. (2002). Sla1p serves as the
targeting signal recognition factor for NPFX(1,2)D-mediated endocytosis. The
Journal of cell biology, 157(2), 315-326. doi:10.1083/jcb.200110027
Huang, W., Richards, S., Carbone, M. A., Zhu, D., Anholt, R. R., Ayroles, J. F., . . .
Mackay, T. F. (2012). Epistasis dominates the genetic architecture of Drosophila
quantitative traits. Proceedings of the National Academy of Sciences of the
United States of America, 109(39), 15553-15559. doi:10.1073/pnas.1213423109
Jarosz, D. F., & Lindquist, S. (2010). Hsp90 and environmental stress transform the
adaptive value of natural genetic variation. Science, 330(6012), 1820-1824.
doi:10.1126/science.1195487
89
Jarvis, J. P., & Cheverud, J. M. (2011). Mapping the epistatic network underlying murine
reproductive fatpad variation. Genetics, 187(2), 597-610.
doi:10.1534/genetics.110.123505
Jimenez-Gomez, J. M., Corwin, J. A., Joseph, B., Maloof, J. N., & Kliebenstein, D. J.
(2011). Genomic analysis of QTLs and genes altering natural variation in
stochastic noise. PLoS genetics, 7(9), e1002295.
doi:10.1371/journal.pgen.1002295
Joseph, B., Corwin, J. A., Zust, T., Li, B., Iravani, M., Schaepman-Strub, G., . . .
Kliebenstein, D. J. (2013). Hierarchical nuclear and cytoplasmic genetic
architectures for plant growth and defense within Arabidopsis. The Plant cell,
25(6), 1929-1945. doi:10.1105/tpc.113.112615
Kang, H. M., Sul, J. H., Service, S. K., Zaitlen, N. A., Kong, S. Y., Freimer, N. B., . . .
Eskin, E. (2010). Variance component model to account for sample structure in
genome-wide association studies. Nature genetics, 42(4), 348-354.
doi:10.1038/ng.548
Kao, C. H., Zeng, Z. B., & Teasdale, R. D. (1999). Multiple interval mapping for
quantitative trait loci. Genetics, 152(3), 1203-1216. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10388834
Kienle, S., & Sommer, R. J. (2013). Cryptic variation in vulva development by cis-
regulatory evolution of a HAIRY-binding site. Nat Commun, 4, 1714.
doi:10.1038/ncomms2711
Kim, H. Y., Lee, S. B., Kang, H. S., Oh, G. T., & Kim, T. (2014). Two distinct domains
of Flo8 activator mediates its role in transcriptional activation and the physical
interaction with Mss11. Biochem Biophys Res Commun, 449(2), 202-207.
doi:10.1016/j.bbrc.2014.04.161
Kim, T. S., Kim, H. Y., Yoon, J. H., & Kang, H. S. (2004). Recruitment of the Swi/Snf
complex by Ste12-Tec1 promotes Flo8-Mss11-mediated activation of STA1
expression. Mol Cell Biol, 24(21), 9542-9556. doi:10.1128/MCB.24.21.9542-
9556.2004
King, E. G., Merkes, C. M., McNeil, C. L., Hoofer, S. R., Sen, S., Broman, K. W., . . .
Macdonald, S. J. (2012). Genetic dissection of a model complex trait using the
Drosophila Synthetic Population Resource. Genome research, 22(8), 1558-1566.
doi:10.1101/gr.134031.111
Kobayashi, O., Suda, H., Ohtani, T., & Sone, H. (1996). Molecular cloning and analysis
of the dominant flocculation gene FLO8 from Saccharomyces cerevisiae. Mol
Gen Genet, 251(6), 707-715. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/8757402
Kover, P. X., Valdar, W., Trakalo, J., Scarcelli, N., Ehrenreich, I. M., Purugganan, M. D.,
. . . Mott, R. (2009). A Multiparent Advanced Generation Inter-Cross to fine-map
quantitative traits in Arabidopsis thaliana. PLoS genetics, 5(7), e1000551.
doi:10.1371/journal.pgen.1000551
Krauthammer, M., Kong, Y., Bacchiocchi, A., Evans, P., Pornputtapong, N., Wu, C., . . .
Halaban, R. (2015). Exome sequencing identifies recurrent mutations in NF1 and
RASopathy genes in sun-exposed melanomas. Nature genetics, 47(9), 996-1002.
doi:10.1038/ng.3361
Kuchin, S., Yeghiayan, P., & Carlson, M. (1995). Cyclin-dependent protein kinase and
cyclin homologs SSN3 and SSN8 contribute to transcriptional control in yeast.
Proceedings of the National Academy of Sciences of the United States of
90
America, 92(9), 4006-4010. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/7732022
Lander, E. S., & Botstein, D. (1989). Mapping mendelian factors underlying quantitative
traits using RFLP linkage maps. Genetics, 121(1), 185-199. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/2563713
Lauter, N., & Doebley, J. (2002). Genetic variation for phenotypically invariant traits
detected in teosinte: implications for the evolution of novel forms. Genetics,
160(1), 333-342. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/11805068
Le Rouzic, A., & Carlborg, O. (2008). Evolutionary potential of hidden genetic variation.
Trends Ecol Evol, 23(1), 33-37. doi:10.1016/j.tree.2007.09.014
Ledon-Rettig, C. C., Pfennig, D. W., & Crespi, E. J. (2010). Diet and hormonal
manipulation reveal cryptic genetic variation: implications for the evolution of
novel feeding strategies. Proc Biol Sci, 277(1700), 3569-3578.
doi:10.1098/rspb.2010.0877
Lee, J. T., Taylor, M. B., Shen, A., & Ehrenreich, I. M. (Submitted). The genotypic
landscape of environmental robustness. PLoS biology.
Lehner, B. (2011). Molecular mechanisms of epistasis within and between genes. Trends
in genetics : TIG, 27(8), 323-331. doi:10.1016/j.tig.2011.05.007
Leinonen, P. H., Remington, D. L., Leppala, J., & Savolainen, O. (2013). Genetic basis of
local adaptation and flowering time variation in Arabidopsis lyrata. Molecular
ecology, 22(3), 709-723. doi:10.1111/j.1365-294X.2012.05678.x
Lempe, J., Lachowiec, J., Sullivan, A. M., & Queitsch, C. (2013). Molecular mechanisms
of robustness in plants. Curr Opin Plant Biol, 16(1), 62-69.
doi:10.1016/j.pbi.2012.12.002
Levy, S. F., & Siegal, M. L. (2008). Network hubs buffer environmental variation in
Saccharomyces cerevisiae. PLoS biology, 6(11), e264.
doi:10.1371/journal.pbio.0060264
Levy, S. F., & Siegal, M. L. (2012). The robustness continuum. Adv Exp Med Biol, 751,
431-452. doi:10.1007/978-1-4614-3567-9_20
Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows-
Wheeler transform. Bioinformatics, 25(14), 1754-1760.
doi:10.1093/bioinformatics/btp324
Liti, G., Carter, D. M., Moses, A. M., Warringer, J., Parts, L., James, S. A., . . . Louis, E.
J. (2009). Population genomics of domestic and wild yeasts. Nature, 458(7236),
337-341. doi:10.1038/nature07743
Liu, H., Styles, C. A., & Fink, G. R. (1996). Saccharomyces cerevisiae S288C has a
mutation in FLO8, a gene required for filamentous growth. Genetics, 144(3),
967-978. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8913742
Lo, W. S., & Dranginis, A. M. (1996). FLO11, a yeast gene related to the STA genes,
encodes a novel cell surface flocculin. J Bacteriol, 178(24), 7144-7151.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8955395
Lorenz, M. C., & Heitman, J. (1998). Regulators of pseudohyphal differentiation in
Saccharomyces cerevisiae identified through multicopy suppressor analysis in
ammonium permease mutant strains. Genetics, 150(4), 1443-1457. Retrieved
from http://www.ncbi.nlm.nih.gov/pubmed/9832522
Lynch M, W. B. (1998). Genetics and analysis of quantitative traits: Sunderland, MA:
Sinauer.
91
Mackay, T. F. (2014). Epistasis and quantitative traits: using model organisms to study
gene-gene interactions. Nature reviews. Genetics, 15(1), 22-33.
doi:10.1038/nrg3627
Mackay, T. F., Richards, S., Stone, E. A., Barbadilla, A., Ayroles, J. F., Zhu, D., . . .
Gibbs, R. A. (2012). The Drosophila melanogaster Genetic Reference Panel.
Nature, 482(7384), 173-178. doi:10.1038/nature10811
Manolio, T. A., Collins, F. S., Cox, N. J., Goldstein, D. B., Hindorff, L. A., Hunter, D. J.,
. . . Visscher, P. M. (2009). Finding the missing heritability of complex diseases.
Nature, 461(7265), 747-753. doi:10.1038/nature08494
Masel, J., & Siegal, M. L. (2009). Robustness: mechanisms and consequences. Trends in
genetics : TIG, 25(9), 395-403. doi:10.1016/j.tig.2009.07.005
Masel, J., & Trotter, M. V. (2010). Robustness and evolvability. Trends in genetics : TIG,
26(9), 406-414. doi:10.1016/j.tig.2010.06.002
Matsui, T., Linder, R., Phan, J., Seidl, F., & Ehrenreich, I. M. (2015). Regulatory
Rewiring in a Cross Causes Extensive Genetic Heterogeneity. Genetics.
doi:10.1534/genetics.115.180661
McDonald, C. M., Wagner, M., Dunham, M. J., Shin, M. E., Ahmed, N. T., & Winter, E.
(2009). The Ras/cAMP pathway and the CDK-like kinase Ime2 regulate the
MAPK Smk1 and spore morphogenesis in Saccharomyces cerevisiae. Genetics,
181(2), 511-523. doi:10.1534/genetics.108.098434
McVicker, G., van de Geijn, B., Degner, J. F., Cain, C. E., Banovich, N. E., Raj, A., . . .
Pritchard, J. K. (2013). Identification of genetic variants that affect histone
modifications in human cells. Science, 342(6159), 747-749.
doi:10.1126/science.1242429
Michelmore, R. W., Paran, I., & Kesseli, R. V. (1991). Identification of markers linked to
disease-resistance genes by bulked segregant analysis: a rapid method to detect
markers in specific genomic regions by using segregating populations.
Proceedings of the National Academy of Sciences of the United States of
America, 88(21), 9828-9832. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/1682921
Milloz, J., Duveau, F., Nuez, I., & Felix, M. A. (2008). Intraspecific evolution of the
intercellular signaling network underlying a robust developmental system. Genes
Dev, 22(21), 3064-3075. doi:10.1101/gad.495308
Moczek, A. P. (2008). On the origins of novelty in development and evolution. BioEssays
: news and reviews in molecular, cellular and developmental biology, 30(5), 432-
447. doi:10.1002/bies.20754
Nadeau, J. H. (2001). Modifier genes in mice and humans. Nature reviews. Genetics,
2(3), 165-174. doi:10.1038/35056009
Nadeau, J. H., Singer, J. B., Matin, A., & Lander, E. S. (2000). Analysing complex
genetic traits with chromosome substitution strains. Nature genetics, 24(3), 221-
225. doi:10.1038/73427
Nelson, R. M., Pettersson, M. E., & Carlborg, O. (2013). A century after Fisher: time for
a new paradigm in quantitative genetics. Trends in genetics : TIG.
doi:10.1016/j.tig.2013.09.006
Neurofibromatosis Fact Sheet. (May 2011). NINDS, 11-2126.
Nijhout, H. F. (2002). The nature of robustness in development. BioEssays : news and
reviews in molecular, cellular and developmental biology, 24(6), 553-563.
doi:10.1002/bies.10093
92
Nuzhdin, S. V., Friesen, M. L., & McIntyre, L. M. (2012). Genotype-phenotype mapping
in a post-GWAS world. Trends in genetics : TIG, 28(9), 421-426.
doi:10.1016/j.tig.2012.06.003
Paaby, A. B., & Rockman, M. V. (2014). Cryptic genetic variation: evolution's hidden
substrate. Nature reviews. Genetics, 15(4), 247-258. doi:10.1038/nrg3688
Paaby, A. B., White, A. G., Riccardi, D. D., Gunsalus, K. C., Piano, F., & Rockman, M.
V. (2015). Wild worm embryogenesis harbors ubiquitous polygenic modifier
variation. Elife, 4. doi:10.7554/eLife.09178
Pan, X., & Heitman, J. (2002). Protein kinase A operates a molecular switch that governs
yeast pseudohyphal differentiation. Mol Cell Biol, 22(12), 3981-3993. Retrieved
from http://www.ncbi.nlm.nih.gov/pubmed/12024012
Pare, G., Cook, N. R., Ridker, P. M., & Chasman, D. I. (2010). On the use of variance per
genotype as a tool to identify quantitative trait interaction effects: a report from
the Women's Genome Health Study. PLoS genetics, 6(6), e1000981.
doi:10.1371/journal.pgen.1000981
Pedrajas, J. R., Kosmidou, E., Miranda-Vizuete, A., Gustafsson, J. A., Wright, A. P., &
Spyrou, G. (1999). Identification and functional characterization of a novel
mitochondrial thioredoxin system in Saccharomyces cerevisiae. The Journal of
biological chemistry, 274(10), 6366-6373. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10037727
Pettersson, M., Besnier, F., Siegel, P. B., & Carlborg, O. (2011). Replication and
explorations of high-order epistasis using a large advanced intercross line
pedigree. PLoS genetics, 7(7), e1002180. doi:10.1371/journal.pgen.1002180
Pfennig, D. W., & Ehrenreich, I. M. (2014). Towards a gene regulatory network
perspective on phenotypic plasticity, genetic accommodation and genetic
assimilation. Molecular ecology, 23(18), 4438-4440. doi:10.1111/mec.12887
Phan, V. T., Ding, V. W., Li, F., Chalkley, R. J., Burlingame, A., & McCormick, F.
(2010). The RasGAP proteins Ira2 and neurofibromin are negatively regulated by
Gpb1 in yeast and ETEA in humans. Mol Cell Biol, 30(9), 2264-2279.
doi:10.1128/MCB.01450-08
Phillips, P. C. (2008). Epistasis--the essential role of gene interactions in the structure and
evolution of genetic systems. Nature reviews. Genetics, 9(11), 855-867.
doi:10.1038/nrg2452
Pickrell, J. K., Marioni, J. C., Pai, A. A., Degner, J. F., Engelhardt, B. E., Nkadori, E., . . .
Pritchard, J. K. (2010). Understanding mechanisms underlying human gene
expression variation with RNA sequencing. Nature, 464(7289), 768-772.
doi:10.1038/nature08872
Pigliucci, M., & Murren, C. J. (2003). Perspective: Genetic assimilation and a possible
evolutionary paradox: can macroevolution sometimes be so fast as to pass us by?
Evolution, 57(7), 1455-1464. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/12940351
Pigliucci, M., Murren, C. J., & Schlichting, C. D. (2006). Phenotypic plasticity and
evolution by genetic assimilation. J Exp Biol, 209(Pt 12), 2362-2367.
doi:10.1242/jeb.02070
Pir, P., Gutteridge, A., Wu, J., Rash, B., Kell, D. B., Zhang, N., & Oliver, S. G. (2012).
The genetic control of growth rate: a systems biology study in yeast. BMC
systems biology, 6, 4. doi:10.1186/1752-0509-6-4
93
Pritchard, J. K., Stephens, M., Rosenberg, N. A., & Donnelly, P. (2000). Association
mapping in structured populations. Am J Hum Genet, 67(1), 170-181.
doi:10.1086/302959
Queitsch, C., Carlson, K. D., & Girirajan, S. (2012). Lessons from model organisms:
phenotypic robustness and missing heritability in complex disease. PLoS
genetics, 8(11), e1003041. doi:10.1371/journal.pgen.1003041
Queitsch, C., Sangster, T. A., & Lindquist, S. (2002). Hsp90 as a capacitor of phenotypic
variation. Nature, 417(6889), 618-624. doi:10.1038/nature749
Rand, D. M., Haney, R. A., & Fry, A. J. (2004). Cytonuclear coevolution: the genomics
of cooperation. Trends Ecol Evol, 19(12), 645-653.
doi:10.1016/j.tree.2004.10.003
Richardson, J. B., Uppendahl, L. D., Traficante, M. K., Levy, S. F., & Siegal, M. L.
(2013). Histone variant HTZ1 shows extensive epistasis with, but does not
increase robustness to, new mutations. PLoS genetics, 9(8), e1003733.
doi:10.1371/journal.pgen.1003733
Ritchie, M. D., Hahn, L. W., Roodi, N., Bailey, L. R., Dupont, W. D., Parl, F. F., &
Moore, J. H. (2001). Multifactor-dimensionality reduction reveals high-order
interactions among estrogen-metabolism genes in sporadic breast cancer. Am J
Hum Genet, 69(1), 138-147. doi:10.1086/321276
Robertson, L. S., & Fink, G. R. (1998). The three yeast A kinases have specific signaling
functions in pseudohyphal growth. Proceedings of the National Academy of
Sciences of the United States of America, 95(23), 13783-13787. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/9811878
Rohner, N., Jarosz, D. F., Kowalko, J. E., Yoshizawa, M., Jeffery, W. R., Borowsky, R.
L., . . . Tabin, C. J. (2013). Cryptic variation in morphological evolution: HSP90
as a capacitor for loss of eyes in cavefish. Science, 342(6164), 1372-1375.
doi:10.1126/science.1240276
Roop, J. I., & Brem, R. B. (2013). Rare variants in hypermutable genes underlie common
morphology and growth traits in wild Saccharomyces paradoxus. Genetics,
195(2), 513-525. doi:10.1534/genetics.113.155341
Rosas, U., Barton, N. H., Copsey, L., Barbier de Reuille, P., & Coen, E. (2010). Cryptic
variation between species and the basis of hybrid performance. PLoS biology,
8(7), e1000429. doi:10.1371/journal.pbio.1000429
Ross, S. J., Findlay, V. J., Malakasi, P., & Morgan, B. A. (2000). Thioredoxin peroxidase
is required for the transcriptional response to oxidative stress in budding yeast.
Mol Biol Cell, 11(8), 2631-2642. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10930459
Rowe, H. C., Hansen, B. G., Halkier, B. A., & Kliebenstein, D. J. (2008). Biochemical
networks and epistasis shape the Arabidopsis thaliana metabolome. The Plant
cell, 20(5), 1199-1216. doi:10.1105/tpc.108.058131
Rutherford, S. L. (2000). From genotype to phenotype: buffering mechanisms and the
storage of genetic information. BioEssays : news and reviews in molecular,
cellular and developmental biology, 22(12), 1095-1105. doi:10.1002/1521-
1878(200012)22:12<1095::AID-BIES7>3.0.CO;2-A
Rutherford, S. L., & Lindquist, S. (1998). Hsp90 as a capacitor for morphological
evolution. Nature, 396(6709), 336-342. doi:10.1038/24550
94
Ryan, O., Shapiro, R. S., Kurat, C. F., Mayhew, D., Baryshnikova, A., Chin, B., . . .
Boone, C. (2012). Global gene deletion analysis exploring yeast filamentous
growth. Science, 337(6100), 1353-1356. doi:10.1126/science.1224339
Sangster, T. A., Lindquist, S., & Queitsch, C. (2004). Under cover: causes, effects and
implications of Hsp90-mediated genetic capacitance. BioEssays : news and
reviews in molecular, cellular and developmental biology, 26(4), 348-362.
doi:10.1002/bies.20020
Sangster, T. A., Salathia, N., Lee, H. N., Watanabe, E., Schellenberg, K., Morneau, K., . .
. Lindquist, S. (2008). HSP90-buffered genetic variation is common in
Arabidopsis thaliana. Proceedings of the National Academy of Sciences of the
United States of America, 105(8), 2969-2974. doi:10.1073/pnas.0712210105
Sangster, T. A., Salathia, N., Undurraga, S., Milo, R., Schellenberg, K., Lindquist, S., &
Queitsch, C. (2008). HSP90 affects the expression of genetic variation and
developmental stability in quantitative traits. Proceedings of the National
Academy of Sciences of the United States of America, 105(8), 2963-2968.
doi:10.1073/pnas.0712200105
Schacherer, J., Shapiro, J. A., Ruderfer, D. M., & Kruglyak, L. (2009). Comprehensive
polymorphism survey elucidates population structure of Saccharomyces
cerevisiae. Nature, 458(7236), 342-345. doi:10.1038/nature07670
Service, P. M. (2004). How good are quantitative complementation tests? Science of
aging knowledge environment : SAGE KE, 2004(12), pe13.
doi:10.1126/sageke.2004.12.pe13
Shao, H., Burrage, L. C., Sinasac, D. S., Hill, A. E., Ernest, S. R., O'Brien, W., . . .
Nadeau, J. H. (2008). Genetic architecture of complex traits: large phenotypic
effects and pervasive epistasis. Proceedings of the National Academy of Sciences
of the United States of America, 105(50), 19910-19914.
doi:10.1073/pnas.0810388105
Sherman, F. (1991). Guide to Yeast Genetics and Molecular. In C. Guthrie & G. R. Fink
(Eds.), Methods in enzymology (pp. 3-21). San Diego, California: Elsevier
Academic Press.
Siegal, M. L., & Bergman, A. (2002). Waddington's canalization revisited:
developmental stability and evolution. Proceedings of the National Academy of
Sciences of the United States of America, 99(16), 10528-10532.
doi:10.1073/pnas.102303999
Siegal, M. L., & Leu, J. Y. (2014). On the Nature and Evolutionary Impact of Phenotypic
Robustness Mechanisms. Annu Rev Ecol Evol Syst, 45, 496-517.
doi:10.1146/annurev-ecolsys-120213-091705
Sinha, H., Nicholson, B. P., Steinmetz, L. M., & McCusker, J. H. (2006). Complex
genetic interactions in a quantitative trait locus. PLoS genetics, 2(2), e13.
doi:10.1371/journal.pgen.0020013
Sollars, V., Lu, X., Xiao, L., Wang, X., Garfinkel, M. D., & Ruden, D. M. (2003).
Evidence for an epigenetic mechanism by which Hsp90 acts as a capacitor for
morphological evolution. Nature genetics, 33(1), 70-74. doi:10.1038/ng1067
Song, W., & Carlson, M. (1998). Srb/mediator proteins interact functionally and
physically with transcriptional repressor Sfl1. EMBO J, 17(19), 5757-5765.
doi:10.1093/emboj/17.19.5757
95
Spiezio, S. H., Takada, T., Shiroishi, T., & Nadeau, J. H. (2012). Genetic divergence and
the genetic architecture of complex traits in chromosome substitution strains of
mice. BMC genetics, 13, 38. doi:10.1186/1471-2156-13-38
Steinmetz, L. M., Sinha, H., Richards, D. R., Spiegelman, J. I., Oefner, P. J., McCusker,
J. H., & Davis, R. W. (2002). Dissecting the architecture of a quantitative trait
locus in yeast. Nature, 416(6878), 326-330. doi:10.1038/416326a
Stelling, J., Sauer, U., Szallasi, Z., Doyle, F. J., 3rd, & Doyle, J. (2004). Robustness of
cellular functions. Cell, 118(6), 675-685. doi:10.1016/j.cell.2004.09.008
Storey, J. D., Akey, J. M., & Kruglyak, L. (2005). Multiple locus linkage analysis of
genomewide expression in yeast. PLoS biology, 3(8), e267.
doi:10.1371/journal.pbio.0030267
Storici, F., Lewis, L. K., & Resnick, M. A. (2001). In vivo site-directed mutagenesis
using oligonucleotides. Nature biotechnology, 19(8), 773-776.
doi:10.1038/90837
Struchalin, M. V., Dehghan, A., Witteman, J. C., van Duijn, C., & Aulchenko, Y. S.
(2010). Variance heterogeneity analysis for detection of potentially interacting
genetic loci: method and its limitations. BMC genetics, 11, 92. doi:10.1186/1471-
2156-11-92
Suzuki, Y., & Nijhout, H. F. (2006). Evolution of a polyphenism by genetic
accommodation. Science, 311(5761), 650-652. doi:10.1126/science.1118888
Swarup, S., Harbison, S. T., Hahn, L. E., Morozova, T. V., Yamamoto, A., Mackay, T.
F., & Anholt, R. R. (2012). Extensive epistasis for olfactory behaviour, sleep and
waking activity in Drosophila melanogaster. Genetics research, 94(1), 9-20.
doi:10.1017/S001667231200002X
Takahashi, K. H. (2013). Multiple capacitors for natural genetic variation in Drosophila
melanogaster. Molecular ecology, 22(5), 1356-1365. doi:10.1111/mec.12091
Tan, Z., Hays, M., Cromie, G. A., Jeffery, E. W., Scott, A. C., Ahyong, V., . . . Dudley,
A. M. (2013). Aneuploidy underlies a multicellular phenotypic switch.
Proceedings of the National Academy of Sciences of the United States of
America, 110(30), 12367-12372. doi:10.1073/pnas.1301047110
Tanaka, K., Nakafuku, M., Satoh, T., Marshall, M. S., Gibbs, J. B., Matsumoto, K., . . .
Toh-e, A. (1990). S. cerevisiae genes IRA1 and IRA2 encode proteins that may
be functionally equivalent to mammalian ras GTPase activating protein. Cell,
60(5), 803-807. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/2178777
Tanaka, K., Nakafuku, M., Tamanoi, F., Kaziro, Y., Matsumoto, K., & Toh-e, A. (1990).
IRA2, a second gene of Saccharomyces cerevisiae that encodes a protein with a
domain homologous to mammalian ras GTPase-activating protein. Mol Cell Biol,
10(8), 4303-4313. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/2164637
Tang, H. Y., Xu, J., & Cai, M. (2000). Pan1p, End3p, and S1a1p, three yeast proteins
required for normal cortical actin cytoskeleton organization, associate with each
other and play essential roles in cell wall morphogenesis. Mol Cell Biol, 20(1),
12-25. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10594004
Tarassov, K., Messier, V., Landry, C. R., Radinovic, S., Serna Molina, M. M., Shames, I.,
. . . Michnick, S. W. (2008). An in vivo map of the yeast protein interactome.
Science, 320(5882), 1465-1470. doi:10.1126/science.1153878
96
Taylor, M. B., & Ehrenreich, I. M. (2014). Genetic interactions involving five or more
genes contribute to a complex trait in yeast. PLoS genetics, 10(5), e1004324.
doi:10.1371/journal.pgen.1004324
Taylor, M. B., & Ehrenreich, I. M. (2015a). Higher-order genetic interactions and their
contribution to complex traits. Trends in genetics : TIG, 31(1), 34-40.
doi:10.1016/j.tig.2014.09.001
Taylor, M. B., & Ehrenreich, I. M. (2015b). Transcriptional Derepression Uncovers
Cryptic Higher-Order Genetic Interactions. PLoS genetics, 11(10), e1005606.
doi:10.1371/journal.pgen.1005606
Taylor, M. B., & Ehrenreich, I. M. (In press). Transcriptional derepression uncovers
cryptic higher-order genetic interactions. PLoS genetics.
Threadgill, D. W., Dlugosz, A. A., Hansen, L. A., Tennenbaum, T., Lichti, U., Yee, D., . .
. et al. (1995). Targeted disruption of mouse EGF receptor: effect of genetic
background on mutant phenotype. Science, 269(5221), 230-234. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/7618084
Tirosh, I., Reikhav, S., Sigal, N., Assia, Y., & Barkai, N. (2010). Chromatin regulators as
capacitors of interspecies variations in gene expression. Molecular systems
biology, 6, 435. doi:10.1038/msb.2010.84
Tong, A. H., Evangelista, M., Parsons, A. B., Xu, H., Bader, G. D., Page, N., . . . Boone,
C. (2001). Systematic genetic analysis with ordered arrays of yeast deletion
mutants. Science, 294(5550), 2364-2368. doi:10.1126/science.1065810
Trotter, E. W., & Grant, C. M. (2002). Thioredoxins are required for protection against a
reductive stress in the yeast Saccharomyces cerevisiae. Mol Microbiol, 46(3),
869-878. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12410842
True, H. L., Berlin, I., & Lindquist, S. L. (2004). Epigenetic regulation of translation
reveals hidden genetic variation to produce complex traits. Nature, 431(7005),
184-187. doi:10.1038/nature02885
True, H. L., & Lindquist, S. L. (2000). A yeast prion provides a mechanism for genetic
variation and phenotypic diversity. Nature, 407(6803), 477-483.
doi:10.1038/35035005
Turner, T. L. (2014). Fine-mapping natural alleles: quantitative complementation to the
rescue. Molecular ecology, 23(10), 2377-2382. doi:10.1111/mec.12719
van Nimwegen, E., Crutchfield, J. P., & Huynen, M. (1999). Neutral evolution of
mutational robustness. Proceedings of the National Academy of Sciences of the
United States of America, 96(17), 9716-9720. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/10449760
Venters, B. J., Wachi, S., Mavrich, T. N., Andersen, B. E., Jena, P., Sinnamon, A. J., . . .
Pugh, B. F. (2011). A comprehensive genomic binding map of gene and
chromatin regulatory proteins in Saccharomyces. Mol Cell, 41(4), 480-492.
doi:10.1016/j.molcel.2011.01.015
Waddington, C. H. (1942). Canalization of development and the inheritance of acquired
characters. Nature, 150, 563-565. doi:DOI 10.1038/150563a0
Waddington, C. H. (1953). Genetic assimilation of an acquired character. Evolution, 7(2),
118-126.
Wagner, G. P., Booth, G., & Bagheri, H. C. (1997). A population genetic theory of
canalization. Evolution, 51(2), 329-347. doi:Doi 10.2307/2411105
97
Wang, Y., Arenas, C. D., Stoebel, D. M., & Cooper, T. F. (2013). Genetic background
affects epistatic interactions between two beneficial mutations. Biology letters,
9(1), 20120328. doi:10.1098/rsbl.2012.0328
Weinreich, D. M., Lan, Y., Wylie, C. S., & Heckendorn, R. B. (2013). Should
evolutionary geneticists worry about higher-order epistasis? Curr Opin Genet
Dev, 23(6), 700-707. doi:10.1016/j.gde.2013.10.007
Wenger, J. W., Schwartz, K., & Sherlock, G. (2010). Bulk segregant analysis by high-
throughput sequencing reveals a novel xylose utilization gene from
Saccharomyces cerevisiae. PLoS genetics, 6(5), e1000942.
doi:10.1371/journal.pgen.1000942
Wilke, C. O., & Adami, C. (2003). Evolution of mutational robustness. Mutat Res, 522(1-
2), 3-11. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12517406
Wilkening, S., Lin, G., Fritsch, E. S., Tekkedil, M. M., Anders, S., Kuehn, R., . . .
Steinmetz, L. M. (2013). An Evaluation of High-Throughput Approaches to QTL
Mapping in Saccharomyces cerevisiae. Genetics.
doi:10.1534/genetics.113.160291
Wilkinson, R., & Wiedenheft, B. (2014). A CRISPR method for genome engineering.
F1000prime reports, 6, 3. doi:10.12703/P6-3
Wray, G. A., Hahn, M. W., Abouheif, E., Balhoff, J. P., Pizer, M., Rockman, M. V., &
Romano, L. A. (2003). The evolution of transcriptional regulation in eukaryotes.
Molecular biology and evolution, 20(9), 1377-1419. doi:10.1093/molbev/msg140
Wu, J., Suka, N., Carlson, M., & Grunstein, M. (2001). TUP1 utilizes histone H3/H2B-
specific HDA1 deacetylase to repress gene activity in yeast. Mol Cell, 7(1), 117-
126. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/11172717
Zheng, W., Zhao, H., Mancera, E., Steinmetz, L. M., & Snyder, M. (2010). Genetic
analysis of variation in transcription factor binding in yeast. Nature, 464(7292),
1187-1191. doi:10.1038/nature08934
Zhu, C. T., Ingelmo, P., & Rand, D. M. (2014). GxGxE for Lifespan in Drosophila:
Mitochondrial, Nuclear, and Dietary Interactions that Modify Longevity. PLoS
genetics, 10(5), e1004354. doi:10.1371/journal.pgen.1004354
Zuk, O., Hechter, E., Sunyaev, S. R., & Lander, E. S. (2012). The mystery of missing
heritability: Genetic interactions create phantom heritability. Proceedings of the
National Academy of Sciences of the United States of America, 109(4), 1193-
1198. doi:10.1073/pnas.1119675109
98
Appendix A: Higher-order genetic interactions and their contribution to
complex traits
This work appears essentially as published in 2014 in
Trends in Genetics, 31(1): 34-40
A.1 Overview
The contribution of genetic interactions involving three or more loci to complex traits is
poorly understood. Because these higher-order genetic interactions (HGIs) are difficult to
detect in genetic mapping studies, very few examples of them have been described.
However, the lack of data on HGIs should not be misconstrued as proof that this class of
genetic effect is unimportant. To the contrary, evidence from model organisms suggests
that HGIs frequently influence genetic studies and contribute to many complex traits.
Here, we review the growing literature on HGIs and discuss the future of research on this
topic.
A.2 HGIs merit deeper investigation as a source of complex trait variation
Genetic interactions (sometimes referred to as epistatic interactions) contribute to many
complex traits. Despite widespread recognition of this point (Carlborg & Haley, 2004;
Frankel & Schork, 1996; Huang et al., 2012; Mackay, 2014; Nelson et al., 2013; Phillips,
2008), relatively little is known about the specific forms of genetic interactions that are
important to heritable phenotypic variation. To date, researchers have mainly reported
genetic interactions involving only two loci (e.g., (Brem et al., 2005; Caicedo,
Stinchcombe, Olsen, Schmitt, & Purugganan, 2004; Gaertner, Parmenter, Rockman,
Kruglyak, & Phillips, 2012; Jarvis & Cheverud, 2011; Rowe, Hansen, Halkier, &
Kliebenstein, 2008)). However, this emphasis on gene-gene interactions over higher-
order genetic interactions (HGIs) involving three or more loci is rooted in technical issues
99
(Fig. A.1), rather than biology. As others have noted, conventional genetic mapping
methods have generally low statistical power to identify genetic interactions and higher
statistical power to detect gene-gene interactions than HGIs (Carlborg, Jacobsson,
Ahgren, Siegel, & Andersson, 2006; Cordell, 2009). These biases mean that most genetic
studies are unlikely to detect HGIs regardless of whether they affect a trait or not, and
make it difficult to determine the overall role of HGIs in heritable phenotypes.
100
Fig. A.1 | The types of genetic effects discussed in this manuscript. Additive effects
(A), gene-gene interactions (B), and a HGI (C). For simplicity, only haploids are
depicted and the HGI is shown with three loci. However, we note that HGIs can be
more complex. In each plot, different genotypes across the involved loci are shown
across the x-axis, while the average phenotype among individuals with a given
genotype is shown along the y-axis. Different alleles are indicated by blue and
orange, a coloring scheme used throughout the paper. For additive effects, loci are
considered in isolation, whereas for interactions, a number of other loci can be
considered to determine if combinatorial genotypes have unique effects on
phenotype.
In this review, we examine the potential contribution of HGIs to complex traits. We write
from the perspective that HGIs merit deeper investigation as a source of ‘missing
101
heritability’ in humans and model systems (Manolio et al., 2009; Zuk, Hechter, Sunyaev,
& Lander). This assertion is based upon not only the technical considerations described
above, but also multiple recent papers that have shown that HGIs can have major
phenotypic effects (e.g., (Chandler et al., 2014; Hanlon et al., 2006; Pettersson et al.,
2011; Taylor & Ehrenreich, 2014)). Our goal with the present manuscript is to summarize
these studies, which are fairly small in total number, and aggregate them with other
research that provides less direct support for HGIs. With the data in hand, readers can
then form their own opinions about the potential importance of HGIs. For researchers
interested in identifying HGIs in their own systems, we also describe some of the
approaches that have successfully detected loci involved in HGIs, and speculate on causal
mechanisms that further characterization of these interacting loci might uncover. We
conclude the paper by raising some ‘big picture’ questions about HGIs that future
research will hopefully address.
A.3 Evidence for HGIs
In this section, we summarize some of the strongest evidence that HGIs contribute to
complex traits. Support for HGIs falls into two main classes—suggestive and direct. We
employ the term ‘suggestive’ to refer to cases where a study’s results indicated that HGIs
might contribute to a trait, but the specific loci involved in a HGI were not identified. In
contrast, we use the term ‘direct’ to refer to instances where multiple loci involved in a
HGI were detected (as summarized in Fig. A.2). We focus exclusively on results from
model organisms, but note that HGIs have also been reported in humans (Collins et al.,
2013; Ritchie et al., 2001).
102
Fig. A.2 | Examples of HGIs from model organism research. Arabidopsis metabolite
levels (Joseph et al., 2013), chicken body weight (Pettersson et al., 2011), fruit fly
wing shape (Chandler et al., 2014), yeast colony morphology (Taylor & Ehrenreich,
2014), and yeast sporulation (Gerke, Lorenz, Ramnarine, & Cohen, 2010). Different
types of evidence for HGIs are distinguished by color and shape.
A.3.1 Suggestive evidence
Unexplained epistatic genetic variance in crosses: Under many experimental designs, it is
possible to estimate the total contribution of genetic interactions to a trait (Falconer DS,
1996; Lynch M, 1998). However, such estimates of the epistatic genetic variance
typically do not distinguish between the effects of gene-gene interactions and HGIs. With
a sufficiently large mapping population, one can attempt to separate these effects by
scanning for gene-gene interactions and determining what proportion of the epistatic
genetic variance they leave unexplained. A recent study focused on a cross of two yeast
strains employed this strategy (Bloom et al., 2013). By using a large mapping population
of 1,006 haploid segregants, genotyping each recombinant across the genome, and
examining the strains for 46 growth traits under controlled conditions, the authors were
103
able to show that genetic interactions played a sizable role in many traits. However, gene-
gene interactions detected for these traits did not explain most of their epistatic genetic
variance, despite the large mapping population and high power for detecting pairwise
interactions. As the authors suggest, there are a number of possible explanations for this
finding, one of which is that HGIs contribute to some of the phenotypes (Bloom et al.,
2013).
Over-explanation of phenotypic differences using genome-wide chromosome substitution
panels: Another source of suggestive information regarding HGIs comes from work on
panels of chromosome substitution strains, which consist of individuals that each carry a
single chromosome from a ‘donor’ parent and the rest of their chromosomes from a
different ‘recipient’ parent (Nadeau, Singer, Matin, & Lander, 2000). Chromosome
substitution panels include one strain for each chromosome, facilitating the measurement
of every chromosome’s contribution to a trait. If loci influencing a trait act in a largely
additive manner, then effects measured in the chromosome substitution panels should add
up to the phenotypic difference between the donor and recipient parents used to generate
a population. However, departures from this expectation imply that genetic interactions
contribute to a trait. In one of the most comprehensive studies of this type, researchers
examined 144 blood, bone, and metabolic traits in mouse and rat genome-wide
chromosome substitution panels (Shao et al., 2008). They found that the cumulative
phenotypic effect of individual chromosome substitutions was often more than 100% of
the phenotypic difference between the two parents of a panel, with more than 500% of
the difference accounted for in some instances (Shao et al., 2008). This finding, which
has been replicated in additional chromosome substitution panels (Spiezio, Takada,
104
Shiroishi, & Nadeau, 2012), suggests that genetic interactions among loci on a number of
chromosomes contributed to many of the examined traits. Whether this involved many
gene-gene interactions, HGIs, or a mixture of the two has not been determined.
Genetic background effects and conditional essential mutations: Work on genetic
background effects also suggests that HGIs make a significant contribution to many
traits. This phenomenon occurs when mutations or sequence polymorphisms vary in their
effects across genetically diverse individuals (e.g., (Sinha et al., 2006; Swarup et al.,
2012; Threadgill et al., 1995)). Research in multiple species has shown that genetic
background effects are very common (as reviewed in (Chandler et al., 2013; Mackay,
2014; Nadeau, 2001)). Recent studies indicate that genetic background effects may arise
due to HGIs among natural genetic variants (Chandler et al., 2014; Dowell et al., 2010).
One of the most detailed studies along these lines was conducted in yeast (Dowell et al.,
2010). Researchers examined the effects of null alleles of nearly every gene in the
Saccharomyces cerevisiae genome on viability in two genetically diverged strains. 57
genes were found to be essential for growth in rich glucose medium in only one of the
two strains. To determine the architecture of this ‘conditional essentiality’, they measured
the frequency at which null alleles caused inviability in crosses of the two strains. Based
on these data, the authors estimated that conditional essentiality was often due to two or
more interacting loci, suggesting that HGIs occurred among the null alleles and multiple
genetic variants present in the two examined strains (Dowell et al., 2010).
105
A.3.2 Direct evidence
Gene-gene interactions that are influenced by genetic background: Further evidence for
HGIs comes from cases in which gene-gene interactions underlying a trait were
identified, but only had effects in specific genetic backgrounds (e.g., (Gerke et al., 2010;
Wang, Arenas, Stoebel, & Cooper, 2013)). A possible explanation for this finding is that
the two initially identified loci participate in HGIs with additional loci. Some of the most
detailed work on genetic background-dependent gene-gene interactions comes from
studies of sporulation efficiency in yeast (Gerke, Lorenz, & Cohen, 2009; Gerke et al.,
2010). In a cross of two isolates, the authors identified four SNPs in the genes IME1,
RME1, and RSF1 that together explain nearly all of the heritable variation in sporulation
efficiency when cultures are shifted from glucose to acetate (Gerke et al., 2009). These
SNPs largely affect the trait through additive effects and gene-gene interactions.
However, a subsequent examination of the 16 different homozygous genotypes involving
these 4 nucleotides across both of the parental genetic backgrounds and 8 different
conditions revealed a high context-dependence to these gene-gene interactions
(Ehrenreich et al., 2010). Indeed, only a single gene-gene interaction was significant
across genetic backgrounds and environments, suggesting that unidentified loci may
participate in HGIs with the polymorphisms in IME1, RME1, and RSF1.
Genetic background-dependent interactions between mutations: Similar to gene-gene
interactions among naturally occurring genetic variants, interactions between mutations
can also be affected by genetic background (Chandler et al., 2013). This implies that
genetic variants among strains may form HGIs with these mutations. Work in Drosophila
melanogaster on the scalloped
E3
(sd
E3
) mutation, which influences wing shape, has
106
provided some of the deepest insights into this phenomenon (Chandler et al., 2014; Chari
& Dworkin, 2013). In a large-scale examination of the effects of modifier mutations on
sd
E3
in two strains, the authors found that 74% of the modifiers were affected by genetic
background (Chari & Dworkin, 2013). They then used a combination of approaches,
including measurement of global gene expression, genetic mapping, and integrative
bioinformatics analysis, to determine the genetic basis of variability in the effects of sd
E3
modifiers (Chandler et al., 2014). Their results indicated that these genetic background-
dependent effects on interactions between sd
E3
and its modifiers were shaped by HGIs
involving these mutations and additional segregating loci (Chandler et al., 2014).
Cytonuclear interactions involving three or more nuclear and cytoplasmic factors: An
additional source of evidence for HGIs is research on cytonuclear interactions, which
arise due to genetic interactions between nuclear loci and cytoplasmic factors (Rand,
Haney, & Fry, 2004). Although the chloroplast and mitochondria are the most frequently
studied cytoplasmic factors, plasmids, prions, and viruses can also have heritable
phenotypic effects (Edwards, Symbor-Nagrabska, Dollard, Gifford, & Fink, 2014;
Halfmann et al., 2012). Work on a number of species has shown that cytonuclear
interactions are pervasive (Rand et al., 2004) and can significantly influence complex
traits, such as longevity in flies (Zhu, Ingelmo, & Rand, 2014), flowering time, growth,
and defense against herbivores and pathogens in plants (Joseph et al., 2013; Leinonen,
Remington, Leppala, & Savolainen, 2013), and the growth effects of gene deletions in
yeast (Edwards et al., 2014). A recent study on heritable variation in metabolism in an
Arabidopsis cross is particularly relevant to the discussion of HGIs (Joseph et al., 2013).
The authors showed that many gene-gene interactions among nuclear loci were affected
107
by an individual’s cytoplasmic genotype. Although the relative contributions of the
chloroplast and mitochondria to this observation could not be determined, this paper
shows that HGIs between two nuclear and at least a single cytoplasmic locus can occur.
Examples of comprehensively dissected HGIs: Arguably the strongest support for HGIs
comes from examples in which the specific loci involved in an HGI have been identified.
Unfortunately, the number of these examples is small (e.g., (Hanlon et al., 2006;
Pettersson et al., 2011; Taylor & Ehrenreich, 2014)). Here, we summarize two of the
papers that have most convincingly demonstrated HGIs: one that described three-locus
interactions that influence body weight in a cross of two chicken lines (Pettersson et al.,
2011), and another that showed a pair of five- and seven-locus genetic interactions that
determine colony morphology in a cross of two yeast strains (Taylor & Ehrenreich,
2014). The work on chickens was a follow-up study to an artificial selection experiment
that identified a set of five loci that each showed a gene-gene interaction with the same
central locus (Gjuvsland et al., 2007). In the subsequent study of these interacting loci,
the authors used a large advanced intercross population to replicate the previously
identified gene-gene interactions and show that some of the loci participated in three-
locus genetic interactions (Pettersson et al., 2011). The HGIs identified in yeast led to a
‘rough’ colony phenotype that unexpectedly segregated at a low frequency in a cross
(Taylor & Ehrenreich, 2014). Although many isolates of S. cerevisiae exhibit unusual
colony morphologies (e.g., (Granek & Magwene, 2010; Ryan et al., 2012)), the cross
employed in this paper involved two parents with the same ‘smooth’ colony morphology
(Taylor & Ehrenreich, 2014). The authors screened thousands of backcross progeny,
sequenced the genomes of hundreds of ‘rough’ progeny, and performed a large number of
108
genetic manipulations to identify alleles at five genes that interact to specify the
phenotype. They then conducted additional experiments to show that one of the non-
causal alleles can lead to the phenotype through an interaction between it and a specific
combination of alleles at two other loci. Therefore, HGIs involving five or seven genes
specify ‘rough’ morphology. The authors also showed that both of these HGIs critically
depend on a spontaneous mutation that occurred in IRA2, a gene that regulates the Ras
pathway.
A.4 Identifying loci and genes involved in HGIs
Given the small number of characterized HGIs, dissection of more examples to their
underlying loci and genes is needed to significantly improve our understanding of this
class of genetic effect. In this section, we summarize strategies that can aid in such
research, emphasizing linkage mapping (Lander & Botstein, 1989) and selective
genotyping (Michelmore et al., 1991) approaches in strain crosses. We note that HGIs
that segregate among isolates can appear as additive effects or gene-gene interactions, or
may not even show effects depending on which individuals are crossed (Fig. A.3). We
also do not address association mapping in populations of wild isolates (e.g., (Andersen
et al., 2012; Atwell et al., 2010)) because the abundance of low frequency alleles (D.L.
Hartl, 2006; Pritchard, Stephens, Rosenberg, & Donnelly, 2000) and population
struct(Kang et al., 2010) in these studies complicates the already challenging task of
studying HGIs.
109
Fig. A.3 | Frequency of HGI progeny depends on parental genotypes. HGIs that
occur within species can segregate in a variety of ways in controlled crosses,
depending on the genotypes of cross parents at involved loci. Here, we show this
phenomenon using haploids and a three-locus HGI. Each individual is illustrated as
an unfilled bar with three loci. The order of loci is arbitrary and is not intended to
imply linkage.
Indeed, a number of factors make it difficult to identify the specific loci and genes
involved in HGIs in crosses. First, exhaustive searches for HGIs require an extremely
large number of statistical tests, resulting in a multiple testing problem and low power
(Cordell, 2009; Phillips, 2008; Storey, Akey, & Kruglyak, 2005). Second, as noted
previously with respect to gene-gene interactions (Turner, 2014), scans for HGIs
involving larger numbers of loci suffer from an added statistical power issue: multi-locus
genotype classes that are critical for detecting a HGI may be represented by only a very
110
small number of individuals in a typically sized mapping population and consequently the
effect of a HGI might be missed. Third, HGIs may be ‘synthetic’ or ‘compositional’; in
other words, loci involved in a HGI may not individually exhibit phenotypic effects (e.g.,
(Taylor & Ehrenreich)). Fourth, efforts to clone genes involved in HGIs using
complementation (Turner, 2014) and genetic engineering techniques (Storici et al., 2001;
Wilkinson & Wiedenheft, 2014) may only succeed in specific genetic backgrounds that
possess the requisite alleles at interacting loci (Service, 2004) (Fig. A.4).
111
Fig. A.4 | HGIs can cause differences in the phenotypic variance of the two alleles at
a locus. (A) Two parents that differ at three interacting loci are crossed. Only one of
the eight recombinant genotypes possesses the alleles required for the HGI (denoted
with a red arrow). (B) Phenotypes of the eight possible recombinants are plotted by
their allele at locus 1. The individual carrying the interacting combination of alleles
denoted with a red arrow and has a higher phenotypic value than the other
genotypes. (C) Potential effect of the HGI on the phenotypic variance associated
with the two alleles at locus 1. The orange allele shows an appreciably higher
phenotypic variance than the blue allele due to some individuals possessing the
interacting combination of alleles. The bars depict confidence intervals of an
arbitrary level (e.g., 95%).
112
Despite these challenges, it is possible to identify HGIs and dissect them to their
underlying components. One way to determine that a HGI might contribute to a trait is by
examining the phenotypic distribution shown by a mapping population. As others have
noted, genetic interactions that affect a continuous phenotype may shift the average trait
value among cross progeny relative to the average trait value shown by their parents
(Brem et al., 2005; Lynch M, 1998). However, assessing whether this effect is due to
HGIs, gene-gene interactions, or a combination of the two typically requires downstream
experiments. Further, HGIs that play a role in qualitative traits may cause these
phenotypes to segregate at a frequency inversely proportional to the number of involved
loci, as was recently shown (Taylor & Ehrenreich).
At the genetic level, a number of strategies have been developed to enable detection of
interacting loci by linkage analysis without performing exhaustive searches (e.g., (Brem
et al., 2005; Carlborg, Andersson, & Kinghorn, 2000; Kao, Zeng, & Teasdale, 1999;
Pare, Cook, Ridker, & Chasman, 2010; Struchalin, Dehghan, Witteman, van Duijn, &
Aulchenko, 2010)). Extensions of these approaches might be applicable to HGIs, in
particular when large genetic mapping populations are examined. A particularly
promising technique is analysis of ‘variance heterogeneity’ (Nelson et al., 2013; Pare et
al., 2010; Struchalin et al., 2010). The idea underlying this method is that individuals that
share the same allele at a locus and exhibit high phenotypic variance may do so because
they differ in their genotypes at other interacting loci (Fig. A.5). Evidence suggests this
strategy can identify loci involved in HGIs (Nelson et al., 2013); however, it is important
to rule out that such variance heterogeneity is not simply due to alleles that cause
increased stochasticity in a trait (e.g., (Ansel et al., 2008; Jimenez-Gomez, Corwin,
113
Joseph, Maloof, & Kliebenstein, 2011)). Beyond more conventional linkage approaches,
genetic mapping techniques that employ phenotypic selection (e.g., (Earley & Jones,
2011; Ehrenreich et al., 2010; Michelmore et al., 1991)) will also likely prove valuable in
studying HGIs. Unlike linkage mapping approaches, these methods identify alleles that
are enriched among or retained by individuals that share a phenotype, and can detect both
additive and interacting loci.
Fig. A.5 | Cloning genes underlying HGIs requires using appropriate genetic
backgrounds for engineering or complementation testing. Here, we focus on allele
replacement strategies. The top panel (black border) shows examples that will fail to
validate a gene’s effect, whereas the bottom panel (red border) shows examples that
will succeed in validating a gene’s effect. In the plots illustrating phenotypic effects,
‘s’ refers to the starting genotype, while ‘e’ refers to the engineered genotype.
114
The work described in the previous section (Chandler et al., 2014; Hanlon et al., 2006;
Pettersson et al., 2011; Taylor & Ehrenreich) also provides valuable insights into
strategies for effectively studying HGIs. For example, identifying HGIs may be easiest in
crosses of two strains (e.g., (Chandler et al., 2014; Dowell et al., 2010; Pettersson et al.,
2011; Taylor & Ehrenreich)). This is likely because they possess less genetic diversity
and fewer allele combinations than natural populations or intercrosses involving a larger
number of parents (e.g., (Aylor et al., 2011; King et al., 2012; Kover et al., 2009)).
Additionally, the previous work shows that examining a very large number of cross
progeny can provide the statistical power necessary to detect loci involved in HGIs
(Pettersson et al., 2011; Taylor & Ehrenreich). A final noteworthy result of some of the
above studies is that mutations and ‘cryptic’ genetic variants that segregate in natural
populations can participate together in HGIs (Chandler et al., 2014; Taylor &
Ehrenreich). Cryptic variants normally do not have phenotypic effects, but can influence
traits when perturbed either by their genetic background or environment (Gibson, 2009;
Gibson & Dworkin, 2004; Paaby & Rockman, 2014). This finding shows that induced
and spontaneous mutations can aid in the search for HGIs.
A.5 Mechanistic basis of HGIs
To this point, we have mainly explored the evidence for HGIs, as well as techniques that
can aid in finding loci involved in HGIs. However, determining the mechanisms that
cause HGIs is also important. Such knowledge will improve our basic understanding of
how an individual’s genotype specifies their phenotype and might also facilitate the
development of new methods for detecting HGIs.
115
Even though present data on mechanisms underlying HGIs is limited, past work on gene-
gene interactions (as reviewed in (Boone, Bussey, & Andrews, 2007; Lehner, 2011)) is
likely relevant. Gene-gene interactions can occur due to loss-of-function mutations in two
genes that act in the same or redundant pathways or protein complexes (Costanzo et al.,
2010; Tong et al., 2001). Given that putatively null alleles and incomplete loss-of-
function alleles are fairly common in natural populations (Abecasis et al., 2012; Clark et
al., 2007; Liti et al., 2009; Mackay et al., 2012), it is possible that sets of three or more of
these variants might lead to HGIs if they occur in the ‘right’ genes. Further, hundreds of
protein complexes possess more than two protein subunits (Gavin et al., 2006). It is
possible that genetic variants in three or more of these subunits might collectively
destabilize a protein complex or even induce new protein complexes to form.
There are a number of other potential causes of HGIs, one of the most likely being
genetic variation perturbing multiple components of a gene regulatory network
(Gjuvsland et al., 2007; Nuzhdin, Friesen, & McIntyre, 2012). If sets of transcription
factors and their upstream regulators possess functional variants, specific combinations of
these alleles might collectively alter the levels of phenotypically important transcripts and
proteins. Supporting such a possibility, it is well established that species harbor extensive
heritable variation in gene regulation (e.g., (Brem, Yvert, Clinton, & Kruglyak, 2002;
McVicker et al., 2013; Pickrell et al., 2010; Wray et al., 2003; Zheng, Zhao, Mancera,
Steinmetz, & Snyder, 2010)). Further, recent work in yeast has shown that
polymorphisms in transcription factors, as well as a polymorphism that might impact
translation, contribute to HGIs (Gerke et al., 2009; Taylor & Ehrenreich). Other
mechanisms that have been hypothesized for gene-gene interactions may also extend to
116
HGIs. These include changes in the levels of critical ‘threshold’ molecules and loss of
genetic buffering (Hartman, Garvik, & Hartwell, 2001; Lehner, 2011). Finally, some
HGIs may be the manifestation of a “diminishing return” for combinations of additive
loci. If a trait has a finite maximum value, combinations of additive loci may appear to
have much lower impacts on the trait than they do in isolation.
Although we have focused on individual mechanisms that can result in HGIs, the
mechanisms described in this section are not mutually exclusive. Indeed, combinations of
these mechanisms, as well as other unforeseen causes, might result in HGIs.
Comprehensive genetic and molecular dissection of example HGIs is therefore critical to
advancing our understanding of this subject, as present data do not allow any definitive
claims.
A.6 Outstanding Questions
A.6.1 What is the typical architecture, effect size distribution, and prevalence of
HGIs? Much of the direct evidence for HGIs comes from examples involving a relatively
small number of loci that collectively exhibit a large effect. Additional research is needed
to determine whether HGIs often involve even more loci or make more subtle
contributions to traits, and to assess the general contribution of HGIs to the total
phenotypic variance within populations. Furthermore, as noted by others (Mackay),
alleles involved in gene–gene interactions can generate apparent additive genetic variance
in natural populations. It is worthwhile to examine the extent to which alleles that
participate in HGIs also contribute to this problem. Lastly, most work on HGIs to date
has focused on haploid organisms or homozygous inbred lines from diploid species.
117
Future work should explore the general prevalence of HGIs across traits and organisms,
and should also examine the importance of HGIs in populations that exhibit
heterozygosity.
A.6.2 Are there general mechanisms that result in HGIs? Identifying loci involved in
HGIs is difficult with current approaches and cloning the genes underlying these loci is
arguably even more challenging. However, characterizing how HGIs arise at the
molecular level is imperative. Such work may identify mechanisms that generate
variability in complex traits are distinct from those found in studies of additive loci or
gene–gene interactions. Additionally, discovering any general mechanisms that
contribute to HGIs may aid in the development of new analytical approaches for
detecting loci that participate in HGIs. Such methods may be necessary for identifying
HGIs in humans and other species that are not amenable to the experimental techniques
described in this review.
A.6.3 What is the relation between evolution and HGIs? Evolutionary geneticists have
begun to recognize the potential importance of HGIs (Weinreich, Lan, Wylie, &
Heckendorn). Within the context of complex traits, many evolutionary processes,
including genetic drift, natural selection, demography, and genome duplication, influence
the genetic architectures of heritable phenotypes (D.L. Hartl). Determining the extent to
which these different processes affect the prevalence of HGIs can help in assessing
whether HGIs might contribute to heritable phenotypic variation in a particular
population or species. Lastly, given that cryptic genetic variants can participate in HGIs,
118
it is worthwhile to examine the broader role of HGIs in facilitating phenotypic novelty
and evolutionary adaptation in natural populations ((Ed.); Hansen).
A.7 Concluding remarks
As we have shown, evidence suggests that HGIs contribute to a number of complex traits.
However, as a general class of genetic effects, HGIs are poorly understood. Thus, it is
likely that research in the near future will play a crucial role in developing our basic
knowledge of HGIs (see Outstanding Questions). If the main goal of genetics as a field is
to determine how an individual’s genotype specifies their phenotype, then research on
HGIs must advance. Although this line of enquiry is in its early days, it has great
potential to significantly improve our understanding of the genetic basis of heritable
traits.
119
Appendix B: The genotypic landscape of environmental robustness
This work is presented as submitted for publication in 2015
The author list is as follows:
Jonathan T Lee, Matthew B Taylor, Amy Shen, and Ian M Ehrenreich
B.1 Overview
Genetic variation can alter the environmental robustness of heritable phenotypes,
thereby causing individuals to show trait differences across conditions. Such variability in
robustness likely has implications for disease susceptibility and evolution, and thus is
important to understand at its genetic and molecular levels. Here, we address this topic by
comprehensively defining the genetic basis of thermal robustness in a cryptic yeast
colony phenotype. The trait is revealed in a yeast cross by a capacitating mutation
in IRA2, a negative regulator of the Ras pathway. We identify alleles of eight loci that
genetically interact in different combinations to determine the ability of IRA2 mutants to
express the phenotype across temperatures. While only a single predominant genotype
involving these variants confers robust expression of the trait in all the examined
environments, multiple genotypes exhibit low robustness and only show the phenotype
specifically at low temperature. Despite their phenotypic similarity across temperatures,
these non-robust genotypes vary significantly in their potential to achieve higher
robustness through segregating genetic variants in the cross. Thus, our study not only
provides a detailed portrait of how differences in environmental robustness shape the
expression of a model complex trait across genetic backgrounds and temperatures, but
also illustrates genetic constraints on the potential to evolve higher robustness.
120
B.2 Introduction
Robustness is the ability of a trait to persist in the face of perturbation (de Visser et al.,
2003; Felix & Barkoulas, 2015; Felix & Wagner, 2008; Lempe, Lachowiec, Sullivan, &
Queitsch, 2013; Levy & Siegal, 2012; Masel & Siegal, 2009; Nijhout, 2002). Phenotypes
can exhibit robustness to macro- or micro-environmental change (i.e., ‘environmental
robustness’), to mutations or standing genetic variants (i.e., ‘genetic robustness’), or to
both environmental and genetic effects. Although our understanding of the mechanisms
that give rise to robustness is likely incomplete (Siegal & Leu, 2014), robustness is
known to be conferred by specific features of biological systems, including chaperones
that buffer cells against the misfolding of other proteins (e.g. heat shock proteins),
functionally redundant genes and pathways, and aspects of gene regulatory network
architecture (Boone et al., 2007; Hartman et al., 2001; Jarosz & Lindquist, 2010; Levy &
Siegal, 2008; Rutherford, 2000; Stelling, Sauer, Szallasi, Doyle, & Doyle, 2004).
Because these features are encoded in an organism’s genome, genetic variants that alter
robustness can arise within populations and cause individuals to differ in their levels of
robustness (Felix & Wagner, 2008; Manolio et al., 2009; Queitsch, Carlson, & Girirajan,
2012).
Heritable variation in robustness may significantly contribute to disease susceptibility
(Manolio et al., 2009; Queitsch et al., 2012) and might also play important roles in
evolution (Bergman & Siegal, 2003; Ehrenreich & Pfennig; Felix & Wagner, 2008; Flatt;
Gibson & Wagner; Masel & Trotter, 2010; Pigliucci & Murren; Pigliucci, Murren, &
Schlichting; Siegal & Bergman, 2002; van Nimwegen, Crutchfield, & Huynen;
Waddington; Wagner, Booth, & Bagheri; Wilke & Adami, 2003). For these reasons, it is
121
important to determine the genetic and molecular basis of such differences in robustness
among individuals (Barkoulas, van Zon, Milloz, van Oudenaarden, & Felix, 2013;
Gibson). However, multiple challenges complicate such an endeavor. First, a trait must
be available that shows measurable differences in environmental robustness across
individuals. Once this trait has been identified, genetic mapping of variability in the
phenotype’s environmental robustness must be performed in a way that can detect a
broad range of genetic architectures, including those that involve genetic interactions
among more than two loci (i.e., ‘higher-order genetic interactions’ (Pettersson et al.,
2011; Taylor & Ehrenreich, 2014, 2015a, In press)). This is because ample theoretical
evidence suggests that changes in robustness likely involve higher-order genetic
interactions or interactions between sets of loci and the environment (Bagheri,
Hermisson, Vaisnys, & Wagner, 2003; Hansen, Alvarez-Castro, Carter, Hermisson, &
Wagner, 2006; Hermisson, Hansen, & Wagner, 2003; Hermisson & Wagner, 2004;
Wagner et al.). Additionally, genetic mapping studies focused on robustness must also be
conducted in a way that can identify the contributions of loci that stabilize a phenotype’s
expression across conditions, regardless of whether these loci are required for expression
of the trait itself (Masel & Siegal, 2009).
We recently described a powerful model system in the budding yeast Saccharomyces
cerevisiae that can be used to study the genetic basis of differences in environmental
robustness. In two previous papers (Taylor & Ehrenreich, 2014, In press), we
demonstrated that a spontaneous frameshift mutation in IRA2, a negative regulator of the
Ras-cAMP-PKA (Ras) pathway (Tanaka, Nakafuku, Satoh, et al., 1990; Tanaka,
Nakafuku, Tamanoi, et al., 1990), can reveal a ‘rough’ colony phenotype in a cross of the
122
BY4716 (‘BY’) lab strain and a derivative of the 322134S (‘3S’) clinical isolate at
standard laboratory temperature (30°C). This mutation (ira2∆2933) results in a truncation
of the cognate Ira2 protein by 117 amino acids and causes a partial loss of Ira2’s function
(Taylor & Ehrenreich, 2014). At 30°C, ira2∆2933 segregants from the BYx3S cross
exhibit the rough phenotype if they have one of two combinations of cryptic genetic
variants at four or more genes. Specifically, cross progeny with the genotypes END3
BY
FLO8
3S
ira2∆2933 MSS11
BY
TRR1
3S
or END3
3S
FLO8
3S
ira2∆2933 MGA1
BY
MSS11
BY
SFL1
BY
express rough morphology (Taylor & Ehrenreich, 2014, In press).
The majority of the interacting genes involved in the expression of rough morphology at
30°C—FLO8, MGA1, MSS11, and SFL1—are transcription factors that act downstream
of the Ras pathway (Conlan & Tzamarias, 2001; Fujita et al., 1989; Gagiano et al.;
Kobayashi et al.; Lorenz & Heitman, 1998; Pan & Heitman, 2002; Robertson & Fink,
1998). Flo8 and Mss11 form a heterodimeric transcriptional activator (H. Y. Kim, Lee,
Kang, Oh, & Kim, 2014), and Mga1 also likely acts as an activator (Borneman et al.;
Lorenz & Heitman, 1998). In contrast, Sfl1 represses Ras target genes when signaling
through the pathway is low (Pan & Heitman, 2002). Based on our past work, it appears
that the role of ira2∆2933 in rough morphology is to induce transcriptional derepression
of Ras target genes, thereby allowing genetic variants that influence the trait to exert their
phenotypic effects (Taylor & Ehrenreich). As for END3 and TRR1, these genes are
involved in vesicle formation (Benedetti et al., 1994; Tang et al.) and detoxification of
endogenous redox stress (Pedrajas et al.; Trotter & Grant, 2002), respectively, and their
roles in the phenotype are less clear. However, we note that these other cellular processes
123
are known to have functional relationships with Ras signaling (Charizanis et al., 1999;
Gourlay & Ayscough, 2006).
Here, we assess how BYx3S ira2∆2933 progeny express rough morphology across a
range of temperatures and find substantial variability in thermal robustness. By using
next generation sequencing to conduct bulk segregant mapping (Ehrenreich et al., 2010;
Michelmore et al.; Wenger, Schwartz, & Sherlock) and selective genotyping (Matsui et
al., 2015; Taylor & Ehrenreich, 2014) of individuals with different levels of
environmental robustness (Fig. B.1), we identify eight loci (the genes END3, FLO8,
FLO11, MGA1, MSS11, SFL1, and TRR1, as well as an undetermined variant on
Chromosome XII) that influence the expression of ira2∆2933-dependent rough
morphology across temperatures. FLO11, which was not identified in our past work
(Taylor & Ehrenreich, 2014, In press), encodes a cell surface glycoprotein that is required
for rough morphology (Bruckner & Mosch; Gerke et al.; Taylor & Ehrenreich, In press).
124
Fig. B.1 | Individuals can express the same phenotype due to distinct genetic
architectures that differ in environmental robustness. In this figure, multiple
genotypes specify the same trait in certain environments at 21°C, but only some of
these genotypes enable expression of the trait at higher temperatures.
In this paper, we examine how combinations of alleles across the eight loci collectively
determine a given genotype’s thermal robustness. We find that only a single genotype
involving these loci results in robust expression of the rough colony phenotype across
temperatures. In contrast to this small genotypic space that confers thermal robustness,
many combinations of alleles facilitate expression of rough morphology specifically at
low temperature. Although these non-robust individuals are phenotypically
125
indistinguishable across temperatures, they dramatically differ in their potential to
achieve higher robustness through segregating variants that are present in the cross. Thus,
our results provide a comprehensive characterization of the relationship between
genotype and environmental robustness for a model phenotype, and also reveal how the
potential for complex traits to evolve robustness is constrained at the genetic level.
B.3 Generation of a mapping population for dissecting thermal robustness
We mated the rough ira2∆2933 segregant used for genetic mapping in (Taylor &
Ehrenreich, 2014) to its BY and 3S parents. Based on preliminary experiments, we found
that many more BYx3S ira2∆2933 segregants show rough morphology at low
temperature (21°C) than at higher temperatures (30 and 37°C; Table B.1), and that
individuals that express rough morphology at low temperature vary in their abilities to
show rough morphology at higher temperatures (Table B.2). For this reason, we focused
on isolating rough backcross segregants at 21°C.
Collection
Temperature (°C)
Number of
Rough Segregants
21 225
30 139
37 85
Table B.1 | Rough morphology is more common at low temperature. Spores from
the BY backcross were plated and scored for rough colonies at each collection
temperature.
126
Temperature (°C)
Collected
Sequenced
21 30 37 BY 3S BY 3S
+ − − 130 202 82 131
+ + − 83 135 55 101
+ − + 17 3 0 0
+ + + 65 53 51 51
Table B.2 | Rough segregants at 21° show various phenotypes at higher
temperatures. Individuals collected at 21°C expressed (+) or did not express (−) the
phenotype at other conditions. A subset of these were sequenced.
Through a screen of ~7500 individuals (Methods), we identified 295 and 393
individuals that expressed rough morphology in at least one temperature from the BY and
3S backcrosses, respectively (Table B.2). These individuals fell into four classes: rough
at 21°C, rough at 21 and 30°C, rough at 21 and 37°C, and rough at 21, 30, and 37°C
(Table B.2). Individuals that were rough at 21 and 37°C represented only 2.9% of the
688 total segregants. Thus, we focused on the other three phenotypic classes, which each
occurred at a frequency of at least 17.2% in the total population of rough segregants. We
refer to these phenotypic classes as ‘fragile’ (rough at 21°C), ‘moderate’ (rough at 21 and
30°C), and ‘robust’ (rough at 21, 30, and 37°C) throughout the remainder of the paper in
reference to their different levels of thermal robustness (Fig. B.2).
127
Fig. B.2 | Ira2∆2933 segregants vary in their ability to express the rough phenotype
at different temperatures. Although the BY and 3S strains form smooth colonies,
BYx3S segregants that carry the ira2∆2933 lesion can show a rough phenotype.
These rough segregants can express the trait across different temperature ranges.
Some individuals can express the phenotype only at 21°C, others can express the
phenotype only at 21°C and 30°C, and others yet can express the phenotype at all
examined temperatures. We refer to these classes of rough segregants as fragile,
moderate, and robust, respectively.
128
B.4 Bulk segregant mapping of different thermal robustness phenotypes
To determine the genetic basis of the fragile, moderate, and robust phenotypes,
we performed bulk segregant mapping by sequencing (Ehrenreich et al., 2010;
Michelmore et al.; Wenger et al.). Between 51 and 126 individuals were selected per
class from each backcross (Table B.2). These individuals were pooled and sequenced to
an average of 292.43-fold coverage. Across the six pools, seven loci were detected in
different combinations (Fig. B.3A-C). Most of these loci overlapped ira2∆2933 or alleles
of genes that we previously cloned: END3
BY
, FLO8
3S
, MGA1
BY
, MSS11
BY
, and TRR1
3S
.
The final locus was detected on Chromosome IX and was seen specifically in the robust
class.
129
Fig. B.3 | Bulk segregant mapping results for different robustness classes. Pooled
sequencing revealed loci associated with the A) fragile, B) moderate, and C) robust
phenotypes. Alleles inherited from the BY strain are denoted in blue, while those
inherited from 3S are denoted in orange. Selectable markers on Chromosomes III
and V are colored gray. The BY and 3S backcross populations for each robustness
class are shown in the top and bottom of each figure panel, respectively.
130
The only allele that showed complete fixation among all three phenotypic classes was
FLO8
3S
. One additional locus corresponding to ira2∆2933 was detected in all phenotypic
classes, and although ira2∆2933 was fixed among moderate and robust individuals, it
segregated among fragile individuals. Specifically, ira2∆2933 was highly enriched
(86.1% frequency) but not completely fixed among fragile individuals in the BY
backcross (Fig. B.3A). Targeted genotyping of ira2∆2933 revealed that 15 fragile
individuals did in fact carry IRA2
BY
instead of the mutation (Methods). However, the
small number of these individuals prevented effective genetic analysis on this subset. For
this reason, we focused our analysis of the fragile class on individuals carrying
ira2∆2933, which we elaborate on later in the paper. However, we note that the presence
of IRA2
BY
individuals may reflect the existence of other genetic architectures that occur
at 21°C and do not require a capacitating mutation.
While no other loci were identified among fragile individuals, moderate individuals
always possessed FLO8
3S
, ira2∆2933, and MSS11
BY
, and almost always carried END3
BY
and TRR1
3S
(Fig. B.3B). The incomplete fixation of these latter alleles occurred because
two genotypes that we previously described—END3
BY
FLO8
3S
ira2∆2933 MSS11
BY
TRR1
3S
and END3
3S
FLO8
3S
ira2∆2933 MGA1
BY
MSS11
BY
SFL1
BY
—confer moderate
robustness (Fig. B.4). Of the two, the END3
BY
-dependent genotype that is predominantly
seen in this study is expected to occur more frequently because it is less genetically
complex. Because the END3
BY
- and END3
3S
-dependent genotypes require five and six
alleles, respectively, the latter genotype is expected to only occur half as often. In
practice, the bias towards the END3
BY
-dependent genotype is sometimes even higher, as
a locus that confers a selective advantage during random spore isolation in the BYx3S
131
cross is closely linked to END3, with the BY allele of this locus conferring a benefit
(Taylor & Ehrenreich, 2014, In press).
Fig. B.4 | A previously identified higher-order genetic interaction leads to a
moderate phenotype. A strain carrying the genotype END3
3S
FLO8
3S
ira2∆2933
MGA1
BY
MSS11
BY
SFL1
BY
results in rough morphology at 21 and 30°C, but not at
37°C.
Lastly, robust individuals had the genotype END3
BY
FLO8
3S
ira2∆2933 MSS11
BY
TRR1
3S
,
but were also fixed for specific alleles at two additional loci—one corresponding to
MGA1
BY
, and a 3S allele on Chromosome IX (Fig. B.3C). To verify the role of MGA1
BY
in the robust phenotype, we used allele replacement in a segregant that possessed the
appropriate genotype and exhibited the robust phenotype (Fig. B.5B). Furthermore, based
on allele replacement in the same strain, we resolved the Chromosome IX locus to the
coding region of FLO11, which, as described earlier, encodes a cell surface glycoprotein
132
that must be actively transcribed for an individual to express the rough phenotype
(Bruckner & Mosch; Gerke et al.; Taylor & Ehrenreich, In press) (Fig. B.5A). In addition
to the seven fixed alleles, we also observed that a second locus on Chromosome XV
showed strong enrichment (~90%) for the BY allele among robust individuals. This site
overlaps SFL1
BY
, which can contribute to moderate robustness (see above), suggesting
that SFL1
BY
might have a quantitative effect on robustness in an END3
BY
FLO8
3S
FLO11
3S
ira2∆2933 MGA1
BY
MSS11
BY
TRR1
3S
genetic background.
Fig. B.5 | FLO11
3S
and MGA1
BY
confer robustness. Allele replacement in a robust
strain was used to confirm the effects of A) MGA1
BY
and B) FLO11
3S
specifically at
37°C.
Our results indicate that just one of the two genotypes that confer moderate robustness—
END3
BY
FLO8
3S
ira2∆2933 MSS11
BY
TRR1
3S
—has the potential to give rise to higher
thermal robustness. Moreover, it suggests that a single predominant genotype in the
133
BYx3S cross confers thermal robustness across all of the examined temperatures in the
presence of ira2∆2933.
B.5 Differences in genetic architectures of fragility in the BY backcross depend on
MSS11
We attempted to determine whether the low genetic complexity detected among fragile
individuals reflects a simple genetic architecture for fragility or the masking of loci due to
the presence of distinct genotypes conferring the same phenotype (i.e., ‘genetic
heterogeneity’, as shown in (Matsui et al., 2015) and (Taylor & Ehrenreich, 2014)). To
investigate these possibilities, we individually sequenced each fragile backcross
segregant at low coverage and looked within each backcross population for pairwise
genotype ratio distortion (Corbett-Detig, Zhou, Clark, Hartl, & Ayroles, 2013), which can
indicate two-locus genetic interactions (Methods).
Using a 1% false discovery rate threshold, we detected two pairs of putatively interacting
loci in the 3S backcross (Fig. B.6A and B; Methods). A locus on Chromosome XIII
corresponding to MSS11 was involved in both of the interactions; the partner loci in these
interactions were on Chromosomes XII and XV. Based on allele replacement
experiments performed in fragile segregants, we determined that SFL1, a gene that we
described earlier, was causal at the Chromosome XV locus (Methods; Fig. B.7). We
deleted the strongest candidate genes underlying the Chromosome XII locus using
genetic engineering (Methods), but did not observe a change in phenotype in any of
these deletion strains. This could occur if the causal variant is a loss-of-function allele,
where deletion of the causal gene would have no phenotypic impact. As we have yet to
134
identify the causal gene at this locus, we refer to simply by its chromosome number ‘XII’
for the remainder of the paper.
135
Fig. B.6 | Multiple sets of genetic interactions cause the fragile phenotype. Genetic
interactions were observed between MSS11
3S
and SFL1
3S
(A) and between MSS11
3S
and XII
BY
(B) in fragile individuals from the 3S backcross. C) MSS11
3S
, SFL1
3S
, and
XII
BY
exhibit a three-way genetic interaction. D) Individuals with MSS11
BY
require
no other loci for fragile trait expression. E) Individuals with MSS11
3S
show fragile
trait expression due to higher-order genetic interaction. F) Genetic mapping reveals
that fragile individuals with the ira2∆2933 mutation require FLO8
3S
and are highly
enriched for FLO11
3S
. For D through E, we note that selectable markers are
denoted in gray, alleles from BY are shown in blue, and alleles from 3S are
indicated with orange.
136
Fig. B.7 | Multiple alleles contribute to rough morphology at 21°C. Replacement of
END3
BY
, MGA1
BY
, MSS11
3S
, or SFL1
BY
with the opposite allele causes a loss of the
phenotype in a fragile MSS11
3S
segregant.
The two pairwise genetic interactions that we identified could represent either negative
genetic interactions that suppress the rough phenotype or synergistic genetic interactions
that potentiate the rough phenotype. To differentiate between these possibilities, we
examined each pair of interacting genes for enriched or depleted genotypes. For both
locus pairs (Fig. B.6A and B), a single two-locus genotype class was depleted, indicating
that the detected interactions reflect negative epistasis. Specifically, we observed that
MSS11
3S
negatively interacts with both SFL1
3S
and XII
BY
. Additionally, when only rough
individuals carrying MSS11
3S
are considered, 79% of them have the genotype SFL1
BY
XII
3S
, suggesting that the two pairwise genetic interactions might in fact represent the
same higher-order genetic interaction (Fig. B.6C).
To more deeply investigate possible higher-order genetic interactions involving MSS11,
we split the fragile individuals from the BY backcross into two groups based on their
137
MSS11 genotypes. We then separately examined genome-wide allele frequencies in each
MSS11 genotype class and found that the fragile individuals in the 3S backcross are
genetically heterogenous. No additional loci were detected among MSS11
BY
individuals
(Fig. B.6D); however, MSS11
3S
individuals largely had the genotype XII
3S
END3
BY
MGA1
BY
MSS11
3S
SFL1
BY
(Fig. B.6E). This suggests that MSS11 allele state plays a
major role in determining which other alleles are required for expression of rough
morphology at low temperature, as the presence of MSS11
BY
is sufficient for the
phenotype in an ira2∆2933 FLO8
3S
background, whereas MSS11
3S
is not. Instead, we
find that MSS11
3S
is involved in a higher-order genetic interaction that enables it to
express the phenotype at low temperature. This finding also indicates that the negative
pairwise genetic interactions described above were detected because allele substitutions
at XII
,
END3, MGA1, MSS11, and SFL1 disrupt a synergistic, higher-order genetic
interaction that confers fragile trait expression to individuals that carry MSS11
3S
.
Additionally, we attempted to identify loci that contribute to fragility in the BY backcross
population by looking at allele frequency skew, focusing specifically on individuals that
carried ira2∆2933. The only additional locus that showed significant enrichment among
these individuals was on Chromosome IX and corresponded to FLO11
3S
(Fig. B.6F). As
FLO11
3S
is also required for the robust phenotype class (see above), it would appear that
this allele has a stabilizing effect on the phenotype in multiple environments and genetic
backgrounds. We attempted to identify additional loci in this population both by
subsetting the data by FLO11 allele and also by testing for genotype ratio distortion
(Methods). However, neither of these approaches were successful, implying that either
FLO11 has a quantitative effect on rough morphology or additional FLO11-dependent
138
heterogeneity exists that we have not detected in the current data. We note that a
quantitative effect would be consistent with our results showing that XII
3S
, END3
BY
,
MGA1
BY
, and SFL1
BY
are nearly but not completely fixed in the MSS11
3S
fragile
background that we observed in the other backcross.
B.6 Discussion
In this paper, we have comprehensively characterized the genetic basis of heritable
variation in thermal robustness in a model phenotype. To do this, we examined a colony
morphology trait that was revealed by a spontaneous IRA2 mutation in the BYx3S cross
and shows variability in thermal robustness among BYx3S segregants. Across the current
manuscript and our previous work (Taylor & Ehrenreich, 2014, In press), we have now
described eight genomic loci that collectively determine whether the rough phenotype is
expressed in the presence of ira2∆2933 at a given temperature (Fig. B.8A).
Combinations of these variants appear to largely act in an epistatic manner to influence
whether an individual can express rough colony morphology and, if so, at what level of
thermal robustness. Thus, regarding the genetic architecture of environmental robustness,
our results are consistent with theoretical predictions that genetic interactions play an
important role in determining the level of robustness associated with a given genotype
(Bagheri et al., 2003; Hansen et al., 2006; Hermisson et al., 2003; Hermisson & Wagner,
2004; Wagner et al.).
139
Fig. B.8 | Higher-order genetic interactions cause varying levels of robustness. A)
Different configurations of eight loci result in various levels of phenotypic
robustness. B) Genetic interactions of increasing complexity lead to different levels
of thermal robustness. Only specific genotypes (black arrows) have the capacity to
lead to greater robustness. In both A and B, blue and orange refer to BY and 3S
alleles, respectively.
Our results also demonstrate that the potential for a trait to evolve robustness through
segregating genetic variation is constrained by the phenotype’s genetic architecture.
140
Specifically, many allele configurations that we have identified can enable expression of
the rough phenotype at low temperature in an ira2∆2933 background. Yet most of these
genotypes result in a trait that is perturbed by increases in temperature, and only a select
few have the potential to evolve higher thermal robustness through genetic variants that
are present in the cross. For example, individuals that carry MSS11
3S
are constrained to
express the rough phenotype only at low temperature, whereas MSS11
BY
individuals have
the potential to show moderate or complete robustness to temperature, depending on their
allele configuration at up to six other genes (Fig. B.8B). A second case is observed at the
transition from moderate to full robustness. END3
3S
relegates individuals to the moderate
phenotype, while END3
BY
provides the potential to obtain higher robustness if FLO11
3S
,
MGA1
BY
, and SFL1
BY
are also present. Moving forward, we plan to characterize the
functional differences between the alleles of MSS11 and END3 in order to understand
how they potentiate increases in thermal robustness.
While certain alleles in our study appear to potentiate robustness, others act as phenotypic
stabilizers (Masel & Trotter). Of note, MGA1
BY
and SFL1
BY
seem to consistently enable
different genotypes to maximize their potential robustness. At this point, the molecular
function of the Mga1 transcription factor remains poorly understood (Borneman et al.;
Lorenz & Heitman, 1998), making it difficult to speculate on this allele’s role in
phenotypic stabilization. However, as we previously described, loss of SFL1 function can
reveal the rough phenotype in a similar manner to ira2∆2933 (Taylor & Ehrenreich)This
is because Sfl1 negatively regulates the transcription of Ras target genes by recruiting the
Ssn6-Tup1 co-repressor complex (Conlan & Tzamarias), which in turn recruits histone
deacetylases that silence adjacent genes, such as FLO11 (Halme et al.; Taylor &
141
Ehrenreich, In press; Wu et al.) (Taylor & Ehrenreich). Given that diminished Sfl1
activity can facilitate expression of rough morphology, it may be that the MGA1
BY
SFL1
BY
allele combination acts as a phenotypic stabilizer by reducing Sfl1 activity.
In general, differences in thermal robustness and potential to evolve higher thermal
robustness may relate to variability in the gene regulatory network that specifies rough
morphology across genetic backgrounds. Indeed, theory suggests that gene regulatory
network architecture can have a major impact on a trait’s robustness to environmental
perturbation (Bergman & Siegal; Siegal & Leu). In our system, half of the identified
genes that influence rough colony morphology in the presence of ira2∆2933 are
transcription factors: FLO8, MGA1, MSS11, and SFL1. Furthermore, as we have shown
in this paper, genetic variants in three of these transcription factors—MGA1, MSS11, and
SFL1—directly influence thermal robustness. The major role of these transcription
factors in our study is consistent with the idea that changes in the gene regulatory
network underlying rough morphology across genotypes contribute to an individual’s
robustness and potential to achieve even higher robustness (Pfennig & Ehrenreich, 2014).
Our results also inform understanding on phenotypic capacitance, the phenomenon in
which unexpected phenotypic variation is revealed by genetic or environmental
perturbation (Gibson & Dworkin; Paaby & Rockman; Paaby et al.; Queitsch et al.;
Rutherford). As we previously described, all of the polymorphisms that influence rough
morphology in the BYx3S cross can be considered cryptic genetic variants (Taylor &
Ehrenreich). This is because BYx3S segregants with the appropriate genotypes at
standing polymorphisms in the cross do not express rough morphology unless ira2∆2933
142
(Taylor & Ehrenreich, 2014), sfl1∆ (Taylor & Ehrenreich)or potentially other mutations
are present. However, here we have found suggestive evidence that such capacitating
mutations are not always necessary, as some individuals can express rough morphology
at only low temperature without ira2∆2933. Thus, low temperature itself might uncover
cryptic variants that cause rough morphology. Although we were unable to define these
ira2∆2933-independent rough genotypes with the current data, the finding of
environmental capacitance raises an interesting question about how low temperature can
act as a capacitor in a manner similar to IRA2 and SFL1. Given our past finding that
mutations in these genes leads to transcriptional derepression of Ras targets, it is possible
that low temperature can have the same molecular effect.
In summary, our study has provided detailed insights into the genetic basis of thermal
robustness in a model complex trait and has also shed light on how the potential for a
phenotype to evolve higher robustness is constrained at the genetic level. Both the rough
phenotype’s expression and its thermal robustness appear to be influenced by genetically
complex changes in Ras signaling and gene regulation. Given that temperature change is
one of the most ubiquitous types of environmental perturbation that organisms experience
in nature and that many components of the Ras pathway are conserved across eukaryotes,
our findings are likely relevant to other species and traits.
143
B.7 Materials and Methods
B.7.1 Generating Backcross Segregants. The lab strain BY4716 (BY) and a haploid
derivative of the clinical isolate 322134S (3S) were used as the original parents. These
strains were backcrossed to the BYx3S rough segregant used for genetic mapping in
(Taylor & Ehrenreich, 2014), which carries ira2∆2933. Diploids were sporulated at 21°C
using standard yeast sporulation methods (Sherman, 1991). Spore cultures were digested
with β-glucoronidase and spores selected using the Synthetic Genetic Array marker
system (Tong et al., 2001) as described previously (Ehrenreich et al., 2010; Taylor &
Ehrenreich, 2014). Spores were plated at low density (<100 spores per plate) on yeast
nitrogen base (YNB) plates containing canavanine, in order to select for MATa haploid
progeny.
B.7.2 Phenotyping. Yeast were grown on media containing yeast extract and peptone
(YP) with either 2% dextrose (YPD) or 2% ethanol (YPE) as a carbon source. Colonies
were replica plated from selective YNB + canavanine media onto YPE. Following five
days of growth at 21°C, colonies were screened as rough or smooth. Strains identified as
rough were inoculated in liquid YPD. Selected strains were grown in liquid YPD for two
days to stationary phase, and then manually pinned to YPE agar. Triplicate pinnings were
incubated at 21, 30, and 37°C for five days, and then imaged using a standard digital
camera.
B.7.3 Bulk segregant mapping across levels of robustness. Segregants were grown to
stationary phase as individual cultures. Cells from cultures of like phenotypes were
mixed in equal volume fractions and DNA extracted using the Qiagen DNeasy Blood and
144
Tissue Kit. Libraries of pooled DNA were constructed using the Illumina Nextera kit,
with each library barcoded with a distinct sequence tag. Libraries were mixed in
equimolar fractions and sequenced on an Illumina NextSeq machine using 75x75 base
pair reads. BY4716 is a derivative of the lab strain S288c. Thus, sequencing reads were
mapped to the S. cerevisiae S288c reference genome and 322134S draft genome
(available through http://www.yeastgenome.org) using the Burrows-Wheeler Aligner
(BWA) version 7 with options mem -t 20 (Li & Durbin, 2009). A 292.43-fold average
coverage across the pools was obtained from these alignments. Genome-wide allele
frequencies were determined at 36,756 unique high confidence SNPs that were
previously identified by mapping Illumina sequencing reads for 3S to the S288c reference
genome (Taylor & Ehrenreich, 2014, In press) (Taylor & Ehrenreich).
We employed the MULTIPOOL method (Edwards & Gifford, 2012) to each segregating
genomic region to identify loci and estimate the location of causal genes . Significantly
enriched loci were identified as having a maximum LOD score > 6 spanning a minimum
of 20 kb, bounded by a 90% credible interval around this point of maximum significance.
To generate allele frequency plots, the data was smoothed by averaging allele frequency
over sliding windows of 25 SNPs (approximately 3-4 kb).
B.7.4 Identification of pairwise interacting loci in the fragile class. Segregants were
grown to stationary phase in liquid media, with DNA extracted from each individual
culture using the Qiagen DNeasy 96 Blood and Tissue Kit. Libraries were prepared for
sequencing using the Illumina Nextera kit, with each individual receiving a unique
sequence barcode. Sequencing was performed on a Illumina NextSeq machine using
145
150x150 base pair reads. As previously described (Taylor & Ehrenreich, 2014), a Hidden
Markov Model (HMM) was used to determine the haplotype of each segregant from the
aforementioned 36,756 SNPs between the BY and 3S genomes. The resulting HMM table
was then collapsed to unique segregating regions. There were 1,421 and 967 segregating
regions in the 3S and BY backcross populations, respectively. Using the base package of
the R statistical program, we calculated genotype ratio distortion (Corbett-Detig et al.,
2013) for each locus pair in a backcross. Statistically significant interactions were
determined using a Χ
2
test, with a false discovery threshold of less than 1% using the R
qvalue() package. Regions within 20,000 bases of the ends of chromosomes were
excluded due to potential mis-mapping of reads at telomeric repeats and gene duplicates.
Based on the results from the interaction screen, individuals were subsetted and allele
frequency plots were generated from the aforementioned HMM tables. Allele frequencies
were averaged over 25 SNP sliding windows, and loci contributing to the trait were
called as regions enriched at greater than 90% or depleted below 10%.
B.7.5 Genetic Engineering. Adapter-mediated allele replacement (Erdeniz et al., 1997;
Taylor & Ehrenreich, 2014) was used to alter the allele state at MGA1 and FLO11 in a
robust strain, as well as END3, MGA1, MSS11, and SFL1 in a MSS11
3S
fragile strain.
Using PCR, an amplicon of the gene of interest was tailed at the 3’ end with the 5’ end of
the kanMx cassette, and an amplicon of kanMx was tailed on the 3’ end with the region
immediately downstream of the gene (Matsui et al., 2015). The two PCR products were
co-transformed into the given strain, and plated on YPD agar containing G418 to screen
for successful integration.
146
The Chromosome XII interval was delimited to a 34,519 base region. Within this
interval, the candidate genes HAP1, HSP60, GSY2, LCB5, PDR8, SYM1, YLR257W, and
YPT6 were independently deleted by replacement with the CORE cassette (Storici et al.,
2001). Regions corresponding to 60 bases upstream and downstream of the target gene
were tailed to the CORE cassette using PCR. This product was transformed into cells
using the Lithium Acetate method (Gietz & Woods, 2002), and selection with G418 was
used to screen for integration of the cassette. The strain used was a 3S backcross
segregant carrying the MSS11
3S
-based fragile genotype.
Abstract (if available)
Abstract
Cryptic genetic variants, which typically do not show an impact on phenotype, can be revealed by environmental or genetic perturbation. This phenomenon, known as capacitance, may play a role in evolution and disease, as it allows for biological systems to accumulate variation free of selection that can later be revealed. Despite a recognition of its potential importance, very little is known about the genetic architectures or molecular mechanisms underlying capacitance. ❧ In this thesis, I describe the development of a system uniquely capable of providing insights into capacitance at the genetic and molecular levels. In a cross of the yeast isolates BY4716 and 322134S, both parents and the vast majority of progeny form smooth colonies. However, a cryptic rough morphology phenotype can be revealed by specific combinations of cryptic variants and capacitating mutations. The work described in chapters 2-4 demonstrates how this phenotype change can function as a reporter for capacitance, and thus allow for precise genetic dissection of this phenomenon. ❧ In chapter 2, I demonstrate how a mutation in a negative Ras regulator can participate in cryptic, higher-order genetic interactions far more complex than those that had previously been demonstrated in the literature. In chapter 3, I explore the role this mutation plays at the molecular level and find that transcriptional derepression can mediate phenotypic capacitance. In chapter 4, I dissect the genetic architecture of several additional instances of phenotypic capacitance, and show that many combinations of capacitors and cryptic variants in the Ras pathway can reveal the same phenotype.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Complex mechanisms of cryptic genetic variation
PDF
Exploring the genetic basis of quantitative traits
PDF
Genome-scale insights into the underlying genetics of background effects
PDF
Genetic and molecular insights into the genotype-phenotype relationship
PDF
Understanding the genetic architecture of complex traits
PDF
The complex genetic and molecular basis of oxidative stress tolerance
PDF
The evolution of gene regulatory networks
PDF
Studies in bivalve aquaculture: metallotoxicity, microbiome manipulations, and genomics & breeding programs with a focus on mutation rate
PDF
Evolutionary genomic analysis in heterogeneous populations of non-model and model organisms
PDF
Understanding the genetics, evolutionary history, and biomechanics of the mammalian penis bone
PDF
Genome sequencing and transcriptome analysis of the phenotypically plastic spadefoot toads
PDF
Genetic architecture underlying variation in different traits in the Pacific oyster Crassostrea gigas
PDF
Genetic diversity and bacterial death in the context of adaptive evolution
PDF
Understanding genetics of traits critical to the domestication of crops using Mixed Linear Models
PDF
Robustness and stochasticity in Drosophila development
PDF
Applications of next generation sequencing in sessile marine invertabrates
PDF
Genome-wide GATC methylation in long-term cultures of Escherichia coli K-12
PDF
Application of genetic association methods in mice to understand phenotypes with a complex etiology
PDF
Natural variation of Arabidopsis thaliana methylome and its impact on genome evolution
PDF
Developing genetic tools to assist in the domestication of giant kelp
Asset Metadata
Creator
Taylor, Matthew Bryce
(author)
Core Title
Genetic architectures of phenotypic capacitance
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Molecular Biology
Publication Date
03/09/2016
Defense Date
02/22/2016
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
capacitance,complex traits,cryptic variation,epistasis,genetic background effects,higher-order genetic interaction,OAI-PMH Harvest,yeast
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Ehrenreich, Ian M. (
committee chair
), Dean, Matthew D. (
committee member
), Hedgecock, Dennis (
committee member
), Nuzhdin, Sergey V. (
committee member
)
Creator Email
mb.taylor@outlook.com,mbtaylor@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-220904
Unique identifier
UC11278310
Identifier
etd-TaylorMatt-4139.pdf (filename),usctheses-c40-220904 (legacy record id)
Legacy Identifier
etd-TaylorMatt-4139.pdf
Dmrecord
220904
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Taylor, Matthew Bryce
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
capacitance
complex traits
cryptic variation
epistasis
genetic background effects
higher-order genetic interaction
yeast