/

University of Southern California Dissertations and Theses

/

Application of genetic association methods in mice to understand phenotypes with a complex etiology

(USC Thesis Other)

Application of genetic association methods in mice to understand phenotypes with a complex etiology

PDF

Transcript (if available)

Content APPLICATION OF GENETIC
ASSOCIATION METHODS IN MICE TO
UNDERSTAND PHENOTYPES WITH A
COMPLEX ETIOLOGY

By
Amanda L. Crow

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GENETIC, MOLECULAR, AND CELLULAR BIOLOGY)

August 2015

Copyright 2015 Amanda L. Crow
i

DEDICATION

For my Grandpa, William Burkett Mason, who encouraged and inspired.

ii

ACKNOWLEDGEMENTS

I first would like to extend my gratitude to my dissertation committee
members, Dr. Gregor Adams and Dr. Rick Friedman. Collaborating with you and
your labs has been a pleasure, and I valued the opportunity to learn from you a bit
more about your fields over the course of my study. It would have been a bit non‐
academic to include elsewhere in the thesis, but I’m very pleased that working with
the HMDP allowed me to broaden my horizons and learn about stem cell and inner
ear biology, which I might not have been otherwise exposed to. Thank you.
Of course, I can’t give enough thanks to my mentor, Dr. Hooman Allayee.
From the time I enrolled in Human Molecular Genetics onward, your passion and
enthusiasm for genetics has been contagious. If I was ever feeling like I had gotten
stuck in my progress, all I needed was to show you were I was and you always had a
new idea and the confidence in me to make something out of it. That confidence
meant so much to me, and I really, really appreciate it. Thank you again so much for
the opportunity to work with you and build a name for myself as “the mouse GWAS
person!”
I also want to thank the current and former members of the Allayee lab who
I’ve worked with: Susanna Vikman, Yang Amy Zhang, and especially Yesha Patel and
Jaana Hartiala. I love taking downtown happy hours by storm with you and
kvetching about everything! Jaana, you know (I hope you know) how crucial your
advice, help, and patience were. I can’t even imagine how I would have done any of
this without you.
To my family, especially Mom, Dad, Punk, and Grandma, you knew when I
needed you (always) and you were there. From the time I was just a fat‐cheeked
bundle of mischief and creative energy, you showed me how to be nothing less
than my best. To my friends, you know who you are, and you know what you did.
Let’s dance forever. To Dan, in the immortal words of Carly Rae Jepsen, “Before you
came into my life I missed you so bad.” Thank you for being a most excellent human
and an even better boyfriend.

iii

LIST OF TABLES

Chapter 1 Page

Table 1. Loci identified in GWAS for HSPCs in the HMDP. 51

Table 2. Genes Exhibiting cis eQTLs Across Multiple Tissues
at the Chromosome 5 Locus for LSKCD150‐CD48‐ Cells. 52

Supplemental Table 1. Correlations of Primitive Hematopoietic
Cell Frequency with Blood Cell Parameters 65

Supplemental Table 2. HMDP Mouse Strains Used in the
Present Study Ordered According to LSK Frequency 66

Chapter 2

Table 1. Loci Identified in HMDP for ABR Thresholds
at Various Tone Burst Frequencies. 104

Table 2. cis eQTL for GWAS Loci at Various Tone Burst
Frequencies. 105

Chapter 3

Table 1. GWA results for NIHL in the HMDP. 157

Table 2. Genes within NIHL 5 association peaks regulated by
local eQTL in the cochlea. 158

iv

LIST OF FIGURES

Chapter 1 Page

Figure 1 (A‐C). Variation in three HSPC populations in the HMDP. 53

Figure 2 (A‐C). Manhattan plots of GWAS results for HSPC
frequency in the HMDP. 54

Figure 3 (A‐D). Regional plots of loci significantly associated with
HSPCs in the HMDP. 57

Figure 4 (A‐C). Relationship between Hopx gene expression and HSPCs. 61

Figure 5 (A‐D). Functional validation of the effects of Hopx on
HSPC physiology. 62

Chapter 2

Figure 1 (A‐F). Variation in ABR in the HMDP. 106

Figure 2 (A‐F). Manhattan plots of GWAS results for ABR at each
tone burst frequency. 108

Figure 3. In situ images exhibiting cochlear mRNA
expression of GWAS candidate genes. 109

Figure 4. Amino acid sequence around predicted Asn2170Ser
substitution. 110

Chapter 3

Figure 1. Characterization of post‐exposure thresholds in the HMDP. 159

Figure 2. GWAS results for post‐noise exposure thresholds in the HMDP. 160

Figure 3. Regional plot of the 8 kHz ABR post noise‐exposure at Chr 17
association in the HMDP centered on the lead SNP at the Nox3 locus
(rs33652818). 161

Figure 4. Nox3het mice have greater PTS (permanent threshold shift)
for 8 kHz. 162
v

Figure 5. Genotypic effects of the peak SNP (rs33652818) at the
Nox3 locus. 163

Figure 6 (A‐B). Topographical analysis of the auditory pathway at
Different frequencies (post‐noise exposure). 164

Figure 7 (A‐B). Detailed analysis of the 8 kHz frequency stimulus. 165

Figure 8 (A‐B). Cytocochleogram of wild‐type and mutant Nox3 mice. 166

Figure 9 (A‐B). 8 kHz Synaptic Cochleogram. 167
vi

ABSTRACT
Genome‐wide studies in humans focused on the discovery of genetic mechanisms
involved in common, complex phenotypes have enjoyed a moderate degree of
success, but are limited by a number of factors. These include statistical power,
feasibility of functional validation of candidate genes, and ability to collect relevant
biological tissue for the trait. These issues can be addressed by performing similar
analyses in mouse models, namely, by using a platform termed the Hybrid Mouse
Diversity Panel (HMDP) that was developed for such studies. The HMDP has been
used in a range of genome‐wide association studies (GWAS) for complex traits, and
in this dissertation, I describe three additional applications for the panel. Here, we
have validated novel genes regulating hematopoietic stem cell frequency and noise
induced hearing loss, and identified a number of loci that are significantly
associated with hearing.
TABLE OF CONTENTS
PAGE
DEDICATION i
ACKNOWLEDGEMENTS ii
LIST OF TABLES iii
LIST OF FIGURES iv
ABSTRACT vi

INTRODUCTION
COMPLEX TRAITS AND GENE DISCOVERY 1
LINKAGE MAPPING STRATEGIES 1
GENOME WIDE ASSOCIATION STUDIES 4
MOUSE MODELS FOR GWAS 5
STATISTICAL METHODS FOR HMDP ANALYSIS 9
MOUSE STUDIES WITH COMPLEX PHENOTYPES 10
REFERENCES 13

CHAPTER 1: THE GENETIC LANDSCAPE OF HEMATOPOIETIC
STEM CELL FREQUENCY IN MICE
SUMMARY 15
TITLE PAGE 17
ABSTRACT 18
INTRODUCTION 19
RESULTS 22
DISCUSSION 31
MATERIALS AND METHODS 36
ACKNOWLEDGEMENTS 41
REFERENCES 42
FIGURE LEGENDS 47
TABLES 51
FIGURES 53
SUPPLEMENTAL MATERIAL 64

CHAPTER 2: THE GENETIC ARCHITECTURE OF HEARING
IMPAIRMENT IN MICE: EVIDENCE FOR FREQUENCY SPECIFIC
GENETIC DETERMINANTS
SUMMARY 75
TITLE PAGE 77
ABSTRACT 78
INTRODUCTION 79
RESULTS 83
DISCUSSION 87
MATERIALS AND METHODS 94
ACKNOWLEDGEMENTS 98
REFERENCES 99
FIGURE LEGENDS 101
TABLES 104
FIGURES 106
SUPPLEMENTAL MATERIAL 112

CHAPTER 3: GENOME‐WIDE ASSOCIATION STUDY IDENTIFIES
NOX3 AS A CRITICAL GENE FOR SUSCEPTIBILITY TO
NOISE‐INDUCED HEARING LOSS
SUMMARY 121
TITLE PAGE 122
ABSTRACT 123
INTRODUCTION 124
RESULTS 127
DISCUSSION 132
MATERIALS AND METHODS 141
ACKNOWLEDGEMENTS 149
REFERENCES 150
FIGURE LEGENDS 154
TABLES 157
FIGURES 159
SUPPLEMENTAL MATERIAL 168

CONCLUSIONS: THE HMDP IN GENETIC STUDIES
A POWERFUL TOOL FOR GWAS 169
CONTROLLING FOR ENVIRONMENTAL VARIABLES 170
FUTURE DIRECTIONS FOR STEM CELL AND HEARING
PHENOTYPE GENETIC STUDIES 171
GENE BY ENVIRONMENT INTERACTIONS IN THE GWAS MODEL 174
GXE AND THE HMDP 175
BEYOND GXE 177
SUMMARY 180
REFERENCES 182

REFERENCES AND BIBLIOGRAPHY 185
INTRODUCTION
Complex Traits and Gene Discovery
The vast majority of common, chronic illnesses are not the result of rare
genetic mutations in single genes, but rather genetic heterogeneity: the cumulative
effects of many genes, gene networks, environmental exposures, and their
interactions. However, even genes with smaller effects on a phenotype can become
a therapeutic target to treat a disease and increase the health and well‐being of the
patient. Thus, concerted research efforts are necessary to identify genes that are
causally associated with increased risk of developing common disorders, such as
cardiovascular, neurodegenerative, and respiratory diseases. As these diseases are
polygenic, and many of the genes involved remain unknown, “forward genetics”
approaches – which leverage naturally occurring variation in the genome to map
positional candidate genes – can be useful to improve our working knowledge of
the complex biological systems underlying these traits. Importantly, advances in
molecular, statistical, and computational technologies and their continued
evolution have provided the necessary tools to elucidate the genetic complexity of
common diseases.

Linkage Mapping Strategies
In humans and model organisms, including the mouse, several unbiased
approaches have been utilized to discover common genetic variants that correlate
1
with phenotypic differences or disease status. Among these are quantitative trait
locus (QTL) linkage studies, which interrogate whether individuals sharing similar
alleles at particular loci are also more likely to be phenotypically similar. In
particular, it follows from the logic that if a mutation on an ancestral chromosome
gives rise to a new or altered trait, then offspring inheriting that mutation will also
exhibit that trait. After generations of breeding, the original ancestral chromosome
will have undergone several recombinations, thereby breaking the ancestral
haplotype into smaller blocks and truncating the region around the mutation.
Identity by descent (IBD) individuals carrying the region containing the critical
mutation should still exhibit the phenotype of interest, but the smaller haplotype
block makes it much easier to identify the causal gene for the trait. In humans,
locating the gene can be accomplished by genotyping family members using
microsatellite markers, or DNA repeat sequences, that exist throughout the
genome. The family members are then classified by phenotype status into
pedigrees, and the location of the candidate is determined by analysis of how
closely any microsatellite markers co‐segregate with the trait of interest. Studies
using this method have been proven successful for identifying the genes underlying
rare Mendelian disorders in humans [1] due to the higher penetrance and effect
sizes of the causal alleles.
Analogous to human families, linkage association studies can also be carried
out in mice. This is performed by first crossing two inbred strains of mice to
2
generate F1 mice, with an equal 50% of their genome contributed from each of the
two parental strains. The F1 mice are then either brother‐sister mated to each
other (F2 intercross) or back to one of the parental strains (N2 backcross), and the
F2 or N2 offspring are phenotyped and genotyped for carrying out linkage analysis.
Despite its feasibility and statistical power for mapping broad chromosomal regions
in the genome, classic linkage studies for complex traits in mice still do not provide
sufficient mapping resolution to directly implicate genes within the identified loci.
This has been addressed to some degree of success by the generation of congenic
strains [2], in which the region of interest from one mouse strain is introgressed
onto the genetic background of a parental strain by selective breeding. Such
congenic mice can help narrow down the QTL by dividing a large locus into smaller
regions through the creation of sub‐congenic strains where residing genes can be
more thoroughly investigated, but this approach still has limitations. For example,
in addition to being very expensive and time‐consuming, congenic mice also may
only exhibit very subtle phenotypic differences from their parental strains [3] due to
small effect sizes of the causal alleles. Further, altering the genetic background may
also influence interactions between the positional candidate genes in the region of
interest with other genes outside of the congenic region, potentially masking any
genetic effects. Conversely, Chromosome Substitution Strains (CSS), also known as
consomic mice [4], are also derived from two parental inbred strains, but they are
bred to select one complete chromosome without any recombinations from a
donor parental strain against the “host” background from the other parent. For
3
example, 21 CSS has been created between strains A/J and C57BL/6 that have been
used to identify QTLs by simply screening the panel for phenotypic differences with
the background strain [5]. CSS are also particularly useful for confirming QTLs
identified in crosses using A/J and C57BL/6 [6]. However, this approach may still be
compromised by epistatic effects, unless the interacting loci are located on the
same chromosome, and the additional breeding that will still be required to narrow
down the critical interval poses the same challenges as with congenic mice [7].

Genome Wide Association Studies
To address some of these concerns, the genome‐wide association study
(GWAS) approach has rapidly replaced linkage analysis as the primary method for
gene discovery of complex diseases in humans. Because GWAS are not limited to
the study of families, hundreds of thousands of individuals can be included, thereby
providing superior power to detect common variants in the genome that contribute
to polygenic diseases. As such, the method has successfully identified numerous
genes in humans for many complex disease phenotypes [8]. However, even this
robust method has drawbacks: studies can still be limited in power, may not
address gene‐gene interactions, and for many phenotypes with a substantial
environmental component, GWAS may not adequately incorporate those variables
into the statistical model or test for interactions between genes and the
environment [9].
4
Importantly, many potential drawbacks of the GWAS method can be
resolved by performing similar analyses in mice. Several strategies currently exist to
address concerns of sufficient statistical power, genetic diversity and mapping
resolution, and time‐ and cost‐efficiency [10]. In particular, an important advantage
of mouse GWAS is the ease of collecting tissues, in contrast to human studies,
where many phenotypes are either difficult or would be nearly impossible to
investigate because of the inability to obtain the relevant disease tissue.
Additionally, because environmental variables can be carefully controlled in the
laboratory, performing GWAS in mice either removes the necessity to adjust for
environmental effects in the statistical model, or allows for analyses that directly
address and interrogate gene‐by‐environment (GxE) interactions. Identifying genes
in the mouse also streamlines the transition from genetic screen to functional
validation of positional candidates because an arsenal of powerful and relatively
inexpensive gene perturbation resources, including commercially available
knockouts for thousands of genes, is available and well‐established [11, 12].

Mouse Models for GWAS
A variety of large‐scale mouse breeding programs are under development
for use in genetic association studies, each with a distinct set of advantages (Figure
1). Strategies using inbred mice, such as the Collaborative Cross and the Hybrid
5
Mouse Diversity Panel, are attractive genetic platforms for several reasons. First, a
panel of classical inbred (CI) laboratory mice with previously obtained dense single
nucleotide polymorphism (SNP) genotypes or whole genome sequences are
commercially available from the Jackson Lab [13, 14], resulting in a comprehensive
map of genomic variation in these strains [15]. As these strains are homozygous at
each locus, statistical power to detect genetic associations also increases [10].
These resources are complemented with the availability of additional recombinant
inbred (RI) strains of mice. These RI strains are generated similarly to the beginning
of a standard intercross, but rather than stopping after the F2 generation, the
brother‐sister pairs are randomly mated to each other for other 20 generations
until the offspring achieve homozygosity across the genome and each newly
derived strain possesses a unique combination of roughly 50% of each parent’s
genome. This mating scheme can produce large sub‐panels of mice; for instance,
the BXD RI panel bred from the mating of C57BL/6J and DBA/2J includes over 70
unique RI strains just from the initial pairing of two CI strains.
Figure 1 Overview of strategies for genome‐wide association in the mouse.
Figure is reproduced, with permission, from REF [10].

6
The Collaborative Cross (CC) represents another strategy with mouse
models that is generating new RI strains derived from a population of eight founder
strains [16]. However, unlike traditional RI strains that are a genetic mixture of only
two founder strains, the derived CC strains are the combined ancestry of eight
strains. The panel of founders includes five CI strains and three wild‐derived strains,
which introduces additional genetic diversity into the RI offspring, as the CI strains
share a high degree of relatedness. Currently, some studies have preliminarily
began mapping studies using the already available CC strains [17], but the entire CC
with 350+ genetically independent RI strains is still under production [18].
In contrast to panels of inbred strains, some investigators have used outbred
heterogeneous stock animals [19], descended from eight CI strains, or developed
the Diversity Outbred (DO) mouse panel [20], which includes strains derived from
the eight founder CC strains. The main advantage of using outbred mice in a GWAS
is the very high mapping resolution afforded by the generation of any number of
genetically diverse animals that allows for large studies and the potential
identification of weaker genetic effects; however, as each DO animal is genetically
unique, this approach is more time consuming and costly since genotyping must still
be performed for every individual in the cross [10].
The Hybrid Mouse Diversity Panel (HMDP) [21], like the CC, leverages the
power of inbred CI and RI strains; however, the strains of this panel are all
commercially available, unlike several of the CC RI strains that are being
7
independently bred across several laboratories. In addition, these strains are not de
novo crosses and genotyping has already been previously performed for each of
them [13]. Using genetically identical inbred mice also allows reproducibility of
experiments [10] since multiple animals of each strain are phenotyped. Thus, this
provides the advantage of screening for phenotypic outliers and controlling for
overall phenotypic variance. The inclusion of approximately 30 CI strains provides
the phenotypic diversity and mapping resolution to facilitate rapid prioritization of
positional candidates [21] whereas the addition of 70 RI strains increase statistical
power [21]. Though hundreds of inbred strains have been derived, the 30 selected
CI strains are the most genetically informative and have been successfully used in
prior mouse association studies. Their genotypes have enough shared ancestry with
the remaining strains that the haplotypes represented among them sufficiently
covers the range of genetic diversity, without compromising mapping resolution
[22‐25]. The RI sets in the HMDP include the previously mentioned BXD strains, as
well as three additional sets of strains derived from crosses between C57BL/6J and
A/J, Balbc/J or C3H/HeJ. Simulation studies have shown that the collection of these
strains in the HMDP provides adequate power to map genes that contribute to as
little as 5% of the total variance of a trait [26]. For these reasons and because
previous studies using the platform have successfully identified novel candidate
genes for complex traits [21, 26‐35], we have elected to use the HMDP for our
GWAS analyses.
8

Statistical Methods for HMDP Analysis
One substantial concern in association studies with mice is that linear
regression, upon which more complex association algorithms are based, carries the
assumption that the analysis is being performed on unrelated individuals. In fact,
inbred laboratory mice and outbred mice derived from inbred strains can be highly
related, resulting in a complex population structure that can generate false
positives and spurious associations [21, 36]. Since traditional methods to correct for
population structure can be computationally intensive, an Efficient Mixed‐Model
Association (EMMA) algorithm specifically suited to GWAS in the HMDP was
developed by Kang et. al. [36]. This method accomplishes the relatedness
correction and is made available on a public webserver
(http://mouse.cs.ucla.edu/emmaserver/) to perform analysis using either a panel of
132,000 genotyped or 4 million genotyped and imputed SNPs that are available in
over 100 CI and RI strains. The first chapter of my dissertation utilizes this method.
More recently, groups such as Lippert et. al. [37, 38], have developed even faster
versions of the EMMA algorithm by proposing alternative methods of correction for
population structure. We incorporated this algorithm, called FaST‐LMM, into an
automated pipeline on a high‐performance computing cluster that was used for the
studies detailed in the second and third chapters of my dissertation.
9
Building on their earlier work, the Eskin group at UCLA has recently also
improved the speed and relatedness correction of the initial EMMA software in a
new implementation called pyLMM [39]. While sharing comparable processing time
to the FaST‐LMM pipeline, in our studies, the pyLMM algorithm provides the most
reliable correction for population structure, both in inbred mice and in admixed
human populations. Additionally, the pyLMM software allows for the inclusion of a
GxE option, in which an environmental variable is included in the model to test for
specific instances where genetic effects are influenced by environmental exposures.
Ongoing and future studies are moving forward with the pyLMM algorithm due to
this important consideration and its efficiency in accounting for genetic relatedness.

Mouse Studies with Complex Phenotypes
My work has taken advantage of the robust applications of the HMDP and
appropriate linear mixed‐model algorithms to investigate complex phenotypes that
have achieved limited results in human studies. Our first study focused on
identifying the genetic determinants of hematopoietic stem cells (HSC) frequency in
the bone marrow compartment. The aims of the study were threefold: first, we
established a phenotypic profile of three HSC populations in the HMDP; second, we
hypothesized whether the GWAS approach had the statistical power to successfully
identify positional candidates for a phenotype that would be logistically challenging
to ascertain in humans; lastly, we validated our most promising candidate using a
10
variety of bioinformatics and functional experimental approaches and
demonstrated that one positional candidate identified in the GWAS analysis does
play a significant role in HSC frequency.
Encouraged by the success of using the HMDP in one phenotype that is
difficult to determine in humans, our second study also leveraged the power of the
HMDP and FaST‐LMM platforms to explore the genetic determinants of hearing.
This is another example of a phenotype where the analogous study in humans
would also be difficult to carry out due to the inability to control for environmental
exposure over time, whereas in the mouse, all animals are exposed to the same
environment and differences in hearing can be more directly attributed to genetics.
The overall goals of this study were to genetically dissect the functionality of the
cochlea. This was accomplished by performing GWAS analyses at six different
frequencies across the hearing spectrum, which permitted us to determine whether
different genes are involved in hearing at regions of the cochlea that correspond to
the register of each frequency. Importantly, in addition to identifying many
positional candidates for general hearing, we observed that some candidates are
unique to one frequency and that they cluster at either the lower or higher ends of
the frequency spectrum. This suggests a complex genetic architecture of hearing,
where different patterns of gene activity across the cochlea impact frequency‐
specific hearing. While functional validation of these candidates is still forthcoming,
11
this study provides evidence that the HMDP can also be used to parse networks of
genes where members are specifically activated in subtly different phenotypes.
My third chapter extended the genetics of hearing study with the HMDP by
incorporating a noise exposure and identifying loci that were associated with
hearing through GxE interactions. By introducing an intentional environmental
component, our aim was to not only consider the complex genetics of cochlear
biology, but to utilize the mouse as a more controlled experimental in vivo model
for noise‐induced hearing loss compared to what humans experience throughout
their lifetime. By surveying genes that appear to be involved in hearing specifically
after the phenotype was affected by noise (i.e. GxE interaction), we established in
this study that there was significant evidence for natural genomic variation to play a
role in the extent of hearing loss after noise exposure.
12
REFERENCES
1. Baron, M., The search for complex disease genes: fault by linkage or fault by
association? Mol Psychiatry, 2001. 6(2): p. 143‐9.
2. Solberg Woods, L.C., QTL mapping in outbred populations: successes and
challenges. Physiol Genomics, 2014. 46(3): p. 81‐90.
3. Flint, J. and R. Mott, Applying mouse complex‐trait resources to behavioural
genetics. Nature, 2008. 456(7223): p. 724‐7.
4. Buchner, D.A. and J.H. Nadeau, Contrasting genetic architectures in different mouse
reference populations used for studying complex traits. Genome Res, 2015. 25(6): p.
775‐791.
5. Nadeau, J.H., et al., Chromosome substitution strains: gene discovery, functional
analysis, and systems studies. Mamm Genome, 2012. 23(9‐10): p. 693‐705.
6. Sa, Q., et al., Quantitative trait locus analysis for hemostasis and thrombosis.
Mamm Genome, 2008. 19(6): p. 406‐12.
7. Nadeau, J.H., et al., Analysing complex genetic traits with chromosome substitution
strains. Nat Genet, 2000. 24(3): p. 221‐5.
8. Hindorff, L.A., et al., A catalog of published genome‐wide association studies. 2014.
Available at: http://www.genome.gov/gwastudies/.
9. Manolio, T.A., et al., Finding the missing heritability of complex diseases. Nature,
2009. 461(7265): p. 747‐53.
10. Flint, J. and E. Eskin, Genome‐wide association studies in mice. Nat Rev Genet,
2012. 13(11): p. 807‐17.
11. Skarnes, W.C., et al., A conditional knockout resource for the genome‐wide study of
mouse gene function. Nature, 2011. 474(7351): p. 337‐42.
12. Ringwald, M., et al., The IKMC web portal: a central point of entry to data and
resources from the International Knockout Mouse Consortium. Nucleic Acids Res,
2011. 39(Database issue): p. D849‐55.
13. Yang, H., et al., A customized and versatile high‐density genotyping array for the
mouse. Nat Methods, 2009. 6(9): p. 663‐6.
14. Yang, H., et al., Subspecific origin and haplotype diversity in the laboratory mouse.
Nat Genet, 2011. 43(7): p. 648‐55.
15. Keane, T.M., et al., Mouse genomic variation and its effect on phenotypes and gene
regulation. Nature, 2011. 477(7364): p. 289‐94.
16. Churchill, G.A., et al., The Collaborative Cross, a community resource for the genetic
analysis of complex traits. Nat Genet, 2004. 36(11): p. 1133‐7.
17. Ram, R., et al., Rapid identification of major‐effect genes using the collaborative
cross. Genetics, 2014. 198(1): p. 75‐86.
18. Collaborative Cross, C., The genome architecture of the Collaborative Cross mouse
genetic reference population. Genetics, 2012. 190(2): p. 389‐401.
19. Valdar, W., et al., Genome‐wide genetic association of complex traits in
heterogeneous stock mice. Nat Genet, 2006. 38(8): p. 879‐87.
20. Svenson, K.L., et al., High‐resolution genetic mapping using the Mouse Diversity
outbred population. Genetics, 2012. 190(2): p. 437‐47.
21. Bennett, B.J., et al., A high‐resolution association mapping panel for the dissection
of complex traits in mice. Genome Res, 2010. 20(2): p. 281‐90.
13
22. Pletcher, M.T., et al., Use of a dense single nucleotide polymorphism map for in
silico mapping in the mouse. PLoS Biol, 2004. 2(12): p. e393.
23. Cervino, A.C., et al., An integrated in silico gene mapping strategy in inbred mice.
Genetics, 2007. 175(1): p. 321‐33.
24. Liu, P., et al., Large‐scale in silico mapping of complex quantitative traits in inbred
mice. PLoS One, 2007. 2(7): p. e651.
25. Liu, P., et al., Candidate lung tumor susceptibility genes identified through whole‐
genome association analyses in inbred mice. Nat Genet, 2006. 38(8): p. 888‐95.
26. Ghazalpour, A., et al., Hybrid mouse diversity panel: a panel of inbred mouse strains
suitable for analysis of complex genetic traits. Mamm Genome, 2012. 23(9‐10): p.
680‐92.
27. Bennett, B.J., et al., High‐resolution association mapping of atherosclerosis loci in
mice. Arterioscler Thromb Vasc Biol, 2012. 32(8): p. 1790‐8.
28. Farber, C.R., et al., Mouse genome‐wide association and systems genetics identify
Asxl2 as a regulator of bone mineral density and osteoclastogenesis. PLoS Genet,
2011. 7(4): p. e1002038.
29. Ghazalpour, A., et al., Comparative analysis of proteome and transcriptome
variation in mouse. PLoS Genet, 2011. 7(6): p. e1001393.
30. Ghazalpour, A., et al., Genetic regulation of mouse liver metabolite levels. Mol Syst
Biol, 2014. 10(5): p. 730.
31. Park, C.C., et al., Gene networks associated with conditional fear in mice identified
using a systems genetics approach. BMC Syst Biol, 2011. 5: p. 43.
32. Orozco, L.D., et al., Unraveling Inflammatory Responses using Systems Genetics and
Gene‐Environment Interactions in Macrophages. Cell, 2012. 151(3): p. 658‐70.
33. Davis, R.C., et al., Genome‐wide association mapping of blood cell traits in mice.
Mamm Genome, 2013. 24(3‐4): p. 105‐18.
34. Hartiala, J., et al., Comparative genome‐wide association studies in mice and
humans for trimethylamine N‐oxide, a proatherogenic metabolite of choline and L‐
carnitine. Arterioscler Thromb Vasc Biol, 2014. 34(6): p. 1307‐13.
35. Parks, B.W., et al., Genetic control of obesity and gut microbiota composition in
response to high‐fat, high‐sucrose diet in mice. Cell Metab, 2013. 17(1): p. 141‐52.
36. Kang, H.M., et al., Efficient control of population structure in model organism
association mapping. Genetics, 2008. 178(3): p. 1709‐23.
37. Lippert, C., et al., FaST linear mixed models for genome‐wide association studies.
Nat Methods, 2011. 8(10): p. 833‐5.
38. Listgarten, J., et al., Improved linear mixed models for genome‐wide association
studies. Nat Methods, 2012. 9(6): p. 525‐6.
39. Sul, J.H., et al., Accounting for population structure in gene‐by‐environment
interactions in genome‐wide association studies using mixed models. 2015,
University of California, Los Angeles: Unpublished manuscript.

14
CHAPTER 1: THE GENETIC LANDSCAPE OF HEMATOPOIETIC STEM CELL FREQUENCY
IN MICE

Summary: In this study, we performed a genome‐wide association study (GWAS) in
a panel of 108 strains of mice selected from the Hybrid Mouse Diversity Panel
(HMDP) to identify genes impacting the frequency of three hematopoietic
stem/progenitor cell (HSPC) populations in bone marrow. Using flow cytometric
analysis, we characterized the composition of HSPCs in the HMDP strains and
established a phenotypic profile that indicates a high degree of variability among
strains. This is important because typically in stem cell candidate gene studies, only
a limited number of strains such as C57BL/6J, DBA/2J, and C3H/HeJ are used. Our
study suggests the imperative of more considered strain selection for these studies,
as using a strain with a less robust presentation may conceal subtle but nonetheless
significant effects on phenotype induced by genetic alteration. For example,
frequency of the most primitive LSKCD150
+
CD48
‐
HSPCs varies in the phenotyped
panel approximately 300‐fold, while “short‐term” LSKCD150
‐
CD48
‐
HSPCs and LSK
progenitors in out phenotyped strains display approximately 150‐fold and 120‐fold
variation, respectively. By leveraging this high degree of natural phenotypic
variation, we performed GWAS for each of the three cell populations and identified
several significant and suggestive loci. The most promising locus for short term
HSPCs, on chromosome 5, contained a number of positional candidate genes which
we prioritized using bioinformatics tools, including mouse eQTL data for several
15
tissues. We found that the gene containing the peak SNP, Hopx, had significant cis
eQTLs in bone, macrophage, liver, and heart, indicating that genetic variation in this
region directly affects Hopx expression. Indeed, qPCR from several HMDP strains
revealed a significant correlation between Hopx expression and short term HSPC
frequency. Subsequent flow cytometry and competitive repopulation analyses
comparing Hopx knockout mice to wild‐type littermates demonstrated significantly
lower short term HSPCs from Hopx knockouts, and impaired re‐engraftment of
immune cells in immunocompromised mice when the donor cells were from Hopx
knockouts. Therefore, we conclude that Hopx is a positive regulator of short term
HSPC frequency in mice. My contributions to this project included gathering and
organization of animals for phenotyping, including collection and assistance with
extraction of bone marrow from the femur. In addition, I performed the GWAS and
bioinformatics analysis and co‐wrote the manuscript. This study was published in
the journal Stem Cell Reports (Stem Cell Reports 2015,
http://dx.doi.org/10.1016/j.stemcr.2015.05.008.)

16
The Genetic Landscape of Hematopoietic Stem Cell Frequency in Mice
Xiaoying Zhou
1
*, Amanda L. Crow
2
*, Jaana Hartiala
2
, Tassja J. Spindler
1
, Anatole
Ghazalpour
3
, Lora W. Barsky
1
, Brian B. Bennett
4
, Brian W. Parks
3
, Eleazar Eskin
5
,
Rajan Jain
6
, Jonathan A. Epstein
6
, Aldons J. Lusis
3
, Gregor B. Adams
1
*, Hooman
Allayee
2
*

1
Eli and Edythe Broad Center for Regenerative Medicine and Stem Cell Research at
USC, Keck School of Medicine, University of Southern California, Los Angeles, CA
90033
2
Department of Preventive Medicine and Institute for Genetic Medicine, Keck
School of Medicine, University of Southern California, Los Angeles, CA 90033
3
Departments of Human Genetics, Medicine, and Microbiology, Immunology, and
Molecular Genetics, David Geffen School of Medicine at UCLA, Los Angeles, CA
90095
4
Department of Genetics and Nutrition Research Institute, University of North
Carolina, Chapel Hill, Kannapolis, NC 28081
5
Department of Computer Science and Inter‐Departmental Program in
Bioinformatics, University of California, Los Angeles, CA 90095
6
Department of Cell and Developmental Biology and Penn Cardiovascular Institute,
Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA 19104

*These authors contributed equally to this work.

Short Title: Genetics of Hematopoietic Stem Cells

17
Abstract
Identification of regulators of hematopoietic stem cell physiology has relied
mainly on candidate gene approaches with genetically modified mice. Here we used
a genome‐wide association study (GWAS) strategy with the hybrid mouse diversity
panel (HMDP) to identify the genetic determinants of hematopoietic
stem/progenitor cell (HSPC) frequency in the adult bone marrow (BM). Among 108
strains characterized, we observed ~120 to 300‐fold variation in three HSPC
populations. A GWAS analysis identified several loci that were significantly
associated with HSPC frequency, including a locus on chromosome 5 harboring
homeodomain only protein (Hopx). Hopx had previously been implicated in cardiac
development but is otherwise not known to influence HSPC biology. Analysis of the
HSPC pool in Hopx
‐/‐
mice demonstrated significantly reduced cell frequencies and
impaired engraftment in competitive repopulation assays, thus providing functional
validation of this positional candidate gene. These results demonstrate the power
of GWAS in mice to identify novel genetic determinants of the hematopoietic
system.

18
Introduction
Genetic studies using knockout mice have elucidated the functions of over
7,000 genes, yet the majority of these analyses can be constrained by the
experimental procedures used to interrogate the gene of interest or the strain of
mouse used [1]. Moreover, the use of candidate gene approaches to identify the
function of genes is inherently biased towards the hypothesis that is being tested.
In the hematopoietic stem cell (HSC) field, the identification of regulators of HSC
physiology has mainly relied on such candidate gene approaches [2], but these
studies have not necessarily provided a complete picture of the complex network of
signals that govern HSC proliferation, differentiation, and function. Therefore non‐
biased strategies, such as transcriptome analyses, have been used. Despite holding
a great deal of promise, these approaches are also sometimes restricted by the
screen itself [3, 4]. For example, many genes have been shown to be expressed at
specific differentiation stages, including the most primitive HSCs [5], but they do not
provide information as to the key regulators of HSC physiology.
An alternative unbiased strategy to elucidate the determinants HSC
frequency/function is to leverage naturally occurring variation in a forward genetics
approach. In this regard, prior studies have shown that hematopoietic
stem/progenitor cell (HSPC) frequency in mice differs as a function of genetic
background [6], and attempts to identify the underlying genes have employed
linkage analysis in crosses between inbred mouse strains. For example, using the
19
long‐term culture‐initiating cell (LTC‐IC; [7] or cobblestone area forming cell (CAFC;
[8] assays, quantitative trait loci (QTL) for HSPC frequency have been identified on
chromosomes 1 and 18 among a panel of recombinant inbred (RI) strains derived
from C57Bl/6 x DBA/2 (BXD). This strategy has also been used to map loci for
dynamic changes in CAFC frequency associated with aging [9, 10]. With the
development of immunophenotypic markers that could identify functional HSPCs
[11], flow cytometric analysis has been used to identify QTLs for HSPC frequency in
the BXD RI panel, but this approach has only been successful in aged mice [12, 13].
In addition, QTLs for HSPCs distinct from those mapped in the BXD RI panel have
been identified using different inbred mouse strains [14‐16], suggesting that the
limited variation in crosses between two strains restricts the identification of other
genetic factors that play a role in HSPC physiology. Furthermore, the classical QTL
approach has inherently low mapping resolution since the regions of interest
typically span large chromosomal intervals and can contain hundreds to thousands
of genes. To overcome these obstacles, a ‘genetical genomics’ approach has been
proposed where transcriptomics analysis is combined with QTL mapping [17‐19].
However, to date, the incorporation of intermediate molecular traits into linkage
analysis has led to the identification of only one gene, latexin, where differential
levels of expression influence the pool of HSPCs in the bone marrow (BM) [20].
Genome‐wide association studies (GWAS) represent another unbiased
approach that has proven particularly successful in humans for gene discovery of
numerous disease phenotypes [21]. Encouraged by these successes, GWAS have
20
also been proposed in mice using a recently developed panel of classic inbred and
RI mouse strains, termed the Hybrid Mouse Diversity Panel (HMDP). Altogether, the
HMDP includes 29 classic inbred strains and 79 RI strains that are maximally
informative for association analysis [22] and has sufficient power to map traits that
contribute to 5% of the overall variance. Importantly, the resolution of the panel is
an order of magnitude better than that achievable using traditional linkage
methods for complex traits. Moreover, the complex genetic relatedness among the
strains can be accounted for by the availability of genotypes of high‐density single
nucleotide polymorphisms (SNPs). The HMDP has been used to map loci for
numerous complex traits relevant to human diseases [22‐32], which collectively
illustrate the power of this approach for gene discovery in mice. In this current
study, we used this genetics platform to carry out a comprehensive phenotypic
screen of 108 different mouse strains from the HMDP to identify novel genetic
regulators of HSPC physiology in the adult BM. We identified multiple loci
associated with the frequencies of three HSPC populations and functionally
validated one positional candidate gene that had previously not been known to play
a role in the hematopoietic system.

21
Results
Variation in HSPC frequencies in the HMDP. While the competitive repopulation
assay is the ‘gold standard’ for measuring HSPC frequency in a given population of
cells [33], this is influenced by a number of parameters, including radiation
sensitivity of the host, homing and lodgment of the injected cells in the BM
microenvironment, and differences in compensatory mechanisms involved in the
differentiation of the mature hematopoietic cells. To limit variation due to any one
or a combination of these parameters, we chose immunophenotypic analysis to
characterize the frequency of primitive HSPC subsets in the adult BM. We obtained
BM mononuclear cells (MNCs) from 12‐week old male mice of 108 HMDP strains
and used flow cytometry to determine the frequency of several HSPC populations.
These included Lineage[Lin]
‐
Sca‐1
+
c‐Kit
+
(LSK) HSPCs, the more immature
LSKCD150
‐
CD48
‐
multipotent progenitors (MPPs), which have been shown to be
able to differentiate into all lineages of the hematopoietic system, but have limited
potential to self‐renew, and LSKCD150
+
CD48
‐
cells, which are the most primitive
long‐term HSCs [34, 35]. In addition, we chose CD150 and CD48 for our analyses
since these “Slam” markers have been validated to identify the different sub‐
populations of HSPCs across multiple inbred mouse strains [36]. Representative
flow cytometry plots that illustrate the gating strategy used for these analyses are
shown in Supplemental Figure 1.
22
As shown in Figure 1, the frequency of LSK cells varied by approximately
120‐fold among the HMDP strains. Analysis of LSKCD150
‐
CD48
‐
and
LSKCD150
+
CD48
‐
cell frequency also revealed ~150‐fold and 300‐fold variation,
respectively, across the strains analyzed (Figure 1). Since all mice were age and sex‐
matched and kept under identical environmental conditions, we attributed these
differences, at least in part, to naturally occurring genetic variation between the
strains. This notion was confirmed by calculating the heritability for each of the
three HSPC sub‐populations, which yielded values of 0.90, 0.92, and 0.70, for LSK,
LSKCD150
‐
CD48
‐
, and LSKCD150
+
CD48
‐
cells, respectively. We note however that
these heritability estimates are somewhat higher than what would be typically
expected for complex traits in humans since phenotype measurements in the
HMDP are obtained from multiple animals of the same genotype (strain).

Relationship between HSPC frequencies and other hematological parameters. The
significant variation observed in the frequencies of LSK, LSKCD150
‐
CD48
‐,
and
LSKCD150
+
CD48
‐
cells across the HMDP provided us the opportunity to explore the
relationship between these sub‐populations and with other hematological
parameters. As shown in Supplemental Figure 2, the three types of primitive HSPCs
were all significantly correlated with each other, with a particularly strong
association between LSK and LSKCD150
‐
CD48
‐
cells (r=0.70; p<0.0001). We next
examined whether the frequencies of the HSPCs correlated with mature
hematological parameters in peripheral blood, including total and specific leukocyte
23
populations. LSK cells exhibited modestly positive, but significant, correlations with
total white blood cell (WBC) count as well as with the number of lymphocytes and
monocytes (Supplemental Table 1). By comparison, LSKCD150
‐
CD48
‐
cells were
negatively correlated with lymphocyte and monocyte counts and positively
associated with granulocytes. With the exception of a weakly positive association
with WBC count, no correlations were observed with the most primitive
LSKCD150
+
CD48
‐
cells. Furthermore, no significant correlations were observed
between any of the three HSPC populations and other red blood cell (RBC) traits,
such as hemoglobin and hematocrit levels (Supplemental Table 1). Taken together,
these data suggest that variation in LSK and LSKCD150
‐
CD48
‐
cells and mature white
blood cells could be controlled, in part, by similar genetic mechanisms whereas
variation in LSKCD150
+
CD48
‐
HSCs as well as RBC parameters may be driven by
distinct factors.

GWAS for HSPC Frequencies. To identify the genetic determinants of HSPC
frequency, we used the phenotype data in the 108 HMDP strains to carry out a
GWAS for the three cell populations (Figure 2A‐C). These analyses utilized ~880,000
genotyped and imputed SNPs and implemented an efficient mixed model algorithm
(EMMA) that takes into account the underlying population structure and genetic
relatedness across the strains [22, 37]. A GWAS for LSKCD150
+
CD48
‐
cells identified
one significantly associated locus at the distal end of chromosome 18 (Figure 2A;
Table 1), where the lead SNP (rs36866074; p=3.2x10
‐6
) mapped to intron 1 of the
24
mitogen‐activated protein kinase 4 (Mapk4) gene (Figure 3A). On chromosome 11
(Figure 2A; Table 1), we also identified an intergenic region located ~464kb
downstream of the neuromedin‐U receptor 2 (Nmur2) and ~507kb upstream of the
glutamate receptor 1 isoform 1 precursor (Gria1) genes that exhibited suggestive
association with LSKCD150
+
CD48
‐
cells (rs29434264; p=2.3x10
‐5
).
A GWAS for LSK cells revealed several suggestively associated loci, including
those on chromosomes 2, 4, 6, and 18, as well as a highly significant (p=3.7x10
‐14
)
locus on chromosome 15 (Figure 2B; Table 1), where the lead SNP (rs31675052) is
located approximately 100kb downstream of Ly6f (Figure 3B). Ly6f is part of a
family of genes located at this locus that encode Sca‐1, which is one of the surface
markers used to immunphenotypically quantitate HSPC frequency. While Sca‐1 is
known to play a role in the function of HSPCs [38], some studies have suggested
that it is not an informative cell surface marker for flow cytometry analysis in
certain mouse strains [39]. To address this potential issue and remove the effect of
the chromosome 15 locus, we re‐performed the GWAS analysis after excluding
strains carrying the low Sca‐1‐expressing haplotype [39] (Supplemental Table 2).
Importantly, exclusion of these strains did not appreciably decrease the heritability
for variation in LSK cells (0.90 vs 0.82). Furthermore, the Sca‐1 locus did not yield an
association signal in this analysis, as expected, but the suggestive peak on
chromosome 18 increased in significance from p=4.3 x10
‐4
to just below the
threshold for genome‐wide significance with a p=9.4x10
‐6
(Table 1; Supplemental
Figure 3A). The peak SNP (rs30267408) on chromosome 18 is not located within a
25
known gene but maps ~363kb distal to the zinc finger protein 521 (Zfp521) and
~290kb proximal to the synovial sarcoma‐associated Ss18‐alpha (Ss18) genes
(Supplemental Figure 3B).
We next performed a GWAS for LSKCD150
‐
CD48
‐
cells and identified three
significantly associated loci on chromosomes 1 (rs8242728; p=9.3x10
‐10
), 5
(rs29633853; p=5.1x10
‐7
), and 15 (rs32350275; p=3.0x10
‐7
) as well as several
suggestively associated loci on chromosomes 3, 11, 17, and 18 (Figure 2C; Table 1).
Since the chromosome 15 locus for LSKCD150
‐
CD48
‐
cells was the same as that
identified for LSKs, we re‐performed the GWAS after excluding the same strains
carrying the low Sca‐1‐expressing haplotype. Similar to LSK cells, excluding these
strains only had a marginal effect on the heritability of LSKCD150
‐
CD48
‐
cells (0.92
vs. 0.88). This analysis strengthened the association signal on chromosome 1
(p=4.4x10
‐12
) but had no effect on chromosome 5 (p=1.1x10
‐6
) or reveal any
additional regions associated with LSKCD150
‐
CD48
‐
cells (Table 1; Supplemental
Figure 4A). The lead SNP on chromosome 1 (rs8242728) fell within a large ~2Mb
linkage disequilibrium (LD) block containing dozens of genes and numerous SNPs
that yielded equivalently significant p‐values (Figure 3C). Given this LD pattern, it
would be difficult to narrow down the underlying causative gene without extensive
follow up. For example, rs8242728 is specifically located within the NADH
dehydrogenase [ubiquinone] iron‐sulfur protein 2 (Ndufs2) gene. However, this
interval also encompasses the “Slam” locus and contains the Cd48 and Slamf1
genes, which encode two of the other flow cytometry markers used for quantitating
26
HSPCs (CD48 and CD150). Interestingly and similar to Sca‐1, deletion of Cd48 in
hematopoietic cells leads to a disruption of the HPC compartment [40]. By
comparison, the most significantly associated SNPs on chromosome 5 are located
within a small LD block, with the peak SNP (rs29633853) localizing to the
homeodomain only protein (Hopx) gene (Figure 3D). Interestingly, the suggestive
association with the peak SNP on chromosome 18 (rs31073841), which maps ~20kb
proximal to the peak SNP of the same Zfp521‐Ss18 locus identified for LSKs, also
increased in significance by nearly one order of magnitude from p=3.8x10
‐4
to p=5.7
x10
‐5
after exclusion of the strains carrying the low Sca‐1‐expressing haplotype
(Table 1; Supplemental Figure 4A). This observation suggests that this chromosome
18 locus may exert pleiotropic effects on multiple HSPC subsets.

Prioritization of positional candidate genes at chromosome 5 locus for LSKCD150
‐
CD48
‐
frequency. Of the loci identified, we focused on the chromosome 5 locus
because of the increased mapping resolution and the possibility that this region
contains a novel gene(s) for LSKCD150
‐
CD48
‐
cells. In particular, rs29633853 is
located within intron 1 of Hopx, which has been shown to play important roles in
cardiac development [41, 42] and lymphoid regulation [43], and serve as a marker
for quiescent intestinal epithelial stem cells [44] and multipotent hair follicle stem
cells [45]. However, Hopx has otherwise not been implicated in the biological
regulation of HSPCs. To evaluate Hopx as a positional candidate, we first searched
the publicly available genomic sequences of ~60 inbred mouse strains but did not
27
identify any amino acid substitutions and/or other protein structure‐altering
variations. We next used gene expression data across multiple tissues in the HMDP
(http://geneeqtl.genetics.ucla.edu/) to determine whether the chromosome 5 locus
exhibited any cis expression QTL (eQTL), defined as those mapping within a 2Mb
interval centered around the lead GWAS SNP (rs29633853). Of the ~20 genes in this
region (Figure 3D), Srd5a3, Tmem165, Exoc1, Cep135, and Hopx exhibited cis eQTLs
in either bone, macrophages, liver, or heart (Table 2). However, only variants in
Hopx yielded both significant associations with LSKCD150
‐
CD48
‐
cells and cis eQTLs
across all four tissues examined. In particular, the lead eQTL SNP in liver
(rs33117479; p=3.4x10
‐18
) was located ~8kb proximal to the lead SNP for LSKCD150
‐
CD48
‐
cells (rs29633853) whereas the lead eQTL SNP in heart (rs33175967; p=
6.2x10
‐19
) was ~17kb distal. The overlap between the association signals for
LSKCD150
‐
CD48
‐
frequency and hepatic Hopx mRNA expression at the chromosome
5 locus is also illustrated in Supplemental Figure 4B.
To further evaluate whether Hopx plays a role in the regulation of HSPCs, we
next performed gene expression analysis by RT‐PCR in BM MNCs of 25 different
mouse strains with varying LSKCD150
‐
CD48
‐
frequencies. These analyses
demonstrated that Hopx was expressed in BM and that its mRNA levels were both
variable and significantly correlated (r=0.44; p=0.03) with LSKCD150
‐
CD48
‐

frequency in these strains (Figure 4A). This latter observation is not entirely
surprising since the coincident mapping of cis eQTLs for Hopx across multiple
tissues and LSKCD150
‐
CD48
‐
frequency to the same chromosome 5 could be
28
expected to result in a correlation between Hopx gene expression and cell number
in BM. To further characterize Hopx expression in BM, we purified various subsets
of HSPCs and other hematopoietic‐derived lineages and determined the cell‐specific
expression pattern of Hopx. These analyses revealed that Hopx was expressed in all
hematopoietic cells and at particularly high levels in mature hematopoietic lineages,
such as CD3
+
T‐lymphocytes (Figure 4B). To confirm these findings, we used flow
cytometry to analyze BM of genetically modified mice that express green
fluorescent protein (GFP) under the control of the endogenous Hopx promoter [45].
As shown in Figure 4C, Hopx was expressed highest in the LSKCD150
‐
CD48
‐

population of HSPCs and in CD3
+
cells, which is consistent with the RT‐PCR results.
Collectively, these data point to Hopx as a strong positional candidate gene since 1)
genetic variation at this locus is functional with respect to Hopx gene expression in
multiple tissues; 2) mRNA levels of Hopx in BM MNCs are positively associated with
LSKCD150
‐
CD48
‐
number; and 3) Hopx is expressed at relatively high levels in the
HSPC subset whose frequency maps to the chromosome 5 locus.

Biological Validation of Hopx as a Gene that Regulates LSKCD150
‐
CD48
‐
Cells. To
biologically validate Hopx as a genetic determinant of HSPCs, we characterized
previously generated Hopx knockout (Hopx
‐/‐
) mice [41] for differences in HSPC
frequency and function. Flow cytometric analysis revealed that Hopx
‐/‐
mice had
significantly reduced numbers of LSKCD150
‐
CD48
‐
cells compared to wildtype
Hopx
+/+
littermates but no differences in the frequency of the LSK (Figure 5A) or
29
LSKCD150
+
CD48
‐
sub‐population of cells (data not shown). Since the reduced
frequency of LSKCD150
‐
CD48
‐
cells in Hopx
‐/‐
mice could due to either intrinsic
defects of the cells or extrinsic factors, we performed competitive repopulation
assays with HSCs from Hopx
‐/‐
mice and wildtype Hopx
+/+
littermates. These
experiments demonstrated that transplantation of HSCs from Hopx
‐/‐
mice did not
affect levels of circulating CD45.2 cells at early time points but led to significantly
impaired engraftment at 16 weeks after transplantation and extending out to 24
weeks (Figure 5B). However, analysis of multi‐lineage engraftment in these mice did
not reveal significant differences in the relative abundance of circulating B‐
lymphocytes, T‐lymphocytes, or myeloid cells (Supplemental Figure 5). We next
analyzed the composition of BM MNCs at the 24 week time point. Mice
transplanted with HSCs from Hopx
‐/‐
mice had a significant and specific reduction in
the frequency of LSKCD150
‐
CD48
‐
cells but not LSK (Figure 5C) or LSKCD150
+
CD48
‐

cells (data not shown) compared to transplantation with Hopx
+/+
HSCs. To
investigate a potential biological mechanism for the functional effects of Hopx
deficiency on HSPC frequency, we analyzed the cell cycle status of LSK and
LSKCD150
‐
CD48
‐
cells. Hopx
‐/‐
mice had significantly reduced numbers of quiescent
LSKCD150
‐
CD48
‐
cells and increased numbers of cells in the G
1
phase (Figure 5D).
However, Hopx deficiency had no cell cycle effects on LSKs (Figure 5D) or
LSKCD150
+
CD48
‐
cells (data not shown). Taken together, these data validate the
GWAS findings and provide functional evidence that Hopx is a positive regulator of
HSPC frequency in mice.
30
Discussion
In the present study, we carried out a phenotypic screen with over 100
inbred strains from the HMDP to investigate how the frequency of three HSPC
populations varies as a function of naturally occurring genetic variation. The results
revealed substantial variability across this panel of mouse strains, at least based on
the specific cell surface markers we used to quantify cell number. However, this
may not necessarily reflect the true variation in the pool size of all stem cells since,
for example, the number of HSPCs in strains carrying the Sca‐1 null allele could be
underestimated. Despite this possibility, we still observed relatively strong
correlations between certain HSPC subsets, but, interestingly, not between HSPCs
and the number of circulating mature blood cells or other hematological
parameters.
Using the strain survey data, we also carried out a GWAS analysis and
identified multiple distinct and overlapping loci that were significantly or
suggestively associated with HSPC sub‐populations. At one particular locus on
chromosome 5, we provide strong evidence that LSKCD150
‐
CD48
‐
cell frequency and
function is regulated by Hopx, a gene that was previously not known to influence
HSPC biology. The collective results of the functional validation studies with Hopx
‐/‐

mice and competitive transplantation assays demonstrated that Hopx deficiency led
to a specific and intrinsic functional defect in the MPP subset of HSCs (LSKCD150
‐
CD48
‐
), possibly related to cell cycle status, but not in the most primitive long‐term
31
subset of cells or more committed progenitor populations. These observations were
also entirely consistent with the gene expression analyses showing that Hopx mRNA
levels in the BM were positively correlated with LSKCD150
‐
CD48
‐
cells and the
GWAS results where the Hopx locus was only associated with this HSPC subset and
not with LSKs or LSKCD150
+
CD48
‐
cells. Of direct relevance to our findings, a recent
study demonstrated how different HSC populations, and LSKCD150
‐
CD48
‐
cells in
particular, play distinct temporal roles in the maintenance of the hematopoietic
system [46]. More specifically, it is the LSKCD150
‐
CD48
‐
sub‐population of cells that
sustains adult hematopoiesis under homeostatic conditions through multiple
rounds of self‐renewal. Furthermore, proliferation of this cell population accounts
for the differentiation of committed progenitors, while only receiving limited input
from the more primitive LSKCD150
+
CD48
‐
subpopulation of cells.
Hopx is a homeodomain only protein that lacks DNA binding activity, yet is
predominantly localized to the nucleus [41, 42]. There are many different functional
roles proposed for this protein depending upon the cell type examined. Through
direct protein‐protein interactions, Hopx was shown to repress the expression of
serum response factor (SRF)‐dependent genes [47], and conditional deletion of Srf
in mouse BM leads to an approximate 3‐fold increase in the frequency of all HSPC
sub‐populations [48]. Therefore, one plausible mechanism for why Hopx deficiency
leads to decreased numbers of LSKCD150
‐
CD48
‐
cells could be as a result of
increased Srf‐mediated transcription of downstream targets. Alternatively, Hopx
may affect HSPC frequency through its effects on cell cycle status. For example, we
32
demonstrated that Hopx
‐/‐
mice have decreased proportion of LSKCD150
‐
CD48
‐
cells
in the quiescent G
0
state and increased numbers in the G
1
phase. Other groups have
shown that Hopx promoter methylation is frequently observed in many
malignancies, including esophageal, gastric, pancreatic, and colorectal cancers,
where it has been suggested that Hopx functions as a tumor suppressor gene [49].
Since it is known that the lack of a number of genes in the hematopoietic system
leads to increased entry into cell cycle, with a consequent loss of HSPC frequency
and function [50], it is possible that perturbation of Hopx expression/activity affects
the intrinsic homeostasis of HSPCs by altering cellular quiescence and thus cell
frequency.
Two other interesting observations from our GWAS results were the overlap
of loci that had previously been mapped for HSPC phenotypes using linkage analysis
and identification of genomic regions that harbored genes encoding cell surface
markers used for flow cytometry. In this regard, we identified a major locus for LSK
cells over the genes encoding Sca‐1 (Ly6e‐Ly6a‐Ly6f) on chromosome 15, which
confirms previous reports that certain strains carry a Sca‐1 haplotype associated
with decreased expression [39]. Exclusion of these strains from the analysis did not
significantly decrease the heritability of LSK variation or reveal additional loci but
yielded highly suggestive evidence for association with a chromosome 18 locus
containing Zfp521. QTL analysis with BXD RI strains also identified a locus for HSC
frequency on chromosome 18 near Zfp521, although the peak of that linkage signal
mapped ~20Mb distal from our lead GWAS SNP [8]. Zfp521 encodes a transcription
33
factor protein with 30 zinc fingers that was originally identified as being specifically
expressed in primitive human CD34
+
cells compared to more mature hematopoietic
cells [51]. More recently, Zfp521 has also been implicated in bone formation [52,
53], but validation of Zfp521 as a gene directly influencing HSPCs will require
additional functional studies. Notably, the loci on chromosomes 15 (Ly6e‐Ly6a‐Ly6f)
and 18 (Zfp521‐Ss18) were identified in the GWAS analyses for LSKCD150
‐
CD48
‐

cells as well. However, aside from these two regions, there was no overlap between
the loci identified for the three HSPC populations or with those previously reported
for blood cell traits in the HMDP [24]. In addition to Ly6e‐Ly6e‐Ly6f, our GWAS
analysis revealed highly significant association of LSKCD150
‐
CD48
‐
cells with the
“Slam” locus on chromosome 1, the strength of which increased by two orders of
magnitude after excluding the low Sca‐1‐expressing strains. Consistent with our
results, Müller‐Sieburg and Riblet [7] also mapped a QTL for HSCs directly over the
“Slam” locus using the CAFC assay and linkage analysis in the BXD RI panel.
However, we still cannot determine with certainty that Cd48 and Slamf1 are the
genes accounting for the signal on chromosome 1 due to extensive LD across this
region. By comparison, we did not observe any associations with SNPs at the locus
encoding c‐Kit, another cell surface marker used to quantitate HSCs.
While GWAS in humans has been successful in identifying genes for many
different disease phenotypes, these types of analysis have had limited success for
HSPCs. This is due, in part, to the logistical challenges of obtaining the relevant
tissue in humans (i.e. BM) in large numbers of subjects in order to have sufficient
34
mapping power. For example, a GWAS in the Framingham Heart Study for levels of
circulating CD34
+
cells in peripheral blood only identified suggestive loci, none of
which could be functionally validated [54]. With respect to the present results, the
relevance of the identified genes and pathways to humans remains to be
determined. However, it is encouraging that recent studies for obesity and blood
cell traits in the HMDP were able to directly correlate mouse association signals
with those identified in human GWAS [24, 32]. For instance, four of the five loci
identified for RBC parameters in the HMDP [24] correspond to loci recently
reported for analogous phenotypes in a large human GWAS [55]. These
observations reinforce the concept that the underlying biological pathways for
complex traits are likely to be conserved between mice and humans.
In summary, we demonstrate the power of the HMDP as a platform to
identify the genetic determinants of highly relevant biomedical traits that would
otherwise not be feasible or logistically challenging to investigate in humans. While
our studies only investigated the genetic determinants of HSPC frequency in the
adult BM, they provide a proof of principle that similar unbiased approaches can be
applied to comparable or more complex phenotypes in other stem cell systems.
Such gene discovery efforts may lead to targets that could be studied towards
further understanding the biological regulation of stem cells and/or manipulated for
therapeutic development.

35
Materials and Methods

Animals. 12 week‐old male mice of 108 inbred strains (n=3‐13) from the HMDP
(Jackson Laboratories) were obtained and used in accordance with the University of
Southern California Institutional Animal Care and Use Committee (IACUC)
guidelines. Mice were housed in sterilized microisolator cages and received
autoclaved food and water ad libitum. Biological validation of Hopx was carried out
using previously generated genetically targeted mice [41, 45].

Flow cytometry. Femurs and tibias were dissected from the mice and BM MNCs
were obtained. To remove any remaining bone fragments or hair, the BM solution
was filtered using a 70μm cell strainer (Becton Dickinson). BM MNCs were
incubated in PBS with fluorescent labeled anti‐mouse CD150, anti‐mouse CD48,
anti‐mouse Sca‐1, and anti‐mouse c‐Kit antibodies (all from eBioscience, San Diego,
CA). Concurrently, cells were incubated with the biotinylated lineage cocktail
(eBioscience). Red cells were then lysed in 1X FACS lysing solution (Becton
Dickinson). The labeled cells were then analyzed by flow cytometry using a LSR II
flow cytometer (Becton Dickinson) and HSPC frequencies were calculated according
to established cell immunophenotypes using FlowJo flow cytometry analysis
software. Lineage‐negative (Lin‐) cells were consistently defined as the lowest 5% of
BM MNCs expressing the lineage cocktail. Gates for other surface markers were
36
standardized and applied across all strains analyzed. Supplemental Figure 1
illustrates the gating strategy used for the flow cytometry analyses in selected
mouse strains that were chosen to represent the range of variation in HSPC
frequency.

Complete blood count analysis. Mice (n=3/strain) were fasted for 4‐5h and bled from
the retro‐orbital plexus under isoflurane anesthesia 2‐3 h after the beginning of the
light cycle. Blood was collected in 20‐μl EDTA‐coated glass capillaries, processed
using standard procedures, and profiled with respect to hematological parameters
using the Heska HemaTrue™ Veterinary Hematology Analyzer (Loveland, CO), as
described previously [24].

Statistical genetics analysis. Association analyses in the HMDP strains were was
carried out using SNP genotype data from the Broad Institute
(www.broadinstitute.org/mouse/hapmap) and the Wellcome Trust Center for
Human Genetics (WTCHG). Using these resources and additional variants
discovered by the NIEHS/Perlegen mouse resequencing project, we imputed the
genotypes of ~4,000,000 SNPs across the genome, with ambiguous genotypes
labeled as “missing.” Of these SNPs, 880,924 were informative in the HMDP with a
minor allele frequency greater than 5% and used in the present GWAS analyses for
HSPC frequency.
37
We applied the following linear mixed model to account for the population
structure and genetic relatedness among strains: y=μ+xβ+u+e where μ represents
mean HSPC frequency, x represents the SNP effect, u represents random effects
due to genetic relatedness with Var(u) = σg2K and Var(e) = σe2I, where K represents
an identity‐by‐descent (IBD) kinship matrix across all genotypes. A restricted
maximum likelihood (REML) estimate of σg2 and σe2 were computed using Efficient
Mixed Model Association (EMMA) [37], and the association mapping was
performed based on the estimated variance component with a standard F test to
test β≠0. The threshold for genome‐wide significance in the HMDP was previously
calculations determined by the family‐wise error rate (FWER) as the probability of
observing one or more false positives across all SNPs per phenotype [22, 37]. These
calculations ran 100 different sets of permutation tests and parametric
bootstrapping of size 1,000 and observed that the genome‐wide significance
threshold at a FWER of 0.05 corresponded to a p‐value of 4.1x10
‐6
, which has been
used in previous studies with the HMDP [22‐32]. This is approximately an order of
magnitude larger than the threshold obtained by Bonferroni correction (1.0x10
−7
),
which would be an overly conservative estimate of significance since nearby SNPs
among inbred mouse strains are highly correlated with each other.

Gene expression analysis. Total RNA was extracted with the use of the RNA
Miniprep Kit (Stratagene) and was reverse‐transcribed into cDNA with the use of
the SuperScript VILO cDNA synthesis kit (Invitrogen) in accordance with the
38
manufacturer’s instructions. Taqman gene expression assay primers and probe sets
(Applied Biosystems and Roche Diagnostics) were used to quantify the expression
of Hopx and Hprt, as an endogenous control. Levels of gene expression were
quantified in triplicate with the use of the 7900HT real‐time PCR system (Applied
Biosystems). Standard curves in these experiments were created with the use of
QPCR mouse reference total RNA (Stratagene).

Competitive repopulation assays. BM MNCs were obtained from Hopx
+/+
and Hopx
‐
/‐
mice and the primitive LSKCD150
+
CD48
‐
HSCs were isolated using cell sorting. To
quantify the ability to reconstitute hematopoiesis, 100 LSKCD150
+
CD48
‐
cells from
Hopx
+/+
or Hopx
‐/‐
mice, along with 250,000 competitor BM MNCs from a B6.SJL
congenic mouse were injected into the tail veins of B6.SJL mice that were lethally
irradiated with 9Gy approximately 24 hours prior to transplantation. Peripheral
blood samples were obtained from the tail vein of hosts starting at 4 weeks after
transplantation and every 4 weeks thereafter, for a total of 24 weeks. PE anti‐
mouse CD45.1, FITC anti‐mouse CD45.2, APC anti‐mouse CD3e, APC‐Cy7 anti‐mouse
CD11b, and biotinylated anti‐mouse B220 antibodies (all from Becton Dickinson)
were used to assess the degree of multi‐lineage reconstitution.

Cell cycle experiments. For cell cycle analysis, cells were incubated with 10µg/ml
Hoechst 33342 (Sigma‐Aldrich) at 37°C for 45 minutes, then stained with lineage
and stem cell antibodies as described above. The stained cells were resuspended
39
and fixed in 10% buffered formalin and incubated at 4°C overnight. To stain for RNA
content, pyronin Y (Polysciences Inc, Warrington, PA) was added to the cells at a
final concentration of 0.75µg/ml and incubated at 4°C for 30 minutes. Cell cycle
status was examined using a LSR II flow cytometer (Becton Dickinson) and FlowJo
software.
40
Acknowledgements
This work was supported in part by California Institute for Regenerative Medicine
grant TG2‐01161 (XZ and ALC); NIH grants T32ES013678 (ALC), K99HL102223 (BBB),
K99HL123021 (BWP), R01ES022282 (EE), R01HL071546 (JAE), P01HL30568 (AJL),
P01HL28481 (AJL), R01ES021801 (HA), 3R01ES021801‐03S1 (GBA and HA); pilot
project awards from the Southern California Clinical and Translational Science
Institute (SC CTSI) through NIH grant UL1TR000130 (GBA and HA) and the Southern
California Environmental Health Sciences Center through NIH grant P30ES007048
(HA); and the Margaret E. Early Medical Research Trust (GBA). The USC core
facilities described were supported in part by P30CA014089 from the National
Cancer Institute. The funders had no role in study design, data collection and
analysis, decision to publish, or preparation of the manuscript.

41
References

1. White, J.K., et al., Genome‐wide generation and systematic phenotyping of
knockout mice reveals new roles for many genes. Cell, 2013. 154(2): p. 452‐
64.
2. Rossi, L., et al., Less is more: unveiling the functional core of hematopoietic
stem cells through knockout mice. Cell Stem Cell, 2012. 11(3): p. 302‐17.
3. Hope, K.J., et al., An RNAi screen identifies Msi2 and Prox1 as having
opposite roles in the regulation of hematopoietic stem cell activity. Cell Stem
Cell, 2010. 7(1): p. 101‐13.
4. Karlsson, C., et al., Identification of the chemokine CCL28 as a growth and
survival factor for human hematopoietic stem and progenitor cells. Blood,
2013. 121(19): p. 3838‐42, S1‐15.
5. Chambers, S.M., et al., Hematopoietic Fingerprints: An Expression Database
of Stem Cells and Their Progeny. Cell Stem Cell, 2007. 1(5): p. 578‐591.
6. de Haan, G., W. Nijhof, and G. Van Zant, Mouse Strain‐Dependent Changes
in Frequency and Proliferation of Hematopoietic Stem Cells During Aging:
Correlation Between Lifespan and Cycling Activity. Blood, 1997. 89(5): p.
1543‐1550.
7. Muller‐Sieburg, C.E. and R. Riblet, Genetic control of the frequency of
hematopoietic stem cells in mice: mapping of a candidate locus to
chromosome 1. J Exp Med, 1996. 183(3): p. 1141‐50.
8. de Haan, G. and G. Van Zant, Intrinsic and extrinsic control of hemopoietic
stem cell numbers: mapping of a stem cell gene. J Exp Med, 1997. 186(4): p.
529‐36.
9. de Haan, G. and G. Van Zant, Dynamic Changes in Mouse Hematopoietic
Stem Cell Numbers During Aging. Blood, 1999. 93(10): p. 3294‐3301.
10. Geiger, H., et al., Age‐ and stage‐specific regulation patterns in the
hematopoietic stem cell hierarchy. Blood, 2001. 98(10): p. 2966‐72.
11. Morrison, S.J. and I.L. Weissman, The long‐term repopulating subset of
hematopoietic stem cells is deterministic and isolatable by phenotype.
Immunity, 1994. 1(8): p. 661‐73.
12. Henckaerts, E., et al., Genetically determined variation in the number of
phenotypically defined hematopoietic progenitor and stem cells and in their
response to early‐acting cytokines. Blood, 2002. 99(11): p. 3947‐54.
42
13. Henckaerts, E., J.C. Langer, and H.W. Snoeck, Quantitative genetic variation
in the hematopoietic stem cell and progenitor cell compartment and in
lifespan are closely linked at multiple loci in BXD recombinant inbred mice.
Blood, 2004. 104(2): p. 374‐9.
14. Jawad, M., et al., QTL analyses of lineage‐negative mouse bone marrow cells
labeled with Sca‐1 and c‐Kit. Mamm Genome, 2008. 19(3): p. 190‐8.
15. Morrison, S.J., et al., A genetic determinant that specifically regulates the
frequency of hematopoietic stem cells. J Immunol, 2002. 168(2): p. 635‐42.
16. van Os, R., et al., Identification of quantitative trait loci regulating
haematopoietic parameters in B6AKRF2 mice. Br J Haematol, 2006. 132(1):
p. 80‐90.
17. Bystrykh, L., et al., Uncovering regulatory pathways that affect
hematopoietic stem cell function using 'genetical genomics'. Nat Genet,
2005. 37(3): p. 225‐32.
18. de Haan, G., et al., A genetic and genomic analysis identifies a cluster of
genes associated with hematopoietic cell turnover. Blood, 2002. 100(6): p.
2056‐2062.
19. Gerrits, A., et al., Combining transcriptional profiling and genetic linkage
analysis to uncover gene networks operating in hematopoietic stem cells and
their progeny. Immunogenetics, 2008. 60(8): p. 411‐422.
20. Liang, Y., et al., The quantitative trait gene latexin influences the size of the
hematopoietic stem cell population in mice. Nat Genet, 2007. 39(2): p. 178‐
88.
21. Hindorff, L.A., et al., A catalog of published genome‐wide association
studies. 2014. Available at: http://www.genome.gov/gwastudies/.
22. Bennett, B.J., et al., A high‐resolution association mapping panel for the
dissection of complex traits in mice. Genome Res, 2010. 20(2): p. 281‐90.
23. Bennett, B.J., et al., High‐resolution association mapping of atherosclerosis
loci in mice. Arterioscler Thromb Vasc Biol, 2012. 32(8): p. 1790‐8.
24. Davis, R.C., et al., Genome‐wide association mapping of blood cell traits in
mice. Mamm Genome, 2013. 24(3‐4): p. 105‐18.
25. Farber, C.R., et al., Mouse genome‐wide association and systems genetics
identify Asxl2 as a regulator of bone mineral density and osteoclastogenesis.
PLoS Genet, 2011. 7(4): p. e1002038.
43
26. Ghazalpour, A., et al., Comparative analysis of proteome and transcriptome
variation in mouse. PLoS Genet, 2011. 7(6): p. e1001393.
27. Ghazalpour, A., et al., Genetic regulation of mouse liver metabolite levels.
Mol Syst Biol, 2014. 10(5): p. 730.
28. Ghazalpour, A., et al., Hybrid mouse diversity panel: a panel of inbred mouse
strains suitable for analysis of complex genetic traits. Mamm Genome, 2012.
23(9‐10): p. 680‐92.
29. Hartiala, J., et al., Comparative genome‐wide association studies in mice and
humans for trimethylamine N‐oxide, a proatherogenic metabolite of choline
and L‐carnitine. Arterioscler Thromb Vasc Biol, 2014. 34(6): p. 1307‐13.
30. Orozco, L.D., et al., Unraveling Inflammatory Responses using Systems
Genetics and Gene‐Environment Interactions in Macrophages. Cell, 2012.
151(3): p. 658‐70.
31. Park, C.C., et al., Gene networks associated with conditional fear in mice
identified using a systems genetics approach. BMC Syst Biol, 2011. 5: p. 43.
32. Parks, B.W., et al., Genetic control of obesity and gut microbiota composition
in response to high‐fat, high‐sucrose diet in mice. Cell Metab, 2013. 17(1): p.
141‐52.
33. Purton, L.E. and D.T. Scadden, Limiting factors in murine hematopoietic stem
cell assays. Cell Stem Cell, 2007. 1(3): p. 263‐70.
34. Kiel, M.J., O.H. Yilmaz, and S.J. Morrison, CD150‐ cells are transiently
reconstituting multipotent progenitors with little or no stem cell activity.
Blood, 2008. 111(8): p. 4413‐4; author reply 4414‐5.
35. Oguro, H., L. Ding, and S.J. Morrison, SLAM family markers resolve
functionally distinct subpopulations of hematopoietic stem cells and
multipotent progenitors. Cell Stem Cell, 2013. 13(1): p. 102‐16.
36. Kiel, M.J., et al., SLAM family receptors distinguish hematopoietic stem and
progenitor cells and reveal endothelial niches for stem cells. Cell, 2005.
121(7): p. 1109‐21.
37. Kang, H.M., et al., Efficient control of population structure in model organism
association mapping. Genetics, 2008. 178(3): p. 1709‐23.
38. Ito, C.Y., et al., Hematopoietic stem cell and progenitor defects in Sca‐1/Ly‐
6A‐null mice. Blood, 2003. 101(2): p. 517‐23.
44
39. Spangrude, G.J. and D.M. Brooks, Mouse strain variability in the expression
of the hematopoietic stem cell antigen Ly‐6A/E by bone marrow cells. Blood,
1993. 82(11): p. 3327‐32.
40. Boles, N.C., et al., CD48 on hematopoietic progenitors regulates stem cells
and suppresses tumor formation. Blood, 2011. 118(1): p. 80‐87.
41. Chen, F., et al., Hop is an unusual homeobox gene that modulates cardiac
development. Cell, 2002. 110(6): p. 713‐23.
42. Shin, C.H., et al., Modulation of cardiac growth and development by HOP, an
unusual homeodomain protein. Cell, 2002. 110(6): p. 725‐35.
43. Albrecht, I., et al., Persistence of effector memory Th1 cells is regulated by
Hopx. Eur J Immunol, 2010. 40(11): p. 2993‐3006.
44. Takeda, N., et al., Interconversion between intestinal stem cell populations in
distinct niches. Science, 2011. 334(6061): p. 1420‐4.
45. Takeda, N., et al., Hopx expression defines a subset of multipotent hair
follicle stem cells and a progenitor population primed to give rise to K6+
niche cells. Development, 2013. 140(8): p. 1655‐64.
46. Busch, K., et al., Fundamental properties of unperturbed haematopoiesis
from stem cells in vivo. Nature, 2015. 518(7540): p. 542‐6.
47. Kook, H. and J.A. Epstein, Hopping to the beat. Hop regulation of cardiac
gene expression. Trends Cardiovasc Med, 2003. 13(7): p. 261‐4.
48. Ragu, C., et al., The transcription factor Srf regulates hematopoietic stem cell
adhesion. Blood, 2010. 116(22): p. 4464‐73.
49. Yamashita, K., H. Katoh, and M. Watanabe, The homeobox only protein
homeobox (HOPX) and colorectal cancer. Int J Mol Sci, 2013. 14(12): p.
23231‐43.
50. Orford, K.W. and D.T. Scadden, Deconstructing stem cell self‐renewal:
genetic insights into cell‐cycle regulation. Nat Rev Genet, 2008. 9(2): p. 115‐
28.
51. Bond, H.M., et al., Early hematopoietic zinc finger protein (EHZF), the human
homolog to mouse Evi3, is highly expressed in primitive human
hematopoietic cells. Blood, 2004. 103(6): p. 2062‐70.
52. Hesse, E., et al., Zinc finger protein 521, a new player in bone formation. Ann
N Y Acad Sci, 2010. 1192: p. 32‐7.
45
53. Kiviranta, R., et al., Coordinated transcriptional regulation of bone
homeostasis by Ebf1 and Zfp521 in both mesenchymal and hematopoietic
lineages. J Exp Med, 2013. 210(5): p. 969‐85.
54. Cohen, K.S., et al., Circulating CD34+ progenitor cell frequency is associated
with clinical and genetic factors. Blood, 2013. 121(8): p. e50‐e56.
55. van der Harst, P., et al., Seventy‐five genetic loci influencing the human red
blood cell. Nature, 2012. 492(7429): p. 369‐75.

46
Figure Legends

Figure 1. Variation in three HSPC populations in the HMDP. The frequency of LSK,
LSKCD150
‐
CD48
‐
, LSKCD150
+
CD48
‐
cells exhibits 120 to 300‐fold variation among
108 HMDP strains. Each dot represents an individual mouse from the respective
strain and the mean values are indicated by the horizontal black bars. BM MNCs
were isolated from the femurs and tibias of 12‐week old male mice (n=3‐8 per
strain; total n=467) and the frequency of different HSPC sub‐populations was
determined by flow cytometry. Data are expressed as a percentage of BM MNCs.

Figure 3. Manhattan plots of GWAS results for HSPC frequency in the HMDP. (A)
The frequency of LSKCD150
+
CD48
‐
cells was significantly associated with a locus on
chromosome 18 harboring Mapk4 and suggestively associated with a region on
chromosome 11 near Nmur2 and Gria1. (B) A GWAS for LSK cells revealed several
suggestively associated loci on chromosomes 2, 4, 6, and 18 as well as a highly
significant locus on chromosome 15 containing Ly6e‐Ly6a‐Ly6f. (C) Three
significantly associated loci on chromosomes 1, 5, and 15 were identified in the
GWAS analysis for LSKCD150
‐
CD48
‐
cells, as well as several suggestively associated
regions on chromosomes 3, 11, 17, and 18. For each significantly associated locus,
the gene(s) nearest to the peak SNP is indicated. The GWAS analyses for each of the
three HSPCs included 880,924 SNPs, whose genomic positions are shown along the
x‐axis with their corresponding ‐log
10
p‐values indicated by the y‐axis. The genome‐
47
wide thresholds for significant (p=4.1x10
‐6
) and suggestive (p=4.1x10
‐4
) evidence of
association are indicated by the horizontal red and blue lines, respectively.

Figure 4. Regional plots of loci significantly associated with HSPCs in the HMDP.
(A) The lead SNP at the chromosome 18 locus for LSKCD150
+
CD48
‐
cells
(rs36866074) maps to intron 1 of Mapk4 (boxed in red). (B) The lead SNP on
chromosome 15 for LSKs (rs31675052) maps to a region harboring Ly6e, Ly6a, and
Ly6f (boxed in red), which are part of a family of genes at this locus that encode one
of the surface markers used to immunphenotypically quantitate HSPC frequency
(Sca‐1). (C) The chromosome 1 locus identified for LSKCD150
‐
CD48
‐
cells
encompasses a large ~2Mb LD block containing dozens of genes and numerous
SNPs that yielded equivalently significant p‐values. Although the lead SNP
(rs8242728) is specifically located in Ndufs2 (boxed in red), this interval also
includes Cd48 and Slamf1 (boxed in red), which encode two of other flow cytometry
markers used for quantitating HSPCs (CD48 and CD150). (D) The peak SNP at the
chromosome 5 locus (rs29633853) associated with LSKCD150
‐
CD48
‐
cells is located
within a small LD block and localizes to Hopx (boxed in red). For each plot, a 1‐2Mb
region is shown and the lead SNP is indicated by a red diamond. In the upper
panels, SNPs positions are shown on the x‐axis with their corresponding ‐log
10
p‐
values indicated by the y‐axis. The location of genes in the selected intervals are
given in the bottom panels.

48
Figure 5. Relationship between Hopx gene expression and HSPCs. (A) Among 25
HMDP strains, Hopx expression levels in BM MNCs are significantly and positively
correlated (r=0.44, p=0.03) with LSKCD150
‐
CD48
‐
frequency. Each data point
represents the mean expression of each mouse strain (n=2‐3 mice/strain) across
two independent experiments. LSKCD150
‐
CD48
‐
frequency is expressed as a
percentage of BM MNCs and expression levels are shown relative to the expression
of Hprt, as an endogenous control, in relative units (ru). (B) Hopx is expressed in
various hematopoietic cell subsets of C57Bl/6 mice, including all three HSPC
populations, but at very low levels in mesenchymal stem cells (MSCs). Expression is
also abundant in mature leukocytes, particularly in T‐lymphocytes (CD3
+
) compared
with B‐lymphocytes (B220
+
) or monocytes (CD11b
+
). Total RNA was isolated from
BM and RT‐PCR was carried out as described in the Methods. Expression levels are
shown relative to the expression of Hprt, as an endogenous control, in relative units
(ru) with error bars representing s.e.m. (C) As determined by relative GFP
fluorescence intensity, Hopx expression was highest in the LSKCD150
‐
CD48
‐

population of HSPCs and especially abundant in CD3
+
cells, consistent with the RT‐
PCR results. Fluorescence was quantitated by flow cytometry in different HSPC sub‐
populations and mature hematopoietic lineages in BM MNCs that were isolated
from the femurs and tibias of 12‐week old genetically modified mice (n=3)
expressing GFP under the control of the endogenous Hopx promoter [45]. Data are
expressed in arbitrary units (au). Note the different X‐axes in the plots.

49
Figure 6. Functional validation of the effects of Hopx on HSPC physiology. (A) The
frequency of LSKCD150
‐
CD48
‐
cells, as a percentage of BM MNCs (left y‐axis), are
significantly reduced in Hopx
‐/‐
mice compared to wildtype littermate controls
(Hopx
+/+
) whereas LSKs are not affected (right y‐axis). BM MNCs were isolated from
the femurs and tibias of 12‐week old male mice and the frequency of different HSPC
sub‐populations was determined by flow cytometry. (B) Competitive repopulation
assay shows significantly decreased levels of engraftment starting at 16 weeks after
transplantation of HSCs isolated from BM of Hopx
‐/‐
mice compared to wildtype
Hopx
+/+
littermates. Mice were bled from the tail vein every 4 weeks after
transplantation and the level of engraftment, indicated by the percent of circulating
CD45.2
+
cells, was quantified by flow cytometry. (C) Frequency of LSKCD150
‐
CD48
‐

cells in BM, as a percentage of MNCs (left y‐axis), is significantly reduced in mice
transplanted with HSCs from Hopx
‐/‐
mice compared to Hopx
+/+
littermate controls
six months following transplantation. There is no difference in the frequency of LSKs
(right y‐axis). (D) LSKCD150
‐
CD48
‐
cells of Hopx
‐/‐
mice exhibit altered cell cycle
status, with decreased and increased percentage of cells in the G
0
and G
1
phases,
respectively, compared to wildtype littermates. No effects were observed in LSKs.
For each experiment in A‐D, error bars represent s.e.m. and n=3‐5 mice per group.
*0<0.05; **p<0.01.

50
Table 1. Loci identified in GWAS for HSPCs in the HMDP.
HSPC Population Chr Position (bp)
1
Lead SNP Nearest Gene(s) MAF p‐value p‐value
2

LSKCD150
+
CD48
‐

11 56318423 rs29434264 Nmur2‐Gria1 0.06 2.3x10
‐5
N/A
18 74202262 rs36866074 Mapk4 0.13 3.2x10
‐6
N/A
LSK
15 75203270 rs31675052 Ly6e‐Ly6a‐Ly6f 0.47 3.7x10
‐14
N/A
18 14494409 rs30267408 Zfp521‐Ss18 0.41 4.3x10
‐4
9.4x10
‐6

LSKCD150
‐
CD48
‐

1 173168046 rs8242728 Ndufs2‐Cd48‐Slamf1 0.50 9.3x10
‐10
4.4x10
‐12

5 77528483 rs29633853 Hopx 0.15 5.1x10
‐7
1.1x10
‐6

15 74669094 rs32350275 Ly6e‐Ly6a‐Ly6f 0.40 3.0x10
‐7
N/A
18 14474199 rs31073841 Zfp521‐Ss18 0.11 3.8x10
‐4
5.7 x10
‐5

Chr, chromosome; MAF, minor allele frequency.
1
base pair position of lead SNP given according to NCBI build 37 of the reference
mouse genome sequence.
2
p‐value obtained from GWAS analysis that excluded strains carrying the uninformative haplotype of
Sca‐1 at the chromosome 15 locus.

51
Table 2. Genes Exhibiting cis eQTLs Across Multiple Tissues at the Chromosome 5 Locus for LSKCD150
‐
CD48
‐
Cells.

eQTL p‐value
Gene Gene Position (bp)
1
SNP SNP Position(bp)
1
GWASp‐value
2
Bone Macrophage Liver Heart
Srd5a3
76569296‐76584529 rs13478330 75857166 0.26 5.8x10
‐7
ns ns ns

rs3153753 76574923 0.99 4.0x10
‐15
ns ns ns

rs4225290 76583956 0.76 ns ns ns 5.0x10
‐8

rs33559924 76254389 0.47 ns ns 9.0x10
‐12
ns

rs3697530 76575062 0.82 ns ns 3.9x10
‐16
ns

rs13478335 76955455 0.41 ns 2.0x10
‐17
ns ns
Tmem165 76612905‐76638270 rs4225290 76583956 0.76 ns 1.5x10
‐11
ns ns
Exoc1 76958336‐76999319 rs4225290 76583956 0.76 9.4x10
‐10
ns ns ns
rs13478334 76817411 0.64 ns 9.4x10
‐12
ns ns
rs33139946 76913636 0.64 ns 9.4x10
‐12
ns ns
rs31555016 76915601 0.64 ns 9.4x10
‐12
ns ns
rs31555860 76921199 0.64 ns 9.4x10
‐12
ns ns
rs29561143 77013954 0.86 ns ns 5.2x10
‐7
ns
Cep135
77075352‐77075401 rs29824030 75977692 0.36 6.2x10
‐8
ns ns ns

rs13478331 76154830 0.38 1.7x10
‐9
ns ns ns
Hopx
77516011‐77544181 rs33077631 77509306 6.1x10
‐6
4.0x10
‐8
ns ns ns
rs33175967 77511463 2.6x10
‐5
ns ns ns 6.2x10
‐19

rs33750358 77534348 6.1x10
‐6
4.0x10
‐8
ns ns ns
rs33117479 77536186 4.8x10
‐5
ns 1.6x10
‐6
3.4x10
‐18
ns
A genomic region ± 1Mb around the peak SNP for LSKCD150
‐
CD48
‐
cells on chromosome 5 (rs29633853; bp position 77528483) was interrogated for
the presence of cis eQTLs in multiple tissues using the UCLA Systems Genetics Resource (http://systems.genetics.ucla.edu/). Only genes exhibiting
cis eQTLs in bone, macrophage, liver, or heart are listed.
1
Base pair (bp) positions of genes and SNPs are given according to NCBI build 37 of the
reference mouse genome sequence.
2
P‐values are given from GWAS results that included all HMDP strains.
52
SK (%)
Figure 1
A
LS
B
50
-
CD48
-
(%)
B
LSKCD15
C
50
+
CD48
-
(%)
Strain
LSKCD1
53
Figure 2
8
A
LSKCD150
+
CD48
-
6
8
Mapk4
Nmur2-Gria1
value)
6
-log
10
(p-v
4
2
17 12 3 4 5 6 7 8 910 16 1819X 11 12 13 14 15
Chromosome
0
54
Figure 2
14
B
Ly6e-Ly6a/Ly6f (Sca-1)
LSK
14
12
value)
10
8
-log
10
(p-v
6 Zfp521-Ss18
2
4
17 12 3 4 5 6 7 8 910 16 1819X 11 12 13 14 15
Chromosome
0
55
Figure 2
10
C
Ndufs2-Cd48-Slamf1
LSKCD150
-
CD48
-
10
8
Hopx
Ly6e-Ly6a-Ly6f (Sca-1)
value)
6
Zfp521-Ss18
-log
10
(p-v
4
2
17 12 3 4 5 6 7 8 910 16 1819X 11 12 13 14 15
Chromosome
0
56
A
Figure 3
rs36866074 6
4
lue)
4
2
-log
10
(p-va
0
73.6 73.8 74.0 74.2 74.4
Position on Chromosome 18 (Mb)
57
B
Figure 3
rs31675052
15
10
value)
5
0
-log
10
(p-v
0
74.4
Position on Chromosome 15 (Mb)
74.6 74.8 75.0 75.4 75.2 75.6
()
58
15
)
C
Figure 3
15
10
5
log
10
(p-value)
rs8242728
0
-l
173.0
Position on Chromosome 1 (Mb)
173.5 174.0 174.5 175.0
59
Figure 3
29633853
8
D
rs29633853
4
g
10
(p-value)
6
2
0
-log
76.5
Position on Chromosome 5 (Mb)
77 77.5 78 78.5
60
Figure 4
61
Figure 5
B
100
/
A
Hopx
+/+
0.14 0.014
*
*
*
5.2
+
Cells
60
80
100
Hopx
+/+
Hopx
-/-
Cells
Hopx
+/+
Hopx
-/-
0.08
0.10
0.12
0.008
0.010
0.012
Cells
% CD45
0
20
40
**
% of
0
0.02
0.04
0.06
0
0.002
0.004
0.006
% of
4 8 12 16 20 24
Weeks after Transplantation
0
LSKCD150
-
CD48
-
LSK
0 0
62
Figure 5
D
100
Hopx
+/+
0.16 0.008
C
f Cells
Hopx
+/+
Hopx
-/-
*
50
60
70
80
90
f Cells
Hopx
/
Hopx
-/-
0.08
0.10
0.12
0.14
0.004
0.005
0.006
0.007
Cells
% of
*
0
10
20
30
40
50
% of
**
0
0.02
0.04
0.06
0.08
0
0.001
0.002
0.003
0.004
% of
LSKCD150
-
CD48
-
LSK
G
1
G
0
S/G
2
/M G
0
G
1
S/G
2
/M
0
LSKCD150
-
CD48
-
LSK
0 0
63

Supplemental Data

The Genetic Landscape of Hematopoietic Stem Cell Frequency in Mice

Zhou et al.

64

Supplemental Table 1. Correlations of Primitive Hematopoietic Cell Frequency with Blood Cell Parameters, Related to
“Relationship between HSPC frequencies and other hematological parameters” in Results.
WBC Ly % Mo % Gr % RBC MCV HCT MCH MCHC RDW% RDWa MPV HGB PLT
LSK
r 0.17 0.14 0.24 0.10 0.02 0.028 ‐0.053 ‐0.040 ‐0.050 0.033 0.030 0.066 0.037 0.04
p‐value 0.005 0.04 <0.0001 0.07 0.75 0.657 0.403 0.525 0.425 0.602 0.633 0.295 0.558 0.43
LSKCD150
‐
CD48
‐

r 0.0 ‐0.18 ‐0.20 0.24 ‐0.05 0.035 ‐0.016 ‐0.002 ‐0.033 0.042 0.064 ‐0.009 ‐0.062 0.04
p‐value 0.90 0.006 0.002 0.0002 0.38 0.574 0.798 0.981 0.601 0.502 0.310 0.891 0.325 0.42
LSKCD150
+
CD48
‐

r 0.15 ‐0.02 ‐0.03 0.04 0.10 ‐0.059 0.035 ‐0.230 ‐0.063 0.088 ‐0.019 ‐0.026 ‐0.009 0.07
p‐value 0.02 0.79 0.69 0.57 0.12 0.352 0.579 < 0.001 0.320 0.161 0.764 0.687 0.891 0.25
WBC: white blood cell count; Ly %: percent lymphocytes; Mo %: percent monocytes; Gr %: percent granulocyte; RBC, red blood
cell count; MCV: mean corpuscular volume; HCT: hematocrit; MCH: mean corpuscular hemoglobin; MCHC: mean corpuscular
hemoglobin concentration; RDW%: red cell distribution width, percent; RDWa: red cell distribution width, area; MPV: mean
platelet volume; HGB: hemoglobin concentration; PLT: platelets. Significant correlations are shown in bold.
65

Supplemental Table 2. HMDP Mouse Strains Used in the Present Study Ordered
According to LSK Frequency, Related to Main Table 2.
1. BXD32/TyJ 29. BXD45/RwwJ 57. C57BL/6J 85. BXD56/RwwJ
2. AXB18/PgnJ 30. BXD42/TyJ 58. BXA7/PgnJ 86. KK/HlJ
3. PL/J 31. BXD34/TyJ 59. DBA/2J 87. BXD16/TyJ
4. BXD14/TyJ 32. BXD15/TyJ 60. BXD64/RwwJ 88. BXHA1
5. BUB/BnJ 33. BXD12/TyJ 61. AXB5/PgnJ 89. BXH2/TyJ
6. BXD70/RwwJ 34. AXB13/PgnJ 62. BXD43/RwwJ 90. CE/J
7. AXB8/PgnJ 35. BXD50/RwwJ 63. BXD75/RwwJ 91. AXB6/PgnJ
8. BXD55/RwwJ 36. BXA11/PgnJ 64. BXD40/TyJ 92. NOD/LtJ
9. CXB12/HiAJ 37. BXD39/TyJ 65. BXD84/RwwJ 93. C3H/HeJ
10. BXD49/RwwJ 38. AXB19/PgnJ 66. BXD66/RwwJ 94. BXA8/PgnJ
11. BXD18/TyJ 39. BXA2/PgnJ 67. 129X1/SvJ 95. BXH8/TyJ
12. BXD29/TyJ 40. BXA12/PgnJ 68. CXB8/HiAJ 96. BXH19/TyJ
13. BXD11/TyJ 41. BXD5/TyJ 69. C58/J 97. SEA/GnJ
14. BXD62/RwwJ 42. AXB12/PgnJ 70. C57BLKS/J 98. CXB3/ByJ
15. BXD73/RwwJ 43. BXA1/PgnJ 71. SM/J 99. CBA/J
16. BXA13/PgnJ 44. BXA4/PgnJ 72. CXB13/HiAJ 100. CXB11/HiAJ
17. AKR/J 45. MA/MyJ 73. BXD86/RwwJ 101. RIIIS/J
18. BXD60/RwwJ 46. CXB7/ByJ 74. BXA26/PgnJ 102. NZB/BlNJ
19. BXD24/TyJ 47. BXH6/TyJ 75. BXD20/TyJ 103. BALB/cJ
20. BXD1/TyJ 48. BXD85/RwwJ 76. BXD61/RwwJ 104. CXB6/ByJ
21. BXH10/TyJ 49. BXD68/RwwJ 77. BXD21/TyJ 105. CXB9/HiAJ
22. LG/J 50. C57L/J 78. FVB/NJ 106. NON/LtJ
23. BXH14/TyJ 51. BXD31/TyJ 79. BXHB2 107. A/J
24. BXD38/TyJ 52. BXD48/RwwJ 80. BXD71/RwwJ 108. I/LnJ
25. SJL/J 53. BTBRT<+>tf/J 81. CXB1/ByJ
26. BXA16/PgnJ 54. BXH4/TyJ 82. BXH9/TyJ
27. AXB15/PgnJ 55. SWR/J 83. BXD44/RwwJ
28. BXD74/RwwJ 56. NZW/LacJ 84. BXA14/PgnJ
Strains are listed from highest to lowest LSK frequency and those carrying the low
Sca‐1‐expressing haplotype (Spangrude & Brooks, 1993), which were excluded from
GWAS analyses for LSK and LSKCD150
‐
CD48
‐
HSPCs, are shown in bold. The 29
classic inbred strains are highlighted in blue.

66

Supplemental Figure 1. Gating strategy for flow cytometric analyses of HSPCs in
the HMDP, Related to Main Table 2. Representative sequential gating strategy for
the Lin
‐
Sca‐1
+
c‐Kit
+
CD150
+
CD48
‐
and Lin
‐
Sca‐1
+
c‐Kit
+
CD150
‐
CD48
‐
cells are
demonstrated. Five mouse strains are shown which have among the highest (CXB‐
8/HiAJ and BXD‐42/TyJ) and lowest (CXB‐9/HiAJ and BALB/cJ) HSPC frequencies of
all strains analyzed. C57BL/6J is also shown for reference. In each case, the Lin
‐
gate
was consistently set as the lowest 5% of BM MNCs.
67

Supplemental Figure 2. Correlation of three HSPC populations in the HMDP,
Related to “Relationship between HSPC frequencies and other hematological
parameters” in Results. The frequency of LSK, LSKCD150
‐
CD48
‐
, LSKCD150
+
CD48
‐

cells are positively correlated with each other, with a particularly strong association
between LSK and LSKCD150
‐
CD48
‐
cells. Each dot represents an individual mouse
from 108 HMDP strains. BM MNCs were isolated from the femurs and tibias of 12‐
week old male mice (n=3‐8 per strain) and the frequency of different HSPC sub‐
populations was determined by flow cytometry. Data are expressed as a percentage
of BM MNCs.
r=0.70
p<0.0001
0
0.15
0.30
0.45
0.60
0.75
00.1 0.2 0.3
LSKCD150
-
CD48
-
(%)
LSK (%)
r=0.33
p<0.0001
0
0.001
0.002
0.003
0.004
0.005
0 0.1 0.2 0.3
LSKCD150
+
CD48
-
(%)
LSKCD150
-
CD48
-
(%)
r=0.32
p<0.0001
0
0.001
0.002
0.003
0.004
0.005
LSKCD150
+
CD48
-
(%)
0 0.15 0.30 0.45 0.60 0.75
LSK (%)
68

Supplemental figure 3A. Manhattan plot of GWAS results for LSK frequency in the HMDP with exclusion of strains carrying the
uninformative Sca‐1‐expressing haplotype, Related to Main Table 2. After the strains carrying the low Sca‐1‐expressing
haplotype were excluded, no locus was significantly associated with LSK cells but a suggestive peak on chromosome 18 harboring
Zfp521 and Ss18 increased in significance to just below the threshold for genome‐wide significance (p=9.4x10
‐6
). These analyses
included 827,406 SNPs, whose genomic positions are shown along the x‐axis with their corresponding ‐log
10
p‐values indicated by
the y‐axis. The genome‐wide thresholds for significant (p=4.1x10
‐6
) and suggestive (p=4.1x10
‐4
) evidence of association are
indicated by the horizontal red and blue lines, respectively.
69

Supplemental figure 3B. Regional plot of the suggestively associated chromosome 18 locus for LSKs after exclusion of strains
carrying the uninformative Sca‐1‐expressing haplotype, Related to Main Table 2. The peak SNP on chromosome 18 for LSK cells
(rs30267408) maps to an intergenic region ~363kb distal to Zfp521 and ~290kb proximal to Ss18 and increased in significance
after the strains carrying the low Sca‐1‐expressing haplotype were excluded (p=9.4x10
‐6
).
70

Supplemental Figure 4A. Manhattan plot of GWAS results for LSKCD150
‐
CD48
‐
frequency in the HMDP with exclusion of strains
carrying the uninformative Sca‐1‐expressing haplotype, Related to Main Table 2. The GWAS analysis that excluded the strains
carrying the low Sca‐1‐expressing haplotype strengthened the association signals for LSKCD150
‐
CD48
‐
cells on chromosomes 1
(p=4.4x10
‐12
) and 18 (5.7 x10
‐5
) but had no effect on chromosome 5 (p=1.1x10
‐6
) or reveal any additional regions. These analyses
included 827,406 SNPs, whose genomic positions are shown along the x‐axis with their corresponding ‐log
10
p‐values indicated by
the y‐axis. The genome‐wide thresholds for significant (p=4.1x10
‐6
) and suggestive (p=4.1x10
‐4
) evidence of association are
indicated by the horizontal red and blue lines, respectively.
71

Supplemental Figure 4B. Coincident association of LSKCD150
‐
CD48
‐
frequency and hepatic Hopx gene expression with a locus
on chromosome 5 in the HMDP, related to main Table 2 and Figure 3B. A Manhattan plot of the GWAS results on chromosome 5
shows overlap of the association signals for LSKCD150
‐
CD48
‐
cells and hepatic Hopx mRNA expression. The lead SNP for the highly
significant cis eQTL in liver for Hopx (rs33117479; p=3.4x10
‐18
) maps ~8kb proximal to the lead SNP for LSKCD150
‐
CD48
‐
cells
(rs29633853; p=1.1x10
‐06
). Blue and red dots correspond to p‐values for LSKCD150
‐
CD48
‐
cells (left y‐axis) and Hopx mRNA levels
(right y‐axis), respectively. Association results for LSKCD150
‐
CD48
‐
frequency are from the GWAS analysis that excluded strains
carrying the low Sca‐1‐expressing haplotype (Spangrude and Brooks, 1993).
72

Supplemental Figure 5. The effect of Hopx deficiency in HSCs on multi‐lineage engraftment after competitive repopulation
assays, Related to Main Figure 5. Despite significantly decreased levels of engraftment starting at 16 weeks after transplantation,
the relative abundance of circulating B‐lymphocytes (B220
+
), T‐lymphocytes (CD3
+
), or myeloid cells (CD11b
+
) was not affected
after transplantation of HSCs from Hopx
‐/‐
mice compared to wildtype Hopx
+/+
littermates. Mice were bled from the tail vein every
4 weeks after transplantation and the level of engraftment as a percentage of total circulating MNCs was determined by flow
cytometry.

73

References

Spangrude GJ, Brooks DM. 1993. Mouse strain variability in the expression of the hematopoietic
stem cell antigen Ly‐6A/E by bone marrow cells. Blood 82(11): 3327‐3332.
74
CHAPTER 2: THE GENETIC ARCHITECTURE OF HEARING IMPAIRMENT IN MICE:
EVIDENCE FOR FREQUENCY SPECIFIC GENETIC DETERMINANTS

Summary: In the present study, we performed a genetic dissection of cochlear
function by utilizing 100 HMDP strains and the FaST‐LMM association algorithm to
detect genetic loci controlling hearing at six different frequencies. First, we
characterized variation among the panel in hearing ability by performing auditory
brainstem response (ABR) tests at 4 kHz, 8 kHz, 12 kHz, 16 kHz, 24 kHz, and 32 kHz.
Across all frequencies, ABR threshold varied 2 to 5‐fold on a scale from 0‐100,
indicating the likelihood of genetic involvement, as all mice experienced the same
environmental exposures. After completing GWAS for each frequency, a total of 9
independent loci were identified on 6 chromosomes. Of note, results for five of the
six frequencies contained a locus that was unique for that frequency, indicating that
different genes are active in different regions of the cochlea in response to certain
frequencies. Additionally, the Gm2447 locus on chromosome 3 was highly
suggestive or significant in analyses for five out of six frequencies, suggesting that
while each frequency is in some part controlled by unique genes, there are some
that appear to influence hearing across the full spectrum. Finally, in accordance
with data from the animals suggesting that ABR is more highly correlated between
frequencies nearer to each other on the spectrum than those on opposite ends,
some loci detected in GWAS appear only at higher or lower frequencies. Together,
these data suggest a highly complex network of genes, some with potentially a
more broad influence throughout the cochlear structure, while others are more
75
specific, that associate with hearing in the mouse. In this study, my role was to
assist in the development of the FaST‐LMM pipeline, perform GWAS and all
subsequent bioinformatics prioritization of positional candidates, and prepare and
write the manuscript. The manuscript for this study has been submitted to the
journal BMC Genomics and is awaiting review.
76
The genetic architecture of hearing impairment in mice: evidence for
frequency specific genetic determinants.

Amanda L. Crow
1
, Jeffrey Ohmen
2
, Juemei Wang
3
, Joel Lavinsky
3
, Jaana Hartiala
1
,
Qingzhong Li
4
, Xin Li
5
, Pezhman Dermanaki
3
, Eleazar Eskin
6
, Calvin Pan
7
, Aldons J.
Lusis
7
, Hooman Allayee
1
, Rick A. Friedman
3

1
Department of Preventive Medicine and Institute for Genetic Medicine, Keck
School of Medicine, University of Southern California, Los Angeles, CA 90033
2
House Ear Institute, Los Angeles, CA 90057
3
Department of Otolaryngology and Zilkha Neurogenetic Institute, Keck School of
Medicine, University of Southern California, Los Angeles, CA, 90033
4
Department of Otolaryngology ‐ Head and Neck Surgery, Eye & ENT Hospital of
Fudan University, Shanghai 200031, China
5
Clinical Laboratory Department, First Affiliated Hospital of Nanchang University,
Donghu District, Nanchang, Jiangxi Province 330006, China
6
Department of Computer Science and Inter‐Departmental Program in
Bioinformatics, University of California, Los Angeles, Los Angeles, CA 90095
7
Departments of Human Genetics, Medicine, and Microbiology, Immunology, and
Molecular Genetics, David Geffen School of Medicine at UCLA, Los Angeles, CA
90095

77
Abstract

Genome‐wide association studies (GWAS) have been successfully applied in
humans for the study of many complex phenotypes. However, identification of the
genetic determinants of hearing in adults has been hampered, in part, by the
relative inability to control for environmental factors that might affect hearing
throughout the lifetime, as well as a large degree of phenotypic heterogeneity.
These and other factors have limited the number of large‐scale studies performed
in humans that have identified candidate genes that contribute to the etiology of
this complex trait. In order to address these limitations, we performed a GWAS
analysis using a set of inbred mouse strains from the Hybrid Mouse Diversity Panel.
Among 99 strains characterized, we observed ~2 to 5‐fold variation in hearing at six
different frequencies, which are differentiated biologically from each other by the
location in the cochlea where each frequency is registered. Among all frequencies
tested, we identified a total of nine significant loci, several of which contained
promising candidate genes for follow‐up study. Taken together, our results indicate
the existence of both genes that affect global cochlear function, as well as
anatomical‐ ‐ and frequency‐specific genes, and further demonstrate the complex
nature of mammalian hearing variation.
78
Keywords
Genome‐wide association study (GWAS), Hybrid Mouse Diversity Panel (HMDP),
genetics, genomics, ABR, hearing, cochlear function
79
Background

Genome‐wide association studies (GWAS) have been successfully applied to
numerous complex traits in humans [1]. These GWAS leverage natural genetic
variation in an unbiased approach that is advantageous for studying complex traits.
We and others have shown that audiometric threshold elevation secondary to aging
in humans is continuously distributed throughout the population [2] and that the
human cochlea has a perceptual frequency spectrum of 20 Hz to 20 kHz. Heritability
studies have shown that the sources of this variance are both genetic and
environmental, with approximately half of the variance attributable to hereditary
factors [3]. As such, a major impediment to progress in the discovery of risk loci for
hearing in humans is the lack of control of the many environmental factors that
affect hearing during the lifetime, and, to date, only a limited number of large‐scale
GWAS for hearing phenotypes have been undertaken in humans. We recently
reported an association between age‐related hearing loss and SNPs within the
GRM7 gene and subsequently demonstrated, in a more comprehensive study, the
polygeneic nature of this disease [4]. Despite these successes, association studies in
humans – particularly for common forms of hearing loss and baseline hearing ability
– are still plagued by small sample sizes and phenotypic heterogeneity.
In order to harness the same investigative power of the GWAS approach in a
more controlled environment that addresses some limitations of human studies,
80
several groups have proposed mouse GWAS [5‐9]. For obvious reasons, mouse
models have several advantages over human studies. The environment can be more
carefully controlled, measurements can be replicated in genetically identical
animals, and the proportion of the variability explained by genetic variation is
increased. Complex traits in mouse strains have been shown to have higher
heritability and genetic loci often have stronger effects on the trait compared to
humans [10‐12]. Furthermore, several recently developed strategies for mouse
genetic studies, such as use of the Hybrid Mouse Diversity Panel (HMDP), provide
much higher resolution for associated loci than traditional approaches with
quantitative trait loci (QTL) mapping [5, 7]. The genetic and functional similarities of
the mouse and human ear coupled with the ability to control environmental
exposure make the mouse the ideal model for the study of age‐related hearing loss.
Both naturally occurring and experimentally induced mutations in mice have
provided much of our current understanding of many diseases of the ear. It has
long been our hypothesis that common forms of hearing impairment in both mice
and humans is a complex trait resulting from susceptibility loci throughout the
genome that contribute to the natural variation in hearing over time. Based upon
literature suggesting that most common disease phenotypes result from sequence
variation, often regulatory, in genes that are different from those underlying
Mendelian traits, e.g., non‐syndromic hearing loss, we undertook a study to
characterize the genetics of hearing by performing an association analysis that
exploits the natural genetic and phenotypic variation of hearing levels in inbred
81
strains of mice.

82
Results
Variation in ABR thresholds in the HMDP. As part of our effort to comprehensively
characterize the genetic basis of strain variation in hearing, we recorded ABR
threshold data in 5‐6 week‐old female mice from 29 common inbred (CI) and 70
recombinant inbred (RI) mouse strains from the HMDP. Figures 1a through 1f
demonstrate the results for ABR thresholds after tone burst stimuli of 4, 8, 12, 16,
24, and 32 kHz in the HMDP. Notably, among the 99 strains characterized, ABR
across the frequencies tested varied by 2 to 5‐fold, strongly suggesting a complex
underlying genetic architecture. Furthermore, ABR thresholds were highly
correlated, particularly among those across frequencies at the same range of the
spectrum (Supplemental Figure 1).

GWAS for ABR threshold variation at each tested frequency. Based on the
phenotypic data in the HMDP, we next sought to identify the genetic factors
affecting hearing by carrying out a GWAS for ABR thresholds at each tone burst
stimulus frequency. These analyses used an efficient mixed model algorithm
(EMMA) that takes into account the underlying population structure and genetic
relatedness of the HMDP strains [13] and has been used successfully for association
mapping in the HMDP. These GWAS analyses yielded several interesting
observations (Figures 2a‐2f.) First, we identified at least one locus for each
frequency that achieved genome‐wide significance, with a total of eight regions
83
distributed on chromosomes 3, 4, 9, 10, 13 and 19 (Table 1). Second, with the
exception of the 24 kHz stimulus, each analysis identified at least one locus that was
frequency specific. Third, a locus on chromosome 3 (rs30259360) that was
significantly associated with ABR threshold after a 12 kHz stimulus also
demonstrated suggestive association (p~10
‐5
) with nearly all other ABR traits (Table
1). Fourth, loci that were associated with ABR threshold across multiple frequencies
tended to cluster at the same end of the spectrum. For example, the chromosome
19 locus (rs30354441) that was associated with ABR threshold after a 4 kHz tone
burst also demonstrated association after the 8 kHz stimulus but not the other
frequencies. Conversely, a locus on chromosome 13 (rs52344209) was only
common to ABR threshold after 24 kHz and 32 kHz tone bursts. Lastly, one locus on
chromosome 10 (rs29362366), which was associated with ABR threshold only after
a 16 kHz stimulus, maps to Otogl, a gene that has been implicated in a Mendelian
form of human deafness.
Bioinformatics analyses to prioritize positional candidate genes. We next used
bioinformatics approaches to prioritize candidate genes at each locus. First, we
used publicly available expression data for several tissues in the HMDP
(http://geneeqtl.genetics.ucla.edu/) to determine whether variation at the
identified loci had cis‐acting effects on gene expression. Expression QTL (eQTL)
were considered local, or cis, if the lead SNP mapped within 2Mb of the peak
GWAS SNP. We identified four genes as having cis eQTL in the liver: three (Higd1a,
84
Ccbp2, and Ano10) are located in the chromosome 9 locus identified in the 4kHz
GWAS, and the fourth cis eQTL in liver was for Naip2, which is located within the
second chromosome 13 region that was identified in the 24 kHz GWAS (Table 2).
As we are interested in hearing, we wanted to interrogate if any of our
genes exhibit cis eQTL in a more relevant tissue, namely, the cochlea. Therefore,
we performed association analysis on microarray‐generated expression data in the
cochlea from a subset of 64 HMDP strains. Notably, several of our GWAS loci
exhibited cis eQTL in cochlear tissue (Table 2). For example, Prpf19, which maps
approximately 700kb upstream of our peak GWAS SNP for ABR thresholds after 4
kHz and 8 kHz stimuli, yielded a significant eQTL in the cochlea (rs30899404; p=
2.7x10
‐6
) but not in other tissues. Additionally, an eQTL for Ms4a6d (rs30768936;
p= 2.7x10
‐8
) was located ~160kb downstream of our peak GWAS SNP for the 12
kHz burst in a different region on chromosome 19 (Table 2). Lastly, Fbp1 had a
highly significant cis eQTL (rs13481847; p= 6.1x10
‐21
) in the cochlea that maps
~100kb upstream of our peak GWAS SNP on chromosome 13, for ABR threshold
after a 24 kHz tone burst. We also examined in situ imaging to visualize where in
the P1 mouse cochlea Prpf19 and Fbp1 are expressed (Figure 3), as expression in
different regions of the organ can indicate different biological functions with
respect to hearing. The sensory epithelium and spiral ganglion demonstrate Prpf19
mRNA expression (Figure 3A); further, Fbp1 mRNA expression was detected in the
basal layer of the striavascularis in P1 mouse cochleas (Figure 3B).
85
In addition to eQTL analysis, we also used the Ensembl genome browser to
determine whether any of the positional candidate genes harbored amino acid
substitutions among the HMDP strains. Of the candidate genes examined, Otogl
had an Asn2170Ser polymorphism that was predicted to have functionally
deleterious consequences by both SIFT (score=0.0) and PolyPhen2 (score=1.0).
Otogl is 87% identical to the human protein at the amino acid level and the
Asn2170Ser is located in a region of high homology that is evolutionarily across
seven mammalian species (Figure 4). Further analysis revealed that Otogl also
harbored 4 additional amino acid substitutions but these were either similar in their
physiochemical properties, computationally not predicted to have functional
consequences, or were located in regions of the protein that were not as
evolutionarily conserved as the region containing Asn2170Ser (Figure 4).

86
Discussion
We have used a GWAS approach with correction for population structure to
map several loci for hearing traits in inbred strains of mice. Our results identify a
number of novel loci and demonstrate two critical additions to the literature. First,
with the exception of the Otogl locus, which was identified in humans, none of the
loci identified in the GWAS with the HMDP overlap with Mendelian loci identified
for hearing loss in mice. This observation supports the notion that variation in
hearing among inbred mouse strains has a complex genetic architecture. Second,
just as the cochlea has a frequency specific functional place map, we demonstrate
that the genetics of hearing sensitivity is frequency specific as well.
We recently published the first meta‐analysis for age‐related hearing loss in
mice using data from several data sets, including a subset of the HMDP strains
(Ohmen et al. 2014). The meta‐analysis identified 5 significantly associated loci,
including Fscn2, the causative gene at the Ahl8 locus identified in a cross between
the DBA/2J and C57BL/6J strains. However, there was no overlap between the loci
detected in the meta‐analysis and those in the present study. For example, the
meta‐analysis contained only 35 of the 99 HMDP strains and predominantly
powered by the inclusion of N2 mice from a DBA/2J X C57BL/6J backcross
generated by Johnson et al (2008).   By contrast, the present study sought to
identify the genetic determinants for common forms of hearing with the entire
HMDP and would thus be more likely to detect allelic variation that is present
87
across all 99 strains. Lastly, our current study assessed hearing at 4 and 24 kHz tone
burst stimuli, in addition to those at 8, 16, and 32 kHz that were used in our prior
study.

There exists frequency specific genetic variation in hearing thresholds in inbred
strains of mice. One of the most important findings from our analysis is the
apparent existence of strong frequency‐specific genetic variation in hearing ability
across inbred mouse strains. Ruben demonstrated through an autoradiographic
study of the mouse cochlea that terminal mitoses in the developing cochlear
epithelium were first detected in the cochlear apex [14]. From this study it was
concluded that the differentiation of the cochlear organ of Corti occurred from the
base to the apex. The mammalian cochlea has numerous specializations throughout
its length that contribute to the tonotopic representation of perceived sound with
the basal portion responding to higher frequencies than the apex [15, 16]. Critical to
the frequency tuning of the cochlea are the morphological and
mechanotransductional properties of the hair cells, the basilar membrane, and the
spiral ganglion neurons along the length of the cochlear spiral. The molecular bases
for these tonotopical specifications along the longitudinal axis of the cochlea
remain largely unknown.
The initial observations made by Zheng et al. (1999) that different strains
exhibit different threshold sensitivities to particular stimulus frequencies were
88
poignant and were a preliminary insight into the strain variation and genetic
variation of sensitivities along the length of the cochlear duct. Son et al., (2012)
using microarray technology, also demonstrated gradients of gene expression in the
adult and developing mouse cochlea, a subset of which (roughly 5% of those genes
assayed) exhibited clear spatial and temporal variations in expression along the
cochlear duct. Utilizing an association‐based approach, we build on these prior
studies to elucidate the genetic architecture of hearing sensitivities along the length
of the cochlear duct in inbred strains of mice. For example, the finding of at least
one specific genome‐wide significant locus for each frequency in our
comprehensive analysis supports this notion. This was true for each tone‐burst
stimulus tested except 24 kHz. Additional support for frequency specific, or cochlear
regionally specific genetic determinants, comes from the identification of a locus for
the lower frequency apical cochlear spectrum on chromosome 19 (4 kHz and 8 kHz)
and of a region on chromosome 13 for the higher frequency basilar cochlear
spectrum (24 kHz and 32 kHz). Although a portion of this frequency‐specific genetic
association can be explained by issues related to study power, we also detected a
genome‐wide significant locus on chromosome 3 for the 12 kHz tone burst that
demonstrated suggestive association with nearly all other ABR stimulus traits
(p~10
‐5
).

Mendelian forms of hearing loss. To date, roughly 18 Mendelian loci for AHL have
89
been described in mice using traditional QTL mapping strategies
(http://hearingimpairment.jax.org/table2.html). None of the regions we identified
in the HMDP correspond to these Mendelian AHL loci. There may be several
reasons for these observations. First, we tested ABR in mice only at 5 weeks of age,
and many strains used in the previously reported AHL mapping studies do not
develop hearing loss until a later age; thus, we would not detect impaired hearing in
these strains at 5 weeks. Second, prior QTL studies were performed using F2 crosses
of two strains of mice, resulting in approximately 50% frequency of each allele in
the population of mice for mapping. With the HMDP, however, the effect allele may
not be frequent enough among the strains used to provide sufficient power with
the linear mixed model mapping algorithm. Interestingly, the locus on chromosome
10 (rs29362366) associated with ABR threshold after a 16 kHz stimulus corresponds
to OTOGL, a gene that has been implicated in a Mendelian form of moderate high
frequency non‐syndromic hearing loss in humans [17]. Furthermore, bioinformatics
analysis of SNP variation in susceptible strains shows a deleterious variant in a
highly conserved region of the Otogl ortholog, making it a strong candidate gene at
this locus.

Hearing impairment in mice: natural variation. Our GWAS for hearing phenotypes
in mice revealed nine statistically significant candidate loci; of these, bioinformatic
analysis suggests prioritizing at least four positional candidate genes for functional
90
validation. As described above, Otogl is a strong candidate for the 16 kHz locus and,
based upon cochlear eQTL analyses, Prpf19, Ms4a6d, and Fbp1 represent strong
candidates within their respective intervals. Prpf19 is a pre‐mRNA processing factor
that has been implicated in mammalian DNA repair [18] but has no known function
in the auditory system. Consistent with the eQTL results, RNAseq data
demonstrates a two‐fold greater expression level in cochlear hair cells compared to
nonsensory cells (Neil Segil, personal communication) and our in situ data indicates
expression in the sensory epithelium and spiral ganglion. Array data similarly
confirms two‐fold greater expression of Prpf19 mRNA in the spiral ganglion neurons
of mice in comparison to the vestibular ganglion beginning at E16 and extending
into adulthood (http://goodrichlabmicroarrays.hms.harvard.edu). Ms4a6d is a
member of a multigene four‐transmembrane family related to CD20, a
hematopoietic‐cell‐specific protein and a high affinity IgE receptor beta chain [19]
with no known function in the mammalian auditory system. Ms4a6d demonstrates
nearly four‐fold greater levels of expression in nonsensory cells of the inner ear and
similar expression levels in the spiral and vestibular ganglia that peak more than
two weeks postnatal, near the time of onset of hearing in mice (Neil Segil, personal
communication). Fbp1 has recently been demonstrated to localize within the cell
nucleus in a cell cycle‐dependent manner and may be involved in RNA processing,
nucleosome assembly, and cell cycle regulation. Like the other putative positional
genes described above, it has no known role in hearing in mammals [20]. RNAseq
data also revealed a two‐fold greater expression of this gene in nonsensory cells of
91
the inner ear and variable levels of expression within the spiral and vestibular
ganglia of mice during pre and postnatal periods); additionally, our in situ data
places Fbp1 mRNA in the striavascularis, which is critical to the maintenance of the
endocochlear potential. Although a specific function has not been ascribed to the
basal cell layer, there are widespread gap junctions between the basal cells and the
basal cells and intermediate cells. This suggests that these cells make up a
functional syncytium similar to that seen in the myocardium, supporting the notion
that this gene may play a role in hearing in mice. Taken together, these results
demonstrate that Prpf19, Ms4a6d, and Fbp1 are expressed in various cochlear cell
types and that the loci harboring these genes contain functional genetic variation
with respect to their expression. Additional functional studies will be required in
order to determine whether these genes play a role in mammalian hearing.
In conclusion, we have carried out a comprehensive analysis in mice to
elucidate the genetic architecture of hearing in response to tone burst stimuli. We
identified multiple novel loci that are for the most part frequency specific and
illustrate the complex nature of mammalian hearing. By combining systems
genetics with bioinformatics, we have also identified plausible positional candidate
genes at several loci that can be pursued for functional validation. Importantly, the
overlap between the effect of the Otogl locus on hearing loss in mice and non‐
syndromic hearing loss in humans reinforces the concept that the underlying
biological pathways are likely to be conserved among mammals. Furthermore, the
results of such genetic studies in mice can potentially be leveraged towards
92
development of novel prevention and treatment strategies.
93
Experimental Methods

Animals. Five week‐old female mice of 99 inbred strains (n=3‐8) from the HMDP
(Jackson Laboratories) were obtained and used in accordance with the University
of Southern California Institutional Animal Care and Use Committee (IACUC)
guidelines. Mice were housed in sterilized microisolator cages and received
autoclaved food and water ad libitum.
Hearing assessment. Mice were anesthetized with an intraperitoneal injection of
a mixture of ketamine (80 mg/kg body weight) and xylazine (16 mg/kg body
weight). Mouse body temperature was maintained through the use of a TCAT‐2DF
temperature controller and the HP‐4 M heating plate (Physitemp Instruments Inc.,
Clifton, NJ). Artificial tear ointment was applied to the eyes during anesthesia.
Lastly, mice recovered from anesthesia on a heating pad.
Auditory signals were presented as tone pips (4, 8, 16, 24, and 32 kHz) in the form
of a hamming wave with a 0.3‐ms rise and fall time (total time of 1 ms). These
signals were presented at a rate of 40 per second. They were then sent to an
amplifier and then to the sound transducer from Intelligent Hearing Systems.
Physiologic responses were recorded with a 20,000 analog‐to‐digital rate and sent
to an 8 channel 150‐gain AC/DC headbox and then onto a secondary Synamps signal
amplifier of 2500 gain before analysis. Filter settings were set at a low‐pass of 3000
94
Hz and a high‐pass of 100 Hz with an artifact rejection of signals with amplitudes
exceeding ± 50 μV. Three thousand waveforms were averaged at each stimulus
intensity.
Hearing assessment using auditory brainstem response. The ability of individual
mice to hear was assessed using an auditory brainstem response test (ABR).
Stainless‐steel electrodes were placed subcutaneously at the vertex of the head and
the left mastoid. A ground electrode was placed at the base of the tail. Test sounds
were presented using an Intelligent Hearing Systems speaker attached to an 8–in.
long tube that was inserted into the ear canal. Due to time and equipment
constraints, only the left ear was assessed. Tone bursts were first presented at a
high intensity to elicit a waveform. Next, the intensity was decreased by 20 dB until
nearing threshold. Intensity was then decreased in smaller steps of 10, 5, and 2 dB
as threshold was approached. Hearing threshold was determined by visual
inspection of ABR waveforms and was defined as the intensity at which two peaks
could be distinguished. Experiments were duplicated at low intensities when the
peaks were not apparent.
Expression Quantitative Trait Loci Analysis. We used previously generated
expression data from liver, etc (http://geneeqtl.genetics.ucla.edu/). For the cochlea,
we isolated cochlea from 64 HMDP strains (2‐4 mice per strain) mice at 6 weeks of
age and used RNeasy (Qiagen Valencia, CA) kit to isolate total RNA.
Association analysis. A GWAS for hearing in the HMDP strains was carried out
95
using SNP genotype data obtained from the Broad Institute
(www.broadinstitute.org/mouse/hapmap) and combined with genotypes from the
Wellcome Trust Center for Human Genetics (WTCHG). Genotypes of RI strains at
the Broad SNPs were inferred from the WTCHG genotypes by imputing alleles at
polymorphic SNPs among parental strains, with ambiguous genotypes labeled as
“missing.” Of the 623,000 SNPs available, 210,000 were informative with an allele
frequency greater than 10% and used in the present GWAS analysis.
We applied the following linear mixed model to account for the population
structure and genetic relatedness among strains: y=μ+xβ+u+e where μ represents
mean HSPC frequency, x represents the SNP effect, u represents random effects
due to genetic relatedness with Var(u) = σg2K and Var(e) = σe2I, where K
represents an identity‐by‐descent (IBD) kinship matrix across all genotypes. A
restricted maximum likelihood (REML) estimate of σg2 and σe2 were computed
using Efficient Mixed Model Association (EMMA) [13], and the association mapping
was performed based on the estimated variance component with a standard F test
to test β≠0. Genome‐wide significance threshold in the HMDP was determined by
the family‐wise error rate (FWER) as the probability of observing one or more false
positives across all SNPs per phenotype. We ran 100 different sets of permutation
tests and parametric bootstrapping of size 1,000 and observed that the genome‐
wide significance threshold at a FWER of 0.05 corresponded to a p‐value of 4.1x10
‐
6
, similar to what has been used in previous studies with the HMDP (Ghazalpour et
al, 2012). This is approximately an order of magnitude larger than the threshold
96
obtained by Bonferroni correction (4.6x10
‐7
), which would be an overly
conservative estimate of significance since nearby SNPs among inbred mouse
strains are highly correlated with each other.
97
Acknowledgements
This work was supported by California Institute for Regenerative Medicine grant
TG2‐01161 (ALC); NIH grants T32ES013678 (ALC), R01ES022282 (EE), P01HL30568
(AJL), P01HL28481 (AJL), R01ES021801 (HA), 3R01ES021801‐03S1 (HA), and
R01DC010856 (RAF); pilot project awards from the Southern California Clinical and
Translational Science Institute through NIH grant UL1TR000130 (HA) and the
Southern California Environmental Health Sciences Center through NIH grant
P30ES007048 (HA). The funders had no role in study design, data collection and
analysis, decision to publish, or preparation of the manuscript.
98
References
1. Welter, D., et al., The NHGRI GWAS Catalog, a curated resource of SNP‐trait
associations. Nucleic Acids Res, 2014. 42(Database issue): p. D1001‐6.
2. Friedman, R.A., et al., GRM7 variants confer susceptibility to age‐related
hearing impairment. Hum Mol Genet, 2009. 18(4): p. 785‐96.
3. Huang, Q. and J. Tang, Age‐related hearing loss or presbycusis. Eur Arch
Otorhinolaryngol, 2010. 267(8): p. 1179‐91.
4. Fransen, E., et al., Genome‐wide association analysis demonstrates the
highly polygenic character of age‐related hearing impairment. Eur J Hum
Genet, 2014.
5. Bennett, B.J., et al., A high‐resolution association mapping panel for the
dissection of complex traits in mice. Genome Res, 2010. 20(2): p. 281‐90.
6. Flint, J. and E. Eskin, Genome‐wide association studies in mice. Nat Rev
Genet, 2012. 13(11): p. 807‐17.
7. Ghazalpour, A., et al., Hybrid mouse diversity panel: a panel of inbred mouse
strains suitable for analysis of complex genetic traits. Mamm Genome, 2012.
23(9‐10): p. 680‐92.
8. Mott, R. and J. Flint, Dissecting quantitative traits in mice. Annu Rev
Genomics Hum Genet, 2013. 14: p. 421‐39.
9. Valdar, W., et al., Genome‐wide genetic association of complex traits in
heterogeneous stock mice. Nat Genet, 2006. 38(8): p. 879‐87.
10. Lindblad‐Toh, K., et al., Large‐scale discovery and genotyping of single‐
nucleotide polymorphisms in the mouse. Nat Genet, 2000. 24(4): p. 381‐6.
11. Wiltshire, T., et al., Genome‐wide single‐nucleotide polymorphism analysis
defines haplotype patterns in mouse. Proc Natl Acad Sci U S A, 2003. 100(6):
p. 3380‐5.
12. Yalcin, B., et al., Unexpected complexity in the haplotypes of commonly used
inbred strains of laboratory mice. Proc Natl Acad Sci U S A, 2004. 101(26): p.
9734‐9.
13. Kang, H.M., et al., Efficient control of population structure in model organism
association mapping. Genetics, 2008. 178(3): p. 1709‐23.
99
14. Ruben, R.J., Development of the inner ear of the mouse: a radioautographic
study of terminal mitoses. Acta Otolaryngol, 1967: p. Suppl 220:1‐44.
15. Davis, R.L., Gradients of neurotrophins, ion channels, and tuning in the
cochlea. Neuroscientist, 2003. 9(5): p. 311‐6.
16. Mann, Z.F. and M.W. Kelley, Development of tonotopy in the auditory
periphery. Hear Res, 2011. 276(1‐2): p. 2‐15.
17. Bonnet, C., et al., Biallelic nonsense mutations in the otogelin‐like gene
(OTOGL) in a child affected by mild to moderate hearing impairment. Gene,
2013. 527(2): p. 537‐40.
18. Mahajan, K.N. and B.S. Mitchell, Role of human Pso4 in mammalian DNA
repair and association with terminal deoxynucleotidyl transferase. Proc Natl
Acad Sci U S A, 2003. 100(19): p. 10746‐51.
19. Ishibashi, K., et al., Identification of a new multigene four‐transmembrane
family (MS4A) related to CD20, HTm4 and beta subunit of the high‐affinity
IgE receptor. Gene, 2001. 264(1): p. 87‐93.
20. Mamczur, P., et al., Cell cycle‐dependent expression and subcellular
localization of fructose 1,6‐bisphosphatase. Histochem Cell Biol, 2012.
137(1): p. 121‐36.

100
Figure Legends

Figure 1. Variation in ABR in the HMDP. The auditory brainstem response (ABR) to
tone burst stimuli at six different frequencies exhibits 2 to 5‐fold variation among
99 HMDP strains. Each dot represents an individual mouse from the respective
strain and the mean values are indicated by the horizontal black bars. Five week old
female mice were exposed to auditory signals at frequencies of 4 kHz, 8 kHz, 12 kHz,
16 kHz, 24 kHz, and 32 kHz. The ability of individual mice to hear these signals was
assessed using an ABR test. ABR is represented by the decibel level at which hearing
threshold was reached, determined visually by an ABR waveform.

Figure 2. Manhattan plots of GWAS results for ABR at each tone burst frequency.
(A) ABR at 4 kHz was significantly associated with a locus on chromosome 9 and a
region on chromosome 19. (B) A GWAS for the 8 kHz tone burst revealed two
significantly associated loci on chromosomes 10 and 19. (C) Three significantly
associated loci on chromosomes 3, 13, and 19 were identified in the GWAS analysis
for the 12 kHz tone burst. (D) One significant locus on chromosome 10 was
associated with ABR at the 16 kHz tone burst. (E) The GWAS for ABR at 24 kHz
identified two different significant loci on chromosome 13. (F) Two significant loci
on chromosomes 4 and 13 were associated with ABR at 32 kHz. For each
significantly associated locus, the gene(s) nearest to the peak SNP is indicated. The
GWAS analyses for each of the six tone burst stimuli included 182,270 SNPs, whose
101
genomic positions are shown along the x‐axis with their corresponding ‐log10 p‐
values indicated by the y‐axis. The genome‐wide thresholds for significant
(p=4.1x10‐6) and suggestive (p=4.1x10‐4) evidence of association are indicated by
the horizontal red and blue lines, respectively.
Figure 3. In situ images exhibiting cochlear mRNA expression of GWAS candidate
genes. (A) Antisense RNA probe for Prpf19 in P1 cochlea demonstrates expression
in the cochlear epithelium, represented by black circle, and spiral ganglion, denoted
by black arrow. Stria vascularis indicated by yellow star. (B) Antisense RNA probe for
Fbp1 in P1 cochlea demonstrates expression in the basal layer of the stria vascularis,
represented by black oval.
Figure 4. Amino acid sequence around predicted Asn2170Ser substitution. The
portion of the Otogl surrounding the predicted deleterious Asn2170Ser substitution
is shown from diverse vertebrate species. Physiochemically dissimilar amino acids in
the same interval are highlighted in red. Among mammalian species, the sequence
in this region is highly conserved and specifically the asparagine (N) residue is
conserved from Xenopus to human.
Supplemental Figure 1. Correlation of ABR between frequencies in the HMDP.
Panels A‐O demonstrate stronger correlations between mean strain ABR at similar
frequencies in the hearing spectrum. Lower (4 kHz‐8 kHz), mid‐range (8 kHz‐12 kHz‐
16 kHz), and higher (24 kHz‐32 kHz) frequencies correlate with each other at R >=
102
0.75, while comparisons between frequencies outside of their range show much
weaker relationships.

103
Table 1. Loci Identified in HMDP for ABR Thresholds at Various Tone Burst Frequencies.
Trait Chr Position (bp)
1
Lead SNP Candidate Gene(s) MAF p‐value
4kHz
3 52509617 rs30259360 Gm2447 0.50 1.3x10
‐5
9 122415312 rs33514183 Higd1a‐Ccbp2‐Ano10 0.39 9.9x10
‐7
19 10834438 rs30354441 Prfp19‐A430093F15Rik 0.06 1.1x10
‐7
8kHz
3 52509617 rs30259360 Gm2447 0.50 3.2x10
‐5
10 91728327 rs38135766 Gm20757 0.28 1.2x10
‐6
19 10834438 rs30354441 Prfp19‐A430093F15Rik 0.06 9.9x10
‐8
12kHz
3 52509617 rs30259360 Gm2447 0.50 1.4x10
‐6
13 63603934 rs51991909 Fbp1‐Fancc‐Ptch1 0.28 9.6x10
‐7
19 11217919 rs36642183 Ms4a6d 0.18 4.1x10
‐6
16kHz
10 107329794 rs29362366 Otogl/Gm6924 0.28 2.3x10
‐7
13 63391591 rs13481847 Fbp1‐Fancc‐Ptch1 0.33 4.3x10
‐5
24kHz
3 52509617 rs30259360 Gm2447 0.50 7.4x10
‐5
13 63391591 rs13481847 Fbp1‐Fancc‐Ptch1 0.33 2.2x10
‐6
13 101851306 rs52344209 Naip2 0.44 3.8x10
‐7
3 52509617 rs30259360 Gm2447 0.50 1.6x10
‐4
32kHz 4 85925264 rs28095334 Adamtsl1 0.22 9.7x10
‐7
13 101851306 rs52344209 Naip2 0.44 1.3x10
‐6
Chr, chromosome; MAF, minor allele frequency. BP: base pair position of lead SNP given, according to NCBI build 37 of the reference
mouse genome sequence. P‐values exceeding the genome‐wide significance threshold (4.1x10
‐6
) are shown in bold.
104
Table 2. cis eQTL for GWAS Loci at Various Tone Burst Frequencies.
Trait Chr Lead SNP BP Nearest Genes eQTL Lead SNP eQTL SNP Position eQTL p‐value eQTL Tissue
4kHz
9 rs33514183 122415312 Higd1a rs13459114 121825029 8.2x10‐20 liver
9 rs33514183 122415312 Ccbp2 rs6299531 122834341 2.0x10‐6 liver
9 rs33514183 122415312 Ano10 rs3713370 122186398 1.8x10‐11 liver
19 rs30354441 10834438 Prpf19 rs30899404 10148177 2.7x10‐6 cochlea
8kHz 19 rs30354441 10834438 Prpf19 rs30899404 10148177 2.7x10‐6 cochlea
12kHz 19 rs36642183 11217919 Ms4a6d rs30768936 11377515 2.7x10‐8 cochlea
24kHz
13 rs13481847 63391591 Fbp1 rs13481847 63290283 6.1x10‐21 cochlea
13 rs52344209 101851306 Naip2 rs3144793 100956146 2.4x10‐11 liver
Chr, chromosome; MAF, minor allele frequency. BP: base pair position of lead SNP given, according to NCBI build 37 of the
reference mouse genome sequence. Genomic regions ± 1Mb around the peak SNPs for each frequency were interrogated for the
presence of cis eQTLs in multiple tissues using the UCLA Systems Genetics Resource (http://systems.genetics.ucla.edu/) and
microarray data from cochlear tissue. Only genes exhibiting cis eQTLs in cochlea and liver are listed.
105
Figure 1
A
80
100
4 kHz
A/J
I/LnJ
C57L/J
BXD16/TyJ
BUB/BnJ
BXA24/PgnJ
AKR/J
LP/J
129X1/SvJ
AXB12/PgnJ
BXD84/Rw w J
NOR/LtJ
CE/J
BXH19/TyJ
BXH6/TyJ
CXB8/HiAJ
NOD/LtJ
SWR/J
BXD32/TyJ
AKXL17a/TyJ
BXD70/Rw w J
BXD55/Rw w J
C3H/HeJ
BALB/CJ
BXD11/TyJ
BXD50/Rw w J
SEA/GnJ
CXB12/HiAJ
BXA1/PgnJ
CBA/J
CXB9/HiAJ
BXD74/Rw w J
BXA16/PgnJ
AXB19a/PgnJ
BXD73/Rw w J
DBA/2J
SM/J
BXH10/TyJ
BXD75/Rw w J
PL/J
FVB/NJ
RIIIs /J
AXB15/PgnJ
MRL/MpJ
BXH22/KccJ
BXA4/PgnJ
AXB19/PgnJ
AXB6/PgnJ
BXA25/PgnJ
BXD12/TyJ
BXD28/TyJ
BXD8/TyJ
CXB11/HiAJ
LG/J
BXH14/TyJ
C57BL/6J
AXB10/PgnJ
AXB24/PgnJ
AXB8/PgnJ
BXA7/PgnJ
BXD1/TyJ
BXD20/TyJ
BXD21/TyJ
BXD6/TyJ
BXH8/TyJ
NON/LtJ
BXH4/TyJ
CXB13/HiAJ
AXB13/PgnJ
BXD34/TyJ
BXD9/TyJ
CXB1/ByJ
BXA14/PgnJ
BXH9/TyJ
SJL/J
BXD18/TyJ
BXD29/TyJ
BXD5/TyJ
AXB1/PgnJ
BXD14/TyJ
BXD15/TyJ
C57BLKS/J
CXB2/ByJ
MA/MyJ
BXD13/TyJ
BXD38/TyJ
BALB/cByJ
BXD2/TyJ
BXD42/TyJ
AXB5/PgnJ
BXA12/PgnJ
BXD31/TyJ
20
40
60
Strain
ABR ABR
B
A/J
BXD16/TyJ
NOR/LtJ
BUB/BnJ
C57L/J
LP/J
I/LnJ
BXA24/PgnJ
BXH6/TyJ
BXD84/Rw w J
CXB8/HiAJ
SWR/J
BXD11/TyJ
CE/J
NOD/LtJ
BXH19/TyJ
129X1/SvJ
BXD20/TyJ
DBA/2J
AKXL17a/TyJ
BXD32/TyJ
BXD55/Rw w J
BXD70/Rw w J
BTBR_T_tf/J
SEA/GnJ
BXD50/Rw w J
AXB12/PgnJ
BXD42/TyJ
BXD75/Rw w J
BXH8/TyJ
NZW/LacJ
AXB15/PgnJ
BXD5/TyJ
BXD73/Rw w J
CBA/J
BXD74/Rw w J
BALB/cByJ
BXA16/PgnJ
AKR/J
SJL/J
C3H/HeJ
AXB6/PgnJ
BXD21/TyJ
BXD38/TyJ
BXD6/TyJ
AXB10/PgnJ
CXB13/HiAJ
NZB/BlNJ
CXB12/HiAJ
AXB19a/PgnJ
BXA25/PgnJ
BXD12/TyJ
BXD8/TyJ
BXD9/TyJ
BALB/CJ
AXB19b/PgnJ
AXB24/PgnJ
AXB5/PgnJ
AXB8/PgnJ
BXA14/PgnJ
BXA4/PgnJ
C58/J
CXB11/HiAJ
RIIIs/J
C57BL/6J
SM/J
BXA1/PgnJ
BXH22/KccJ
BXD13/TyJ
BXD14/TyJ
BXD18/TyJ
BXD31/TyJ
NON/LtJ
BXH14/TyJ
AXB1/PgnJ
FVB/NJ
BXH9/TyJ
AXB13/PgnJ
BXD2/TyJ
CXB1/ByJ
PL/J
AXB19/PgnJ
BXA7/PgnJ
BXD29/TyJ
BXH10/TyJ
BXH4/TyJ
CXB2/ByJ
CXB9/HiAJ
KK/HlJ
LG/J
BXD1/TyJ
BXD28/TyJ
MA/MyJ
MRL/MpJ
BXA12/PgnJ
BXD15/TyJ
BXD34/TyJ
C57BLKS/J
A ABR
C
A/J
NOR/LtJ
BUB/BnJ
BXD16/TyJ
NOD/LtJ
BXA24/PgnJ
DBA/2J
LP/J
C57L/J
BXD20/TyJ
BXH19/TyJ
CXB8/HiAJ
BXH6/TyJ
129X1/SvJ
BXD32/TyJ
BXD84/Rw w J
I/LnJ
BTBR_T_tf/J
BXD55/Rw w J
BXD8/TyJ
CE/J
BXD70/Rw w J
AXB19a/PgnJ
BXA25/PgnJ
BXD75/Rw w J
AKR/J
NZW/LacJ
BXD73/Rw w J
AXB15/PgnJ
AXB5/PgnJ
BXD11/TyJ
BXD42/TyJ
BXD6/TyJ
NZB/BlNJ
BXD74/Rw w J
CXB13/HiAJ
SWR/J
CXB12/HiAJ
BXH8/TyJ
BXD9/TyJ
C58/J
MA/MyJ
C57BL/6J
BXA1/PgnJ
AKXL17a/TyJ
AXB19/PgnJ
BXD12/TyJ
BXD21/TyJ
CBA/J
C3H/HeJ
AXB19b/PgnJ
BXA16/PgnJ
BXD2/TyJ
BALB/CJ
BXA4/PgnJ
BXD50/Rw w J
BXH14/TyJ
AXB8/PgnJ
PL/J
SEA/GnJ
FVB/NJ
AXB10/PgnJ
AXB12/PgnJ
BALB/cByJ
BXD13/TyJ
BXD18/TyJ
BXD5/TyJ
CXB11/HiAJ
CXB9/HiAJ
LG/J
NON/LtJ
RIIIs/J
BXH22/KccJ
AXB1/PgnJ
BXA14/PgnJ
AXB13/PgnJ
BXD15/TyJ
BXD31/TyJ
BXD38/TyJ
CXB1/ByJ
KK/HlJ
MRL/MpJ
SJL/J
SM/J
BXH4/TyJ
BXH9/TyJ
BXH10/TyJ
BXA7/PgnJ
BXD1/TyJ
BXD14/TyJ
BXD29/TyJ
BXD34/TyJ
BXH7/TyJ
C57BLKS/J
AXB24/PgnJ
AXB6/PgnJ
BXA12/PgnJ
BXD28/TyJ
CXB2/ByJ
106
Figure 1
D
A/J
NOR/LtJ
NOD/LtJ
BXD16/TyJ
BUB/BnJ
BXD32/TyJ
DBA/2J
BXD42/TyJ
BXD21/TyJ
BXA24/PgnJ
LP/J
BXD12/TyJ
I/LnJ
BXH19/TyJ
BXD11/TyJ
C57L/J
AXB5/PgnJ
BXD20/TyJ
BXH6/TyJ
CE/J
CXB8/HiAJ
BXH8/TyJ
129X1/SvJ
AXB19/PgnJ
BXD84/Rw w J
AXB19a/PgnJ
BXA1/PgnJ
BXA25/PgnJ
BXD6/TyJ
SWR/J
BXD55/Rw w J
BXD73/Rw w J
AXB15/PgnJ
CXB13/HiAJ
SEA/GnJ
AXB19b/PgnJ
BXD75/Rw w J
FVB/NJ
MA/MyJ
BXD9/TyJ
PL/J
C3H/He J
CXB12/HiAJ
BTBR_T_tf/J
BXD50/Rw w J
RIIIs/J
BXD74/Rw w J
BXD70/Rw w J
BXD38/TyJ
MRL/MpJ
CXB11/HiAJ
LG/J
CBA/J
BALB/cByJ
BXA16/PgnJ
BXD15/TyJ
C58/J
CXB1/ByJ
BXH22/KccJ
CXB2/ByJ
C57BL/6J
NZW/LacJ
SM/J
BXH4/TyJ
BXA4/PgnJ
AKR/J
AKXL17a/TyJ
AXB10/PgnJ
AXB12/PgnJ
BALB/CJ
BXA14/PgnJ
BXD13/TyJ
BXD18/TyJ
NZB/BlNJ
BXH14/TyJ
BXA12/PgnJ
BXH9/TyJ
AXB13/PgnJ
AXB24/PgnJ
BXD28/TyJ
BXD29/TyJ
BXD5/TyJ
BXD8/TyJ
BXH10/TyJ
CXB9/HiAJ
NON/LtJ
AXB8/PgnJ
AXB6/PgnJ
BXA7/PgnJ
BXD1/TyJ
BXD14/TyJ
BXD2/TyJ
BXD34/TyJ
BXH7/TyJ
C57BLKS/J
KK/HlJ
SJL/J
AXB1/PgnJ
BXD31/TyJ
ABR ABR
E
A/J
BXD32/TyJ
NOR/LtJ
NOD/LtJ
AXB19a/PgnJ
DBA/2J
AXB15/PgnJ
AXB19/PgnJ
BXD38/TyJ
BXD12/TyJ
AXB19b/PgnJ
BXD20/TyJ
AXB5/PgnJ
BUB/BnJ
MA/MyJ
BXD16/TyJ
BXH19/TyJ
BXD42/TyJ
BXD11/TyJ
BXD8/TyJ
LP/J
BXA25/PgnJ
BXD2/TyJ
129X1/SvJ
BXD34/TyJ
CE/J
C57L/J
BXA24/PgnJ
BXD55/Rw w J
I/LnJ
BXH6/TyJ
BXD73/Rw w J
BXD21/TyJ
BXA1/PgnJ
AXB12/PgnJ
BXD15/TyJ
BXD84/Rw w J
BXH10/TyJ
C57BLKS/J
BXD50/Rw w J
BXA16/PgnJ
BXD14/TyJ
BXD9/TyJ
C58/J
CXB8/HiAJ
SEA/GnJ
BXA4/PgnJ
BXD1/TyJ
BXH8/TyJ
BALB/cByJ
BXD5/TyJ
AXB13/PgnJ
CXB9/HiAJ
BXD29/TyJ
BXA7/PgnJ
AKXL17a/TyJ
CXB13/HiAJ
LG/J
BXD70/Rw w J
BTBR_T_tf/J
BXD18/TyJ
BXD6/TyJ
BXH22/KccJ
NZW/LacJ
AKR/J
BXD75/Rw w J
SWR/J
CXB1/ByJ
CXB12/HiAJ
BXH4/TyJ
RIIIs/J
BXA14/PgnJ
BALB/CJ
AXB10/PgnJ
BXA12/PgnJ
BXD74/Rw w J
CXB2/ByJ
C57BL/6J
SJL/J
C3H/He J
AXB1/PgnJ
BXD31/TyJ
CXB11/HiAJ
NON/LtJ
PL/J
BXH9/TyJ
NZB/BlNJ
BXH14/TyJ
AXB6/PgnJ
FVB/NJ
MRL/MpJ
CBA/J
AXB8/PgnJ
AXB24/PgnJ
BXD13/TyJ
BXH7/TyJ
SM/J
BXD28/TyJ
KK/HlJ
A ABR
F
A/J
NOR/LtJ
BXD32/TyJ
NOD/LtJ
AXB15/PgnJ
AXB19a/PgnJ
BXA25/PgnJ
AXB5/PgnJ
BXD16/TyJ
AXB19/PgnJ
AXB19b/PgnJ
BXD20/TyJ
BXD8/TyJ
DBA/2J
BXH19/TyJ
129X1/SvJ
BXD12/TyJ
LP/J
AKR/J
BUB/BnJ
BXD34/TyJ
MA/MyJ
BXD11/TyJ
BXD2/TyJ
BXD84/RwwJ
CXB9/HiAJ
BXD55/RwwJ
C57BLKS/J
AXB13/PgnJ
BXD29/TyJ
BXD73/RwwJ
BXD42/TyJ
BXD9/TyJ
C57L/J
BXD21/TyJ
BXD50/RwwJ
BXA4/PgnJ
BXD15/TyJ
BXD38/TyJ
BXH6/TyJ
CXB13/HiAJ
C58/J
BXD6/TyJ
BXH10/TyJ
BXA1/PgnJ
BXA24/PgnJ
BXD1/TyJ
CE/J
BXD14/TyJ
BXD5/TyJ
BXA16/PgnJ
AXB12/PgnJ
CXB8/HiAJ
AXB10/PgnJ
BXD18/TyJ
I/LnJ
BXD75/RwwJ
SEA/GnJ
BXD70/RwwJ
BXD28/TyJ
BXA14/PgnJ
AKXL17a/TyJ
AXB8/PgnJ
BTBR_T_tf/J
BXA7/PgnJ
BXH8/TyJ
RIIIs/J
MRL/MpJ
NZW/LacJ
BXA12/PgnJ
BALB/cByJ
CXB2/ByJ
LG/J
AXB1/PgnJ
BXH4/TyJ
BXD74/RwwJ
CXB11/HiAJ
SWR/J
C57BL/6J
BXH22/KccJ
AXB6/PgnJ
BXD13/TyJ
NON/LtJ
SJL/J
CXB1/ByJ
FVB/NJ
NZB/BlNJ
PL/J
CXB12/HiAJ
BXH14/TyJ
BALB/CJ
AXB24/PgnJ
BXD31/TyJ
SM/J
BXH9/TyJ
KK/HlJ
BXH7/TyJ
C3H/HeJ
CBA/J
107
A
4 kHz
Higd1a‐Ccbp2 ‐Ano10
Prpf19 ‐A430093F15Rik
Figure 2
B
8 kHz
Prpf19 ‐A430093F15Rik
Gm20757
C
12 kHz
Gm2447
Fbp1 ‐Fancc ‐Ptch1
Ms4a6d
108
D
16 kHz
Otogl/Gm6924
Figure 2
E
24 kHz
Fbp1 ‐Fancc ‐Ptch1
Naip2
F
32 kHz
Adamts11
Naip2
109
Figure 3
A
B
110
Figure 4
Human (homo sapiens) S D Y G CCG T C K N VSCK F H M E NGTS V VY
Chimp (Pan troglodytes) S D Y D C C G T C K N VSCK F H M E NGTS V VY
Dog (Canis familiaris) S D Y D C C G T C K N VSCK F Q M E NGTS V I Y
Cow (Bos taurus) S D Y D C C G T C K N ISCKFQMENGTTV IY
Mouse (Mus musculus) S D Y D C C G T C K N I S C K F I M E N G T S V I Y Mouse (Mus musculus) S D Y D C C G T C K N I S C K F I M E N G T S V I Y
Rat (Rattus norvegicus) S D Y D C C G T C K N ISCKFHMENGTSV IY
Chicken (Gallus gallus) S H H S CCG T C Q N VSC SFL T EN GT R IVY
Frog (Xenopus tropicalis) SQQ M CCG T C K N VSC S F YS DN GT L VL Y
111
R = 0.78
p < .0001
R = 0.66
p < .0001
Supplemental Figure 1
AB
(8kHz)
12kHz)
R = 0.53 R = 0.26
C D
ABR (4kHz)
ABR (4kHz)
ABR
ABR (
p < .0001 p = .01 C D
ABR (16kHz)
ABR (24kHz)
R = 0.26
p = .01
E
ABR (4kHz) ABR (4kHz)
R (32kHz)
ABR (4kHz)
ABR
112
R = 0.87
p < .0001
R = 0.75
p < .0001
FG
Supplemental Figure 1
12kHz)
16kHz)
ABR (8kHz) ABR (8kHz)
ABR (1
ABR (1
R = 0.44
0001
R = 0.36
0002
HI
ABR (24kHz)
ABR (32kHz)
p < .0001 p = .0002
ABR (8kHz) ABR (8kHz)
113
R = 0.87
p < .0001
R = 0.6
p < .0001
JK
Supplemental Figure 1
6kHz)
24kHz)
L M
ABR (12kHz) ABR (12kHz)
ABR (16
ABR (2
R = 0.52
p < .0001
R = 0.72
p < .0001
L M
ABR (24kHz)
ABR (32kHz)
NO
ABR (12kHz) ABR (16kHz)
(32kHz)
(32kHz)
R = 0.57
p < .0001
R = 0.87
p < .0001
ABR (16kHz) ABR (24kHz)
ABR
ABR
114
Supplemental Figure 2
A
4 kHz –Chromosome 9
B
4 kHz –Chromosome 19
B
115
Supplemental Figure 2
C
8 kHz –Chromosome 10
D
8 kHz –Chromosome 19
D
116
Supplemental Figure 2
E
12 kHz –Chromosome 13
F
12 kHz –Chromosome 19
F
117
Supplemental Figure 2
G
16 kHz –Chromosome 10
118
Supplemental Figure 2
H
24 kHz –Chromosome 13
I
24 kHz –Chromosome 13
I
119
Supplemental Figure 2
J
32 kHz –Chromosome 4
K
32 kHz –Chromosome 13
K
120
CHAPTER 3: GENOME‐WIDE ASSOCIATION STUDY IDENTIFIES NOX3 AS A
CRITICAL GENE FOR SUSCEPTIBILITY TO NOISE‐INDUCED HEARING LOSS

Summary: In the present study, a preliminary GWAS using the FaST‐LMM EMMA
algorithm in 64 strains identified a candidate locus on chromosome 17 for noise‐
induced hearing loss (NIHL.) This study represents an additional key advantage
of genome‐wide analysis in the HMDP, which is that the environmental variable
can be exactly measured and identically delivered to each animal, ensuring that
any resulting change in phenotype is strictly from differences in genetic
interaction with that environmental variable. The chromosome 17 locus that
was implicated in the GWAS for NIHL at 8kHz was not significant in the GWAS for
baseline hearing at this frequency (Chapter 2), indicating that positional
candidates in this region are associated with ABR as a direct consequence of
noise exposure. The peak SNP at the significant locus centers over Nox3, NADPH
oxidase‐3, which has been previously shown to be highly expressed in the inner
ear. Compared to wild‐type mice, Nox3
het
and Nox3
‐/‐
mice both demonstrate
hearing impairment after noise insult, suggesting that the gene product of Nox3
may be protective against cochlear damage. My role in this study included
assistance on the development of the FaST‐LMM analysis and subsequent
execution of GWAS using this method. Additionally, I performed bioinformatic
analyses to prioritize Nox3 as the best likely candidate for functional followup,
and I co‐wrote the manuscript. This manuscript has been published in PLOS
Genetics (PLoS Genet 11(4): e1005094. doi:10.1371/journal.pgen.1005094).
121
Genome‐wide association study identifies Nox3 as a critical gene
for susceptibility to noise‐induced hearing loss.

Joel Lavinsky
1,2
, Amanda L. Crow
3
, Calvin Pan
4
, Juemei Wang
2
, Ksenia A. Aaron
2
,
Maria K. Ho
2
, Qingzhong Li
2
, Pehzman Salehide
2
, Anthony Myint
2
, Maya Monges‐
Hernadez
2
, Eleazar Eskin
5
, Hooman Allayee
3
, Aldons J. Lusis
3,6
, Rick A. Friedman
2

1
Graduate Program in Surgical Sciences, Federal University of Rio Grande do Sul,
Porto Alegre, Rio Grande do Sul, Brazil
2
Department of Otolaryngology, Zilkha Neurogenetic Institute, USC Keck School
of Medicine, University of Southern California, Los Angeles, California, United
States of America
3
Department of Preventive Medicine and Institute for Genetic Medicine, USC
Keck School of Medicine, University of Southern California, Los Angeles,
California, United States of America
4
Department of Human Genetics, University of California, Los Angeles, Los
Angeles, California, United States of America
5
Department of Computer Science, University of California, Los Angeles, Los
Angeles, California, United States of America
6
Department of Microbiology, Immunology, and Molecular Genetics, University
of California, Los Angeles, Los Angeles, California, United States of America
122
Abstract
In the United States, roughly 10% of the population is exposed daily to
hazardous levels of noise in the workplace. Twin studies estimate heritability for
noise‐induced hearing loss (NIHL) of approximately 36% and strain specific
variation in sensitivity has been demonstrated in mice. Based upon the
difficulties inherent to the study of NIHL in humans we have turned to the study
of this complex trait in mice. We exposed 5 week‐old mice from the Hybrid
Mouse Diversity Panel (HMDP) to a 10 kHz octave band noise at 108 dB for 2
hours and assessed the permanent threshold shift 2 weeks post exposure using
frequency specific stimuli. These data were then used in a genome‐wide
association study (GWAS) using the Efficient Mixed Model Analysis (EMMA) to
control for population structure. In this manuscript we describe our GWAS, with
an emphasis on a significant peak for susceptibility to NIHL on chromosome 17
within a haplotype block containing NADPH oxidase‐3 (Nox3). Our peak was
detected after an 8 kHz tone burst stimulus. Nox3 mutants and heterozygotes
were then tested to validate our GWAS. The mutants and heterozygotes
demonstrated a greater susceptibility to NIHL specifically at 8 kHz both on
measures of distortion product otoacoustic emissions (DPOAE) and on auditory
brainstem response (ABR). We demonstrate that this sensitivity resides within
the synaptic ribbons of the cochlea in the mutant animals specifically at 8 kHz.
Our work is the first GWAS for NIHL in mice and elucidates the power of our
approach to identify tonotopic genetic susceptibility to NIHL.

123
Introduction
Noise‐induced hearing loss (NIHL) is a worldwide leading occupational
health risk in industrialized countries and is the second most common form of
sensorineural hearing impairment, after presbyacusis [1]. In the United States,
roughly 10% of the total population is exposed daily to hazardous levels of noise
in the workplace [2]. The most extreme workplace environment for NIHL is the
Armed Forces. According to the Department of Veterans Affairs, hearing loss is
the most common disability among U.S. troops in the Middle East. The financial
impact of these disability claims on the VA is staggering and likely will continue
to grow. According to the American Tinnitus Association (http://www.ata.org/),
the number of disability claims from hearing injury is expected to increase by
18% per year with a total cost of $1.2 billion annually [3]. Risk could be reduced
with a better understanding of the biological processes that modulate
susceptibility to damaging noise. It is believed that NIHL is a complex disease
resulting from the interaction between environmental and genetic factors and it
is well recognized that people with similar exposures to noise show variation in
the amount of hearing loss, indicative of a genetic component [4].Twin studies
estimate heritability for noise‐induced hearing loss (NIHL) of approximately 36%
[5].
The discovery of gene by environment interactions in human disease, such
as susceptibility to NIHL, has many inherent difficulties, most notably,
controlling for exposure. Although several candidate gene association studies
for NIHL in humans have been conducted, each is underpowered, un‐replicated,
124
and accounts for only a fraction of the genetic risk. In addition, no heritability
studies have been performed, since families, where all subjects are exposed to
identical noise conditions, are almost impossible to collect.
The genetic basis of NIHL has been clearly demonstrated in animals as
different susceptibilities to noise have been seen in different inbred stains of
mice [4]. Mouse strains (C57BL/6J) exhibiting age‐related hearing loss (AHL)
were shown to be more susceptible to noise than other strains [6]. Also, several
knockout mice including SOD1‐/‐ [7], GPX1‐/‐ [8], PMCA2‐/‐ [9] and CDH23+/‐
[10] were shown to be more sensitive to noise than their wild‐type littermates.
The mouse has been an essential animal model for studies in hearing loss, and
advances in mouse genetics, including genome sequence and high density
single‐nucleotide polymorphism (SNP) maps, provide a suitable system for the
study of a complex trait such as NIHL [6]. The identification of novel genes is
crucial for the discovery of new pathways and gene networks that will improve
our knowledge of basic hearing biology and identify new therapeutic targets
with the potential to combat NIHL.
  Due to the limitations of human genome‐wide association study (GWAS) and
quantitative trait locus (QTL) analyses in mice, we have chosen to use a genome‐
wide association strategy incorporating the Hybrid Mouse Diversity Panel
(HMDP). The HMDP is a collection of classical inbred (CI) and recombinant
inbred (RI) strains whose genomes have been sequenced and/or genotyped at
high resolution [11]. Power calculations have demonstrated that this panel is
superior to traditional linkage analysis and is capable of detecting loci
125
responsible for 5% of the overall variance. Several studies have successfully
mapped candidate loci for complex traits using this panel and we have recently
published a meta‐analysis for age‐related hearing loss incorporating the HMDP
[12] [13] [14] [15].
  In this manuscript we describe, for the first time, an association analysis
with correction for population structure in the mapping of several loci for
susceptibility to NIHL in inbred strains of mice. After completing a preliminary
screen of the HMDP, an intriguing locus appeared warranting further
exploration. Herein, we describe a genome‐wide significant peak on (Chr.) 17
within a haplotype block containing NADPH oxidase‐3 (Nox3) and provide
evidence supporting its role in susceptibility to NIHL. Furthermore, we
demonstrate frequency‐specific genetic susceptibility within the mouse cochlea.

126

Results
There exists phenotypic variation in susceptibility to NIHL within the HMDP. In
an effort to identify genomic regions associated with NIHL susceptibility, we
phenotyped 5‐week old female mice (n=297) from 64 HMDP strains (n= 4‐
5/strain) for thresholds after noise exposure using Auditory Brainstem Response
thresholds at specific ABR stimulus frequencies. The stimuli consisted of 4, 8, 12,
16, 24 and 32 kHz tone bursts. A wide range of ABR thresholds were observed
across the HMDP with differences of 3.22‐fold between the lowest and the
highest strains for thresholds at 8 kHz post‐noise exposure (Fig 1). Frequencies
of 4, 12, 16, 24 and 32 kHz demonstrated differences of 1.55, 3.25, 3.57, 2.74
and 3.75‐fold, respectively.

Genome‐wide association analysis of NIHL reveals frequency specific genetic
susceptibility. EMMA algorithm was applied to each phenotype separately to
identify genetic associations for the six tone‐burst stimuli [18]. Adjusted
association p‐values were calculated for 108,064 SNPs with minor allele
frequency of > 5% (p < 0.05 genome‐wide equivalent for GWA using EMMA in
the HMDP is p=4.1 x 10
‐6
,
‐log10P=5.39). At this threshold, genome‐wide significant associations on Chr. 2
(rs27972902; p=8.6x10
‐7
) and Chr. 17 (rs33652818; p=2.3x10
‐6
) were identified
for the 8 kHz stimuli (Table 1, Fig 2). Additionally, a significant association signal
on Chr. 15 (rs32934144; p=1.7x10
‐6
) was identified for the 16 kHz tone burst and
127
two significant regions on Chr. 3 (rs30795209; p=5.5x10
‐7
) and Chr. 15
(rs32278602; p=5.9x10
‐7
) were identified at 32 kHz.

Characterization of NIHL GWAS peaks. Within each association peak there were
4 (Chr. 15), 11 (Chr. 3), 10 (Chr. 17) and 2 (Chr. 2) unique RefSeq genes. We next
identified genes within each of the five intervals possessing functional
alterations. Genes were selected based upon their regulation by a local
expression QTL (eQTL) in the HMDP or if they harbored a non‐synonymous (NS)
SNP that was predicted to have functional consequences. For the eQTL analysis,
we generated gene expression microarray profiles using RNA isolated from
cochleae in 64 HMDP strains (n =3 arrays per strain). EMMA was then used to
perform an association analysis between all SNPs and array probes mapping
within each region. A total of 18,138 genes were represented by at least one
probe, after excluding probes that overlapped SNPs, present among the classical
inbred strains used in the HMDP (see Methods). Of these, 6 genes (4 within Chr.
3 association and 2 within Chr. 17 association) were identified with at least one
probe whose expression was regulated by a local eQTL (Table 2). However, the
only probe whose expression was regulated by a significant local eQTL in the
cochlea was located on Chr. 17.
We determined whether any of the 27 genes implicated in our
preliminary GWAS had a defined role in the inner ear. The associations on Chr.
2, 3 and 15 did not harbor known cochlear genes. Only NADPH oxidase 3 (Nox3)
on Chr. 17 had been implicated in inner ear biology with mutants lacking
128
otoconia in the utricular and saccular maculae [22] and its high expression in the
inner ear [23].

Detailed analysis of the Chr 17 association highlights Nox3 as a candidate
gene. Of all genes at the chromosome 17 locus, one gene, Tfb1m, had a
significant (1.08x10
‐6
) eQTL (Fig 3). Of note, Nox3, the gene in which our peak
GWAS SNP is located, does not have an eQTL in the cochlea; however, there was
a clear demonstration [23] that Nox3 is highly expressed (at least 50‐fold higher
than in any other tissues) in specific portions of the inner ear. Based on these
data and the location of our peak GWAS SNP (rs33652818), we focused on Nox3
as a plausible candidate gene for NIHL at the chromosome 17 locus.

Nox3
het
mice are more susceptible to NIHL. To directly test the hypothesis that
Nox3 was associated with susceptibility to NIHL, we characterized previously
generated Nox3
het
mice for pre‐ and post‐noise exposure ABR thresholds and
PTS after 4, 8, 12, 16, 24 and 32 kHz tone‐burst stimuli. Consistent with our
original GWAS finding, this analysis revealed a statistically significant reduction
in the PTS in wild‐type mice (C57BL/6JEiJ strain) compared to Nox3
het
/
+
and
Nox3
het
/Nox3
het
at 8 kHz (Fig 4). As a comparison, the effects of the peak SNP
(rs33652818) at the Nox3 locus on ABR at various frequencies is shown in Fig 5.
Interestingly, there were significant differences as a function of genotype at
both the 4 kHz and the 8 kHz test frequencies, although the level of significance
at 4 kHz (p=1.1x10
‐4
) is only suggestive (Fig S1) and does not reach genome‐wide
129
significance (Table 1). Thus, the significant and highly suggestive association of
rs33652818 with ABR at 8 and 4 kHz, respectively, in the HMDP, as well as the
frequency‐specific phenotype exhibited by the Nox3
het
/Nox3
het
mice, suggests
that Nox3 may be involved in NIHL at the lower end of the frequency spectrum.
For a detailed analysis of the entire auditory pathway, we next evaluated
outer hair cell (OHC) activity using DPOAE and the inner hair cell (IHC) and
neuronal responses by ABR wave I peak‐to‐peak amplitudes. Despite the
absence of a statistically significant difference in DPOAE thresholds (Fig 6A) at 8,
16, 22 and 32 kHz, there was a pronounced difference at 8 kHz in the wave 1
ABR peak‐to‐peak amplitudes (Fig 6B).
The DPOAE (Fig 7A) suprathreshold amplitudes (dB SPL) and ABR wave 1
amplitudes (µV) (Fig 7B) for the 8 kHz tone burst were compared at different
stimulus intensities. Both analyses demonstrated statistically significantly less
noise damage in the wild‐type in comparison to the heterozygous and mutant
mice.
To confirm these electrophysiological findings, we collected cochleae
from pre‐ and post‐noise exposure Nox3
het
mice and wild‐type. First, we
assessed OHC loss throughout the entire cochlea by creating a cytocochleogram
(Fig 8A) of immunolabeled (Fig 8B) whole‐mount organs of Corti to correlate
with the DPOAE findings. Subsequently, the IHC afferent synaptic density (Fig 9)
was analyzed as a marker of the neuronal responses (suprathreshold ABR wave
1 amplitude).
130
Despite the absence of a statistical significance in OHC loss, the Nox3
het
/+
and Nox3
het
/Nox3
het
mice demonstrated a significantly reduced post‐noise
exposure density of synaptic ribbons (at the 8kHz tonotopic location).

131
Discussion
NIHL Genome‐wide Association Study We have, for the first time, used
association analysis with correction for population structure to map several loci
for hearing traits in inbred strains of mice. Our results identify a number of
novel loci for susceptibility to NIHL. Additionally, our study demonstrates
frequency‐specific genetic susceptibilities to noise within the cochlea and the
power of our GWAS to detect frequency‐specific loci that are precisely
recapitulated in a mutant mouse model.
Mouse GWAS has revolutionized the field of genetics and has lead to the
discovery of hundreds of genes that are involved in complex traits [24]. Our
successful mapping largely came from the initial observation that there was a
clear strain variation at all post noise exposure hearing phenotypes, reiterating
the contribution of genetic factors to NIHL susceptibility. This wide distribution
of phenotypes and genotypes facilitated our high‐resolution genetic mapping.
We used a combined set of 64 classic inbred and recombinant inbred
strains, a portion of the HMDP, as an extension of the classical inbred strain
association. This increased the statistical power of the classical association
studies by including a set of recombinant inbred strains in the mapping panel
[25]. The HMDP provided significant statistical power and resolution to identify
a locus for NIHL susceptibility that was precisely modeled in a mutant strain
[26]. Although this panel is composed of 100 commercially available inbred
strains, with roughly two‐thirds of this panel we were able to map 5 loci,
reflecting the power to detect loci with moderate effect. In addition to the
132
power present in this resource, the resolution of this panel is, in some cases,
two orders of magnitude better than that achieved with linkage analysis, as we
have recently demonstrated in our mouse GWAS for age‐related hearing loss
[27].
In an unprecedented manner, this new paradigm was applied to the first
high‐resolution mapping of candidate genes for NIHL susceptibility. Our GWAS
generated significant associations in at least five loci at three different post‐
noise exposure stimulus frequencies, corresponding to a total 27 candidate
genes. All of these candidate genes require adequate characterization, but the
first gene to be validated by a genetic mutant mouse model was Nox3. Nox3 was
selected for further investigation based upon its relatively restricted expression
in the cochleo‐vestibular epithelium and spiral ganglion neurons [23].

Nox3 and the Inner Ear. The Nox3 gene was described in 2000 based upon its
sequence similarity to other Nox isoforms (encodes an NADPH oxidase) [28]. The
overall structure of Nox3 is highly similar to that of Nox1 and Nox2 [29] and
Nox3 shares 56% amino acid with Nox2 [30]. Encoded by Nox3, the six‐
transmembrane NADPH‐binding protein interacts with a two‐transmembrane
protein (encoded by Cyba) and a cytosolic protein (encoded by Noxo1). This
activation releases a functional NADPH oxidase complex that is able to
transporting electrons across membranes towards oxygen (O2) generating
superoxide (O2•‐) and subsequent reactive oxygen species (ROS) [19].
133
First studies on the Nox3 function were published in 2004 and generated
the definition of Nox3 as an NADPH oxidase of the inner ear [23][22]. Banfi, et
al., performed analysis of Nox3 distribution (real time PCR and in situ
hybridization) and reported high Nox3 expression in the inner ear
(cochlear/vestibular sensory epithelia and the spiral ganglion). Following
exposure to cisplatin, HEK293 cells transfected with Nox3 produced O2•‐
spontaneously and generated a dramatic increase in O2•‐ production [23].
Paffenholz et al. [22] reported that mutations of the het locus affect Nox3 and
that these head tilt mice (het) have impaired otoconial formation in the utricle
and saccule resulting in balance defects, such as the inability to detect linear
acceleration or gravity. Based upon this finding we chose to pursue
interrogation of Nox3, a gene within our locus on Chr. 17.
Subsequent studies have established a role for the Nox3 gene as the
primary source of ROS generation in the cochlea, especially induced by cisplatin
ototoxicity [31]. The knockdown of Nox3 (pretreatment with siRNA) prevented
cisplatin ototoxicity with preservation of hearing thresholds and hair cells. Also,
it reduced the expression of Nox3 and biomarkers of damage (TRPV1 and KIM‐1)
in cochlear tissues [32]. siRNA‐mediated gene silencing of Nox3 alleviated
cisplatin‐induced hearing loss in rats and reduced apoptosis of the sensory hair
cells in the cochlea [33]. Although there was no similar evidence regarding NIHL,
this key role for Nox3 in the development of cisplatin ototoxicity confirming its
role in regulatory mechanisms of cochlear damage encouraged us to validate
this candidate gene for NIHL.
134
The only study exploring NIHL and the NOX family (including Nox3) was
completed in rats [34]. This study did not indicate whether the Nox3 gene
decreased or increased the susceptibility to noise, but instead it evaluated Nox3
expression levels after noise exposure. Some members of the NADPH oxidase
family (Nox1 and Duox2) were up‐regulated in the rat cochlea after noise
exposure, suggesting that these isoforms could be linked to cochlear injury. In
contrast, the Nox3 isoform was down‐regulated after exposure to 100 dB SPL
and 110 dB SPL by seven and fivefold respectively, which could represent an
endogenous protective mechanism against oxidative stress. This protective
mechanism may have decreased the impact of the noise among wild‐type rats
by reducing the expression of Nox3 and decreasing the difference related to
mutants. However, the in vivo data was based on the use of a non‐specific Nox
inhibitor that targeted multiple members of this enzyme without conclusively
demonstrating that Nox3 plays a role in NIHL. Our study, by contrast, has used
animal models with naturally occurring genetic variation and specific genetic
perturbation of Nox3 to directly implicate this oxidative stress enzyme in
hearing.
According to our study, noise exposure might have an opposite effect to
cisplatin on Nox3 expression, suggesting differential involvement of Nox3 on
noise and cisplatin‐induced cochlear damage. Based upon this literature we
hypothesized that the absence or reduction of the Nox3 gene product,
responsible for the production of ROS in the cochlea, would reduce
susceptibility to noise and were startled by our findings. A review of the
135
literature shows there are several key protective mechanisms attributed to the
Nox family of genes. These mechanisms include: host defense and inflammation
(ROS‐dependent killing, inactivation of microbial virulence factors, regulation of
pH and ion concentration in the phagosome and anti‐inflammatory activity),
regulation of gene expression (TNF‐alpha, TGF‐beta1 and angiotensin II), cellular
redox potential, cellular signaling (inhibition of phosphatases, activation of
kinases, regulation of ion channels and Ca2+ signaling), oxygen sensing (kidney,
carotid body and lungs), biosynthesis, regulation of blood pressure, cell growth,
angiogenesis, differentiation and senescence [30]. These protective mechanisms
may very well play a role in the findings of susceptibility to NIHL in the wild‐type
animals.

Validation of Nox3’s role in NIHL. We were able to validate our frequency‐
specific GWAS findings in isolation by studying Nox3
het
mutant mice. After noise
exposure there was a statistically significant difference between the wild‐type
mice in comparison to the homozygous mutants and the heterozygotes on
several measures of auditory function specifically and solely after the 8 kHz
exposure. Contrary to the initial expectations, the presence of the Nox3 gene
was clearly protective against noise damage. Also we were able to demonstrate
the genotypic effect of the peak SNP at the same GWAS phenotype at 8 kHz. We
also show genotypic effect on 4 kHz, but this finding was only suggestive in
GWAS and not confirmed in Nox3
het
mutants.
136
We dissected this phenotype in detail physiologically by assessing OHC
function using DPOAEs and IHC/auditory nerve function using ABR. Although
there was no statistically significant difference in DPOAE thresholds amongst the
genotypes, there was a marked difference in the amplitude of wave 1 of the ABR
after suprathreshold stimulation with the 8 kHz tone burst. This suggested that
the mechanism of hearing loss, in relation to Nox3, resided in the spiral ganglion
neurons and likely at 8 kHz along the cochlear place map.
There are many genes differentially expressed along the tonotopic axis of
the cochlea, and this has been shown for Nox3 [35]. It is likely that our
frequency specific finding of variation in susceptibility to NIHL is the result of
this tonotopic expression pattern.
Considering that all of the results pointed to the area of 8 kHz, we
initiated a thorough electrophysiological and histological dissection at this
particular frequency. The evaluation of the DPOAEs and suprathreshold wave 1
ABR amplitudes was performed at multiple stimulus intensity levels. For each
study, the wild‐type were more resistant to NIHL at only at 8 kHz. We performed
immunohistochemistry two weeks after the noise exposure. Although the
difference in OHC loss was not significant, we demonstrated a significantly
higher density of synaptic ribbons in wild‐type mice. Thus, the
electrophysiological findings were verified by the immunohistochemistry,
demonstrating that the presence of Nox3 is protective at the neuronal level and
that the sensory neural hearing loss after noise exposure occurred at this level
of the peripheral auditory system.
137
The absence of differences in outer hair cell count was also verified by its
corresponding electrophysiological measure of DPOAE thresholds. However,
through the evaluation of DPOAE suprathreshold amplitudes, we were able to
observe a statistically significant higher amplitude in the wild‐type mice. These
three different measures of the integrity of the outer hair cells (outer hair cell
count, DPOAE thresholds and DPOAE suprathresholds amplitudes) have
different sensitivity profiles to demonstrate the impact of noise. Probably
DPOAE suprathreshold amplitude is the most sensitive measurement, since
there is greater signal‐noise ratio. This metric indicates that there is significantly
less impact on the activity of the outer hair cells in wild‐type mice.

Nox3 Plays a Protective Role in the Cochlea of Mice. Although Nox3 is
associated with production of O2•‐ in the inner ear, the Nox family has several
physiological and potentially protective mechanisms. Definitely, this protective
role explains the fact that the absence of Nox3 increased susceptibility to NIHL
in our mouse models. However, there is a lack of specific studies about the
mechanisms of the Nox3 gene due to this very focal expression in the inner ear
and functional data on Nox3 have been only gathered in overexpression systems
[36]. Most evidence regarding these mechanisms is derived from other isoforms,
like Nox2, which is functionally similar to Nox3 [29]. Thus, due to the limited
literature, we relied on the other isoforms to formulate hypotheses about the
mechanisms of susceptibility to NIHL.
138
Since ROS are commonly related to inflammation, an anti‐inflammatory
activity of NOX enzymes would seem illogical. However, over recent years there
has been a striking number of publications pointing in the opposite direction.
Most of the data about the anti‐inflammatory activity of Nox enzymes comes
from studies using mice deficient in the phagocyte NADPH oxidase Nox2 as
demonstrated by a decreased capacity to degrade phagocytized material in
Nox2‐deficient cells leading to the accumulation of debris [37]. Also, this
hyperinflammation might be due to a lack of ROS‐dependent signaling in Nox2‐
deficient phagocytes and ROS‐dependent attenuation of Ca2+ signaling
contributing to enhanced inflammation. Lastly, impairment of oxidative
inactivation of proinflammatory mediators leads to a prolongation of the
inflammatory response [30].
Hyperinflammation in NADPH oxidase‐deficient mice was demonstrated
in mouse models of Helicobacter gastritis [38][39], arthritis [40], demyelinating
disease [41], and sunburn [42]. In experimental lung influenza infection, Nox2
deficient mice demonstrated larger inflammatory infiltrates [43]. Also, by
studying endothelial dysfunction, the absence of Nox4 resulted in reduction of
endothelial nitric oxide synthase expression, nitric oxide production, and heme
oxygenase‐1 expression, which was associated with apoptosis and inflammatory
activation [44].
There is mounting evidence that NOX enzymes have a role in limiting the
inflammatory response and we have shown this to be true in noise‐induced
cochlear damage. This anti‐inflammatory activity of NOX enzymes is poorly
139
understood in the cochlea. So far, as described in other isoforms, our initial
hypothesis is that there are important protective mechanisms, such as an anti‐
inflammatory response resulting from noise exposure. This anti‐inflammatory
mechanism would be crucial to protect the cochlea against noise injury,
overcoming its potential for damage caused by the release of ROS.

140
Materials and Methods
Ethics statement. The Institutional Care and Use Committee (IACUC) at
University of Southern California, Los Angeles, approved the animal protocol for
the HMDP strains and the Nox3
het
mice (IACUC 12033). HMDP strains and
C57BL/6JEiJ Nox3
het
(Nox3
het
/Nox3
het
, Nox3
het
/+ and wild‐type) were
anesthetized with an intraperitoneal injection of a mixture of ketamine (80
mg/kg body weight) and xylazine (16 mg/kg body weight).

The Hybrid Mouse Diversity Panel. A detailed description of the HMDP (strain
selection, statistical power and mapping resolution) is provided in Bennett BJ, et
al. 2010. [11]. Approximately four female mice for each HMDP strain were
purchased from the Jackson Laboratory (Bar Harbor, ME). Only female mice
were tested to avoid confounding effects of sex. Mice were 4 weeks of age, and
to ensure adequate acclimatization to a common environment, mice were aged
until 5 weeks. 5‐week‐old mice were selected to eliminate the potential effects
of age‐related hearing loss contributing to our phenotype. All mice were
maintained on a chow diet until sacrifice.

Genotyping. Common and recombinant inbred strains were previously
genotyped by the Broad Institute (www.mousehapmap.org). Of the 140,000
SNPs available, 108,064 were informative (allele frequency ≥ 5% and less than
20% missing data) and were used for the association analysis.

141
Pre and Post Noise Exposure Hearing Thresholds. Stainless‐steel electrodes
were placed subcutaneously at the vertex of the head and the right mastoid,
with a ground electrode at the base of the tail. Body temperature was
maintained and monitored. Artificial tear ointment was applied to the eyes.
Each mouse was recovered on a heating pad at body temperature. Auditory
signals were presented as tone pips with a rise and a fall time of 0.5 msec and a
total duration of 5 msec at the frequencies 4, 8, 12, 16, 24, and 32 kHz. Tone
pips were delivered below threshold and then increased in 5 dB increments until
goal of 100 dB. Signals were presented at a rate of 30/second. Responses were
filtered with a 0.3 to 3 kHz pass‐band (x10,000 times). For each stimulus
intensity 512 waveforms were averaged. Hearing threshold was determined by
inspection of auditory brainstem response (ABR) waveforms and was defined as
the minimum intensity at which wave 1 could be distinguished. Data was stored
for offline analysis of peak‐to‐peak (P1‐N1) values for wave 1 amplitudes. Post‐
exposure thresholds were evaluated by the same method 2 weeks post‐
exposure.

Pre and Post Exposure DPOAE Determination. Distortion product otoacoustic
emissions (DPOAEs) were analyzed as input/output (I‐O) functions with 2f1‐ f2
(primary measure). Primary tones were set at a ratio of f2/f1 = 1.2 with the f2
between 8 to 32 kHz(f2 level set 10 dB less than the f1 level)and L2 ranging from
20 to 70 dB. The noise floor was measured by averaging 6 spectral points (above
and below the 2f1‐ f2). After both waveform and spectral averaging DPOAEs
142
were extracted. Threshold was defined as the L2 level needed to produce a
DPOAE of 0 dB SPL with a signal to noise ratio (SNR) ≥ 3 dB.

Noise exposure and audiometric equipment. 6 week‐old mice were exposed for
2 hours to 10 kHz octave band noise (OBN) at 108 dB SPL using a method
adapted from Kujawa and Liberman (2009) [16]. The OBN noise exposure was
previously described [17]. For 2 hours, mice were placed in a circular ¼‐inch
wire‐mesh exposure cage with four shaped compartments and were able to
move about within the compartment. The cage was placed in a MAC‐1
soundproof chamber designed by Industrial Acoustics (IAC, Bronx, NY) and the
sound chamber was lined with soundproofing acoustical foam to minimize
reflections. Noise recordings were played with a Fostex FT17H Tweeter Speaker
built into the top of the sound chamber. Calibration of the damaging noise was
done with a B&K sound level meter with a variation of 1.5 dB across the cage.
  A data acquisition board from National Instruments (National Instruments
Corporation, Austin, Texas) was regulated by custom software (used to generate
the stimuli and to process the responses). Stimuli were provided by a custom
acoustic system, made up of two miniature speakers, and sound pressure was
measured by a condenser microphone. Testing involved the right ear only. All
hearing tests were performed in a separate MAC‐1 soundproof chamber to
eliminate both environmental and electrical noise.

143
Cochlear RNA Extraction. For each HMDP strain, both cochleae from each 8‐
week‐old mouse were removed. The inner ear was micro‐dissected and the
surrounding soft tissue and the vestibular labyrinth was removed. The dissected
cochleae were then frozen in liquid nitrogen and then ground to powder. RNA
was extracted and purified by placing cochlea samples in RNA lysis buffer
(Ambion). The sample was incubated overnight (4
o
C), centrifuged (12,000g for 5
minutes) to pellet insoluble materials and RNA isolated (following
manufacturer’s recommendations). This procedure generates approximately
300 ng of total RNA per mouse.

Gene Expression Analysis. Illumina’s Mouse whole genome expression,
BeadChips, was used for the gene expression measurements. Amplifications and
hybridizations were performed according to Illumina’s protocol (Southern
California Genome Consortium microarray core laboratory at UCLA). RNA was
reverse transcribed to cDNA using Ambion cDNA synthesis kit (AMIL1791) and
then converted to cRNA and labeled with biotin. Further, 800ng of biotinylated
cRNA product was hybridized to prepare whole genome arrays and was
incubated overnight (16‐20 hrs) at 55
o
C. Arrays were washed and then stained
with Cy3 label. Excess stain was removed by washing and then arrays were
scanned on an Illumina BeadScan confocal laser scanner.

Efficient Mixed‐Model Association (EMMA). EMMA is a statistical test for
association mapping correcting for genetic relatedness and population structure
144
and consider the mean per strain and also individual measurement per mouse
to increase the statistical power. We have previously demonstrated that p <0.05
genome‐wide equivalent for GWA using EMMA in the HMDP is P =
4.1×10
−6
(−log10P = 5.39)[18]. An R package implementation of EMMA is
available online at http://mouse.cs.ucla.edu/emma.

Candidate Gene Characterization. RefSeq genes were downloaded from the
UCSC genome browser (https://genome.ucsc.edu/cgi‐bin/hgGateway) using the
NCBI Build37 genome assembly to characterize genes located in each
association. EMMA was used to calculate association (P‐values) for the probes
corresponding to the RefSeq genes. The confidence interval (95%) for the
distribution of distances between the most significant and the true causal SNPs,
for simulated associations that explain 5% of the variance in the HMDP, is 2.6
Mb [11]. Only SNPs mapping to each associated region were used in this
analysis. We selected SNPs that were variant in at least one of the HMDP
classical inbred strains. Non‐synonymous SNPs within each region were
downloaded from the Mouse Phenome Database (http:// phenome.jax.org/).

Characterizing the Nox3
het
mice. The generation and initial characterization of
Nox3
het
allele was previously described [19]. The Nox3
het
allele arose
spontaneously (endogenous retroviral insertion into intron 12) on the GL/Le
strain, but has since been made congenic onto the C57BL/6JEiJ strain. To
circumvent the probability of additional alleles from the donor strain this
145
congenic region was backcrossed for more than 10 generations. Since the
downless mutant allele is not present in this strain the congenic interval
containing Nox3 is likely less than 5 centimorgans
(http://jaxmice.jax.org/strain/002557.html). Nox3
het
(known as the head‐tilt or
het mice) carry autosomal recessive, spontaneous mutations that lead to
otoconial absence with no apparent abnormalities in other organs. The otoconia
deficit results in head‐tilting behavior and absent vestibular‐evoked potentials
(VsEPs) but normal thresholds ABR [20].
Pre exposure ABR, DPOAE and VsEP in male and female mice (5 weeks
old) of varying Nox3
het
genotype (Nox3
het
/Nox3
het
and Nox3
het
/+) and wild‐type
(C57BL/6JEiJ strain) was measured as described above. Pre‐exposure threshold
levels were obtained at 1 week prior to noise exposure and the animals were
assessed for noise damage 2 weeks after exposure. The ABR permanent
threshold shift (PTS) was defined as the difference between pre‐exposure and
post‐exposure thresholds at each tested frequency. One‐way ANOVA was used
to test the significance and post hoc Tukey test for multiple comparisons.

Cochlear Whole Mount Preparation. Mice were sacrificed less than 24 hours
after the post exposure ABR. Cochleae were dissected from the surrounding
tissues and openings were made into the coils by piercing the apex and
rupturing both the oval and the round windows. The dissection was done in cold
PBS. After dissection, cochleae were fixed in 4% paraformaldehyde for overnight
at 4
o
C and then washed with PBS. Further dissection was done to expose the
146
organ of Corti. For permeabilization and blocking, tissue was immersed for 1
hour in PBS containing 0.2% Triton X‐100 (Sigma Chemical) and 16% normal goat
serum (SouthernBiotech). Samples were incubated overnight at room
temperature with primary antibodies (rabbit anti‐myosin6, 1:500, Proteus
Biosciences and purified mouse anti‐CtBP2, 1:500, BD Biosciences) for doubled‐
staining. Secondary antibody was then applied and tissue was incubated in the
dark overnight (Alexa 594 donkey anti‐rabbit, 1:500, Life technologies and Alexa
Fluor‐488 anti‐mouse, 1:500, Life technologies). After, samples were washed
three times in PBS and mounted on glass slides using Fluoromount G
(SouthernBiotech). Microscopy was carried out with a laser confocal microscope
(Olympus IX81) with epifluorescence light (Olympus Fluoview FV1000). Outer
hair cell loss (% per 100µm) was counted and plotted as cytocochleogram by
relating distance of cochlear apex to the tonotopic map of mice of strain CBA
[21]. Percentages indicate the normalized location of the inner and outer hair
cells in the cochlea (0%, apical and 100%, basal end) in 10% steps.
Synaptic ribbon density was plotted for each correspondent ABR
frequency (4, 8, 12, 16, 24 and 32 kHz) against the same tonotopic map. Inner
hair cells were analyzed in a row (50 μm) for each frequency. CtBP2
immunofluorescence spots were counted in z‐stacks and divided by the number
of inner hair cells (measured as the quantity of nuclei) in the sample.

Nox3
het
PCR. Polymerase chain reaction (PCR) was performed for Nox3 using the
following primers: Nox3‐int12F, GTTCTGGAGCACCACCTTGT; Nox3‐int12R
147
CCCATAGGGAGCCAAGAAAT; and ERV‐R, TGTCAAGCTGACTCCACCAG [19]. PCR
products were separated on a 1.5% agarose gel containing 0.5 mg/ml ethidium
bromide.
148
Acknowledgments
The authors acknowledge the technical contributions of Robert Rainey and Litao
Tao (USC). We thank Yesha W. Lundberg, PhD (Boys Town National Research
Hospital) for providing Nox3
het
mice (Nox3
het
/Nox3
het
and Nox3
het
/+).


149
References
1. Stucken EZ, Hong RS (2014) Noise‐induced hearing loss: an occupational
medicine perspective. Curr Opin Otolaryngol Head Neck Surg 22: 388–393.
doi:10.1097/MOO.0000000000000079.
2. Dobie RA (2008) The burdens of age‐related and occupational noise‐
induced hearing loss in the United States. Ear Hear 29: 565–577.
doi:10.1097/AUD.0b013e31817349ec.
3. Yankaskas K (2013) Prelude: noise‐induced tinnitus and hearing loss in
the military. Hear Res 295: 3–8. doi:10.1016/j.heares.2012.04.016.
4. Sliwinska‐Kowalska M, Pawelczyk M (2013) Contribution of genetic
factors to noise‐induced hearing loss: a human studies review. Mutat Res 752:
61–65. doi:10.1016/j.mrrev.2012.11.001.
5. Heinonen‐Guzejev M, Vuorinen HS, Mussalo‐Rauhamaa H, Heikkilä K,
Koskenvuo M, et al. (2005) Genetic component of noise sensitivity. Twin Res
Hum Genet Off J Int Soc Twin Stud 8: 245–249. doi:10.1375/1832427054253112.
6. Erway LC, Shiau YW, Davis RR, Krieg EF (1996) Genetics of age‐related
hearing loss in mice. III. Susceptibility of inbred and F1 hybrid strains to noise‐
induced hearing loss. Hear Res 93: 181–187.
7. Ohlemiller KK, McFadden SL, Ding DL, Flood DG, Reaume AG, et al. (1999)
Targeted deletion of the cytosolic Cu/Zn‐superoxide dismutase gene (Sod1)
increases susceptibility to noise‐induced hearing loss. Audiol Neurootol 4: 237–
246. doi:13847.
8. Ohlemiller KK, McFadden SL, Ding DL, Lear PM, Ho YS (2000) Targeted
mutation of the gene for cellular glutathione peroxidase (Gpx1) increases noise‐
induced hearing loss in mice. J Assoc Res Otolaryngol JARO 1: 243–254.
9. Kozel PJ, Davis RR, Krieg EF, Shull GE, Erway LC (2002) Deficiency in
plasma membrane calcium ATPase isoform 2 increases susceptibility to noise‐
induced hearing loss in mice. Hear Res 164: 231–239.
10. Holme RH, Steel KP (2004) Progressive hearing loss and increased
susceptibility to noise‐induced hearing loss in mice carrying a Cdh23 but not a
Myo7a mutation. J Assoc Res Otolaryngol JARO 5: 66–79. doi:10.1007/s10162‐
003‐4021‐2.
11. Bennett BJ, Farber CR, Orozco L, Kang HM, Ghazalpour A, et al. (2010) A
high‐resolution association mapping panel for the dissection of complex traits in
mice. Genome Res 20: 281–290. doi:10.1101/gr.099234.109.
12. Farber CR, Bennett BJ, Orozco L, Zou W, Lira A, et al. (2011) Mouse
genome‐wide association and systems genetics identify Asxl2 as a regulator of
150
bone mineral density and osteoclastogenesis. PLoS Genet 7: e1002038.
doi:10.1371/journal.pgen.1002038.
13. Park CC, Gale GD, de Jong S, Ghazalpour A, Bennett BJ, et al. (2011) Gene
networks associated with conditional fear in mice identified using a systems
genetics approach. BMC Syst Biol 5: 43. doi:10.1186/1752‐0509‐5‐43.
14. Smolock EM, Ilyushkina IA, Ghazalpour A, Gerloff J, Murashev AN, et al.
(2012) Genetic locus on mouse chromosome 7 controls elevated heart rate.
Physiol Genomics 44: 689–698. doi:10.1152/physiolgenomics.00041.2012.
15. Davis RC, van Nas A, Bennett B, Orozco L, Pan C, et al. (2013) Genome‐
wide association mapping of blood cell traits in mice. Mamm Genome Off J Int
Mamm Genome Soc 24: 105–118. doi:10.1007/s00335‐013‐9448‐0.
16. Kujawa SG, Liberman MC (2009) Adding insult to injury: cochlear nerve
degeneration after “temporary” noise‐induced hearing loss. J Neurosci Off J Soc
Neurosci 29: 14077–14085. doi:10.1523/JNEUROSCI.2845‐09.2009.
17. White CH, Ohmen JD, Sheth S, Zebboudj AF, McHugh RK, et al. (2009)
Genome‐wide screening for genetic loci associated with noise‐induced hearing
loss. Mamm Genome Off J Int Mamm Genome Soc 20: 207–213.
doi:10.1007/s00335‐009‐9178‐5.
18. Kang HM, Zaitlen NA, Wade CM, Kirby A, Heckerman D, et al. (2008)
Efficient control of population structure in model organism association mapping.
Genetics 178: 1709–1723. doi:10.1534/genetics.107.080101.
19. Flaherty JP, Fairfield HE, Spruce CA, McCarty CM, Bergstrom DE (2011)
Molecular characterization of an allelic series of mutations in the mouse Nox3
gene. Mamm Genome Off J Int Mamm Genome Soc 22: 156–169.
doi:10.1007/s00335‐010‐9309‐z.
20. Zhao X, Jones SM, Yamoah EN, Lundberg YW (2008) Otoconin‐90 deletion
leads to imbalance but normal hearing: a comparison with other otoconia
mutants. Neuroscience 153: 289–299. doi:10.1016/j.neuroscience.2008.01.055.
21. Müller M, von Hünerbein K, Hoidis S, Smolders JWT (2005) A
physiological place‐frequency map of the cochlea in the CBA/J mouse. Hear Res
202: 63–73. doi:10.1016/j.heares.2004.08.011.
22. Paffenholz R, Bergstrom RA, Pasutto F, Wabnitz P, Munroe RJ, et al.
(2004) Vestibular defects in head‐tilt mice result from mutations in Nox3,
encoding an NADPH oxidase. Genes Dev 18: 486–491. doi:10.1101/gad.1172504.
23. Bánfi B, Malgrange B, Knisz J, Steger K, Dubois‐Dauphin M, et al. (2004)
NOX3, a superoxide‐generating NADPH oxidase of the inner ear. J Biol Chem
279: 46065–46072. doi:10.1074/jbc.M403046200.
151
24. Civelek M, Lusis AJ (2014) Systems genetics approaches to understand
complex traits. Nat Rev Genet 15: 34–48. doi:10.1038/nrg3575.
25. Flint J, Eskin E (2012) Genome‐wide association studies in mice. Nat Rev
Genet 13: 807–817. doi:10.1038/nrg3335.
26. Ghazalpour A, Rau CD, Farber CR, Bennett BJ, Orozco LD, et al. (2012)
Hybrid mouse diversity panel: a panel of inbred mouse strains suitable for
analysis of complex genetic traits. Mamm Genome Off J Int Mamm Genome Soc
23: 680–692. doi:10.1007/s00335‐012‐9411‐5.
27. Ohmen J, Kang EY, Li X, Joo JW, Hormozdiari F, et al. (2014) Genome‐
wide association study for age‐related hearing loss (AHL) in the mouse: a meta‐
analysis. J Assoc Res Otolaryngol JARO 15: 335–352. doi:10.1007/s10162‐014‐
0443‐2.
28. Kikuchi H, Hikage M, Miyashita H, Fukumoto M (2000) NADPH oxidase
subunit, gp91(phox) homologue, preferentially expressed in human colon
epithelial cells. Gene 254: 237–243.
29. Cheng G, Cao Z, Xu X, van Meir EG, Lambeth JD (2001) Homologs of
gp91phox: cloning and tissue expression of Nox3, Nox4, and Nox5. Gene 269:
131–140.
30. Bedard K, Krause K‐H (2007) The NOX family of ROS‐generating NADPH
oxidases: physiology and pathophysiology. Physiol Rev 87: 245–313.
doi:10.1152/physrev.00044.2005.
31. Rybak LP, Mukherjea D, Jajoo S, Kaur T, Ramkumar V (2012) siRNA‐
mediated knock‐down of NOX3: therapy for hearing loss? Cell Mol Life Sci CMLS
69: 2429–2434. doi:10.1007/s00018‐012‐1016‐3.
32. Mukherjea D, Whitworth CA, Nandish S, Dunaway GA, Rybak LP, et al.
(2006) Expression of the kidney injury molecule 1 in the rat cochlea and
induction by cisplatin. Neuroscience 139: 733–740.
doi:10.1016/j.neuroscience.2005.12.044.
33. Mukherjea D, Jajoo S, Kaur T, Sheehan KE, Ramkumar V, et al. (2010)
Transtympanic administration of short interfering (si)RNA for the NOX3 isoform
of NADPH oxidase protects against cisplatin‐induced hearing loss in the rat.
Antioxid Redox Signal 13: 589–598. doi:10.1089/ars.2010.3110.
34. Vlajkovic SM, Lin SC‐Y, Wong ACY, Wackrow B, Thorne PR (2013) Noise‐
induced changes in expression levels of NADPH oxidases in the cochlea. Hear
Res 304: 145–152. doi:10.1016/j.heares.2013.07.012.
35. Son EJ, Wu L, Yoon H, Kim S, Choi JY, et al. (2012) Developmental gene
expression profiling along the tonotopic axis of the mouse cochlea. PloS One 7:
e40735. doi:10.1371/journal.pone.0040735.
152
36. Brandes RP, Weissmann N, Schröder K (2014) Nox family NADPH
oxidases: Molecular mechanisms of activation. Free Radic Biol Med.
doi:10.1016/j.freeradbiomed.2014.07.046.
37. Brown JR, Goldblatt D, Buddle J, Morton L, Thrasher AJ (2003) Diminished
production of anti‐inflammatory mediators during neutrophil apoptosis and
macrophage phagocytosis in chronic granulomatous disease (CGD). J Leukoc Biol
73: 591–599.
38. Blanchard TG, Yu F, Hsieh C‐L, Redline RW (2003) Severe inflammation
and reduced bacteria load in murine helicobacter infection caused by lack of
phagocyte oxidase activity. J Infect Dis 187: 1609–1615. doi:10.1086/374780.
39. Keenan JI, Peterson RA, Hampton MB (2005) NADPH oxidase
involvement in the pathology of Helicobacter pylori infection. Free Radic Biol
Med 38: 1188–1196. doi:10.1016/j.freeradbiomed.2004.12.025.
40. Olofsson P, Holmberg J, Tordsson J, Lu S, Akerström B, et al. (2003)
Positional identification of Ncf1 as a gene that regulates arthritis severity in rats.
Nat Genet 33: 25–32. doi:10.1038/ng1058.
41. Hultqvist M, Olofsson P, Holmberg J, Bäckström BT, Tordsson J, et al.
(2004) Enhanced autoimmunity, arthritis, and encephalomyelitis in mice with a
reduced oxidative burst due to a mutation in the Ncf1 gene. Proc Natl Acad Sci U
S A 101: 12646–12651. doi:10.1073/pnas.0403831101.
42. Komatsu J, Koyama H, Maeda N, Aratani Y (2006) Earlier onset of
neutrophil‐mediated inflammation in the ultraviolet‐exposed skin of mice
deficient in myeloperoxidase and NADPH oxidase. Inflamm Res Off J Eur
Histamine Res Soc Al 55: 200–206. doi:10.1007/s00011‐006‐0071‐3.
43. Snelgrove RJ, Edwards L, Rae AJ, Hussell T (2006) An absence of reactive
oxygen species improves the resolution of lung influenza infection. Eur J
Immunol 36: 1364–1373. doi:10.1002/eji.200635977.
44. Schröder K, Zhang M, Benkhoff S, Mieth A, Pliquett R, et al. (2012) Nox4
is a protective reactive oxygen species generating vascular NADPH oxidase. Circ
Res 110: 1217–1225. doi:10.1161/CIRCRESAHA.112.267054.

153
Figure Legends

Figure 1. Characterization of post‐exposure thresholds in the HMDP. Mean ±
SEM for 8 kHz post‐noise exposure hearing thresholds in 64 HMDP inbred
mouse strains. The difference between the strains with the lowest and the
highest values were 3.22‐fold.

Figure 2. GWAS results for post‐noise exposure thresholds in the HMDP.
Manhattan plot showing the association (‐log10) p‐values (‐logP) for 8 kHz in 64
HMDP inbred mouse strains. The analysis was performed using 108,064 SNPs
with a minor allele frequency > 5%. Each chromosome is plotted on the x‐axis in
alternating brown and blue colors. SNPs on Chr. 2 and Chr. 17 for 8 kHz
exceeded the predetermined genome‐wide significance threshold (‐logP = 5.39).

Figure 3. Regional plot of the 8 kHz ABR post noise‐exposure at Chr 17
association in the HMDP centered on the lead SNP at the Nox3 locus
(rs33652818). The blue diamond represents the most significant SNP (p = 9.63E‐
06) and SNPs are colored based on their LD with the most significant SNP being:
red SNPs in LD at r2>0.8, orange SNPs in LD at r2>0.6 and green SNPs in LD at
r2>0.4. The positions of all RefSeq genes are plotted using genome locations
(NCBI’s Build37 genome assembly).

Figure 4. Nox3het mice have greater PTS (permanent threshold shift) for 8 kHz.
Nox3het/Nox3het and Nox3het/+ display significantly greater PTS in
154
comparison to wild‐type controls. Data shown are mean comparisons analyzed
by one‐way ANOVA (post hoc Tukey test for multiple comparisons). *p < 0.05.
Homozygous = Nox3het/Nox3het: Heterozygous = Nox3het/+ : Wild‐type =
C57BL/6JEiJ strain.

Figure 5. Genotypic effects of the peak SNP (rs33652818) at the Nox3 locus.
Comparison between alleles GG and AA across the various frequencies. There is
a statistically significant difference between the alleles at 4 and 8 kHz. * = p
value < 0.001. Error bars +/‐ 1 SE.

Figure 6. Topographical analysis of the auditory pathway at different
frequencies (post‐noise exposure). Although no significant difference was seen
for the DPOAE thresholds (6A), the wild‐type controls show a higher wave 1
amplitude (p = 0.010) only at 8 kHz compared to Nox3het/+ and
Nox3het/Nox3het (6B). One‐way ANOVA (Tukey test for multiple comparisons).
* = p < 0.05.  Homozygous = Nox3het/Nox3het: Heterozygous = Nox3het/+ :
Wild‐type = C57BL/6JEiJ strain.

Figure 7. Detailed analysis of the 8 kHz frequency stimulus. Dissection of the 8
kHz frequency by DPOAE I/O function (7A) and ABR wave 1 amplitudes (7B)
consistently indicate more impairment in the Nox3het/Nox3het and Nox3het/+
than wild‐type. One‐way ANOVA (Tukey multiple comparisons). * = p < 0.05.
155
Homozygous = Nox3het/Nox3het: Heterozygous = Nox3het/+ : Wild‐type =
C57BL/6JEiJ strain.

Figure 8. Cytocochleogram of wild‐type and mutant Nox3 mice (8A). No
significant difference in OHC counts were detected among wild‐type, Nox3het/+
and Nox3het/Nox3het. OHC preparations of immunostained (40x) post‐noise
exposure cochleae (8B) at the 8 kHz tonotopic location (red, rabbit anti‐
myosin6) demonstrate a small impact with the loss of apical OHC (2 weeks post‐
noise exposure). Homozygous = Nox3het/Nox3het: Heterozygous = Nox3het/+:
Wild‐type = C57BL/6JEiJ strain.

Figure 9. 8 kHz Synaptic Cochleogram. Synaptic ribbon count measured at the 8
kHz tonotopic position within the cochlea. Despite the absence of a statistical
difference in OHC counts, the wild‐type mice demonstrated significantly greater
post‐noise exposure synaptic ribbon density per IHC (9A). Projections (60x, 3x
zoom, oil immersed) of confocal stacks (9B) of immunostained pre and post‐
noise exposure mouse IHC synaptic ribbons (green, mouse anti‐CtBP2). ‐/‐ =
Nox3het/Nox3het: ‐/+ = Nox3het/+: +/+ = wild‐type.
156
Table 1. GWA results for NIHL in the HMDP.

a
Post noise‐exposure hearing thresholds in different frequencies.
b
Locations based on genome assembly (NCBI´s Build37)
c
MAF, minor allele frequency
d
Number of RefSeq genes (NCBI´s Build37 assembly) located in the mouse association confidence interval (2.6 Mb)
* Genome‐wide significant (p<4.1E‐6)

Trait
a
Chr SNP Position
(Mb)
b
‐logP MAF
c
No. of
Genes
d
Human Region
(Chr: Start Mb – End Mb)
4 kHz 17 rs33652818 3.8 1.1E
‐04
0.222 10 Chr6:155.0‐155.7
8 kHz 2 rs27972902 68.1 8.58E
‐07
* 0.222 2 Chr2:168.8‐169.6
8 kHz 17 rs33652818 3.8 2.25E
‐06
* 0.222 10 Chr6:155.0‐155.7
16 kHz 15 rs32934144 28.2 1.74E
‐06
* 0.389 2 Chr5:13.6‐14.8
32 kHz 3 rs30795209 96.7 1.95E
‐07
* 0.278 11 Chr1:145.6‐147.4
32 kHz 15 rs32278602 22.3 5.86E
‐07
* 0.389 2 Chr5:19.4‐22.2
157
Table 2. Genes within NIHL 5 association peaks regulated by local eQTL in the cochlea.
Gene RefSeq Chr txStart (bp)
a
txEnd (bp)
b
Local eQTL P
c
Pias3 ILMN_2631014 3 96696384 96706070 6.78E‐02
CD160 ILMN_2707181 3 96798763 96829351 1.49E‐02
Gja8 ILMN_2625168 3 96918863 96926020 3.09E‐01
Gja5 ILMN_2678477 3 97032416 97053634 4.07E‐01
Tiam2 ILMN_2836875 17 3326573 3519397 3.71E‐01
Tfb1m ILMN_2690441 17 3519263 3557713 1.08E‐06

a
txStart, location of transcription (NCBI Build37 genome assembly) start.
b
txEnd, location of transcription (NCBI Build37 genome assembly) end.
c
Statistically significant p value ≤ 5.1E‐04 (Bonferroni corrected for the number of probes tested)

158

Figure 1

159

Figure 2

160

Figure 3

161

Figure 4

162

Figure 5

163

Figure 6
164

Figure 7
165

Figure 8

166

Figure 9
167
Supplementary Material

Figure S1. GWAS results for 4 kHz post‐noise exposure thresholds in the HMDP.
Manhattan plot showing the association (‐log10) p‐values (‐logP) for 4 kHz in 64
HMDP inbred mouse strains. The analysis was performed using 108,064 SNPs with a
minor allele frequency > 5%. The level of significance at 4 kHz (rs33652818) is
suggestive and does not reach genome‐wide significance (p=1.1x10
‐4
).
168
THE HDMP IN GENETIC STUDIES

A powerful tool for GWAS
Taken together with others, our studies have demonstrated the efficiency
and the versatility of using a large panel of inbred mice to identify positional
candidates for complex phenotypes. The HMDP, in particular, achieves a
combination of sufficient statistical power, high mapping resolution, and
convenience that makes it an ideal platform to study traits that have previously had
limited success for gene discovery.
As described in the first chapter of this dissertation, we first applied the
HMDP in genetic studies of hematopoietic stem cell frequency in the bone marrow
compartment. Though reverse genetics approaches such as the disruption of
candidate genes has provided an initial understanding of many of the
developmental mechanisms behind stem cell behavior [1], there is much that
remains unclear. Unbiased genetic approaches such as a GWAS contribute
significantly to our comprehensive knowledge of the genetic networks that control
these traits, but a GWAS powered to detect hematopoietic stem cell regulators in
humans with more subtle effects requires bone marrow from thousands of
subjects. As collection of bone marrow is a semi‐invasive, painful procedure, this is
obviously not feasible. The success of our study confirms utility of the HMDP for
previously inaccessible phenotypes.
169

Controlling for Environmental Variables
Environmental factors are an ongoing concern in genetic studies.
Participants in large human cohorts may provide unreliable estimates of their
exposure, or quantitative variables approximating a swath of exposures may still
not represent the full extent of environmental impact. In the statistical model, the
addition of multiple variables decreases power and thereby diminishes the
likelihood of detecting smaller genetic effects, even if the exposure is accurately
measured. For a trait such as common variation in hearing, where environmental
exposure can have a significant impact, the limiting factor for human GWAS has not
been the ability to phenotype, since performing hearing assessments in humans is
common and reliable. Rather, the challenge is the difficulty of accurately
quantitating the amount of noise exposure accumulated by adults throughout their
lifetimes. Some studies have surveyed a few hundred or a few thousand members
of isolated communities with the assumption that these relatively homogenous
populations will have minimal variation in noise exposure, but this approach is
imperfect, and these studies have remained underpowered and hampered by issues
related to population stratification [2, 3].
By contrast, our GWAS studies with the HMDP have addressed the
environmental problem two‐fold: we first utilize the mouse model in an isolated
environment that ensures every animal receives the same exposure, and second,
170
we directly apply a uniform environmental exposure to affect a change in the post‐
noise damage auditory brainstem response (ABR) of the animals, which represents
a type of experimental intervention that cannot be done in humans due to ethical
concerns.

Future directions for stem cell and hearing phenotype genetic studies
Our initial GWAS with the HMDP have identified novel genetic mechanisms
behind HSPC frequency and noise‐induced hearing loss (NIHL), which provide the
foundation for future studies and further exploration of the underlying biological
mechanisms. For instance, we validated Hopx as a regulator of short‐term HSCs;
however, in the early stages of our prioritization of positional candidates, we found
that Spink2, a gene that neighbors Hopx, was very highly expressed in the bone
marrow as well. Studies have suggested the possibility of multiple causal genes at
significant GWAS loci [4], thus raising the possibility further testing of Spink2 could
validate this gene as a second regulator of short‐term HSC frequency. Further,
though we initially chose to pursue the locus containing Hopx and Spink2 for
functional validation, it was not the only significant locus identified by our GWAS.
Therefore, an immediate extension of these initial analyses would be to explore the
role of positional candidates at other significant or highly suggestive loci,
particularly those for the two other HSPC populations.
171
While we have only quantitated the frequency of HSCs in the HMDP, such an
approach could also conceivably be applied the HSC function as well. For example,
the ATP bioluminescence signal detection assay described by Rich [5] evaluates
hematopoietic proliferation, self‐renewal, and expansion potential, which can be
carried out for the HMDP strains as well. However, such a study would be more
labor intensive and more challenging than our flow cytometry analyses.
Our GWAS also suggests that similar studies may be successful in
investigating a range of phenotypes in stem cell populations from other tissues. For
example, the recent discovery of cardiac stem progenitor cells (CSPCs) [6] carries a
range of implications for cardiovascular disease (CVD) and recovery after
myocardial infarction (MI). These cells have self‐renewal capabilities and can
differentiate into a range of terminal cardiac cell types; however, very little is still
known about the mechanisms that drive their proliferation and differentiation [7].
Groups such as Davis et. al. [8] have described methods of growing and collecting
spherical cardiac aggregates called “cardiospheres” in cell culture from an initial
population of CSPCs harvested from biopsy. Reproducing these experiments with
HMDP strains could provide the opportunity to determine whether there is
significant phenotypic variation among strains for several CSPC parameters, such as
growth rate in culture or proportions of terminally differentiated cell types in
collected cardiospheres. Any phenotypes capturing the turnover of CSPCs to mature
cardiac cells represent the potential for a GWAS analysis in the HMDP.
172
Our study in hearing also presents many avenues for follow‐up work. Our
first step should be the validation of positional candidates identified for hearing and
the study of the mechanisms by which the genes influence hearing. Hearing has a
mechanical component, where vibrations from auditory waves are received by hair
cells in the inner ear, and a neural component, where the inner ear cells transduce
those waves into nerve impulses that reach the brain [9]. Exploring whether our
candidate genes are active in cells in the inner ear or in neurons will be valuable for
explaining their functional mechanisms and potentially help to clarify this complex
phenotype at the molecular level. With the observation that most of the loci from
our GWAS exhibit associations at specific frequencies in the hearing spectrum, it
would be of interest to determine whether the causal gene(s) from those loci exert
their effects in hair cells or spiral ganglion neurons in certain regions of the cochlea.
Such an observation would reinforce the importance of cochlear anatomy in
determining frequency‐specific hearing.
In addition, ongoing GWAS analyses are currently being pursued for inner
ear phenotypes. While our studies thus far have sought to dissect hearing and
cochlear biology, the vestibular system, which is involved in balance and spatial
orientation, is also being investigated by characterizing the HMDP for phenotypic
variation in linear vestibular sensory evoked potentials (VsEP). As both the cochlea
and the vestibule are complex systems with many different methods to assess
173
function, HMDP studies can conceivably be utilized for other measurable
phenotypes in the inner ear as well.

Gene by environment interactions in the GWAS model
A remaining question from our GWAS on a “pre” and “post” noise exposure
dataset is how to evaluate the strength of interactions between genes and the
environment in the same way that we can measure the effect size of a SNP by itself.
Our validation of Nox3 as a gene involved in hearing “post” noise exposure is an
example of an association between a gene and a phenotype that is detected after
physiochemical damage induced by an environmental insult; however, this does not
allow us to draw conclusions about the statistical relationship between Nox3 and
noise.
One approach to addressing this issue would be to include an interaction
variable in the model, in addition to main effects terms for the SNP. This
modification is very costly in terms of statistical power and it has been estimated
that detection of any interaction effect requires a sample size of at least four times
larger than that used to detect a main SNP effect of similar size [10]. Additionally,
studies have suggested that in the same way that genetic relatedness can result in
spurious associations for main SNP effects, population stratification can also
generate false positive interaction effects when individuals are both genetically
similar and share a similar environment [11, 12]. For example, previous studies
174
using a GxE model may not have adequately addressed the effect of population
structure on GxE interactions, or were underpowered and detected only suggestive
or minimally significant associations [13‐17].
While no algorithm can independently increase statistical power without
removing variables, an algorithm that sufficiently corrects for population structure
and addresses relatedness effects on main SNP effects and GxE interactions could
be integrated with the HMDP platform. For this purpose, the Eskin group’s pyLMM
software is ideal [12], since in preliminary testing the genomic inflation factor for
SNPs in GWAS and GxE GWAS using the HMDP is closer to the ideal value of 1
compared to those achieved with previous algorithms. As the effect of population
stratification is more severe when highly related individuals also share similar
environments, or whose exposure is denoted by the same value, the pyLMM model
adjusts both for genetic similarity and genetic‐with‐environment similarity.

GxE and the HMDP
While the resolution of the HMDP is excellent at most loci, the power with
the panel of 100 strains may only allow identification of a subset of the loci
contributing to complex traits. The issue of power becomes even more crucial for
GxE studies. However, power can be supplemented by performing integration with
results from traditional crosses. This approach was recently demonstrated by
carrying out a meta‐analysis with a subset of the HMDP, N2 mice from a C57BL/6J x
175
CAST/EiJ cross, and 226 CI strains to identify new loci for age‐related hearing loss
(AHL) [18]. In addition, the HMDP can be expanded with additional strains already
available from The Jackson Laboratory. These include 62 strains in the LXS RI panel
and 50 more BXD strains that have recently been generated. Also available are
several sets of cyropreserved strains: 19 strains from the AKXD RI panel, 13 from
the AKXL panel, and 15 from the NXSM panel. Thus it is possible to increase the size
of the HMDP to over 260 strains. Inclusion of other types of genetically derived
strains that can complement the HMDP, such as RI congenic lines, chromosome
substitution strains, and genome tagged mice [19] (CITE NADEAU’S REVIEW HERE
AS WELL) could also be employed to increase both power and resolution This could
even include the incorporation of strains from the CC [20, 21]. Taken together, the
GWAS approach in mice built on the backbone of the HMDP represents an exciting
platform for GxE investigations that may not be as limited by issues related to
power, false positive associations, and gene‐gene or gene‐environment
interactions.
The mouse provides additional important advantages beyond mapping loci,
particularly with respect to functional validation of positional candidates. For
example, even successful human GxE GWAS are still limited by the difficulty of
replicating human environmental conditions in in vitro and in vivo models to test
candidates, whereas collection of exposed tissues and thorough exploration of
molecular phenotypes is inherent in the mouse GxE GWAS. However, mice in which
176
positional candidates have been genetically perturbed either by inactivation or
overexpression can be exposed using precisely the same protocol as the inbred
mice that were used to initially identify the GxE loci of interest. Furthermore, mRNA
expression of the candidate can be assessed in tissues collected from multiple
exposure groups and strains. While testing candidate genes in response to exposure
is already a common and established approach, the behavior of the gene in this
case would be assessed according to genotype and exposure, therefore confirming
a true interaction.

Beyond GxE
It is widely accepted that complex phenotypes are substantially affected by
the environment and, as such, GWAS focusing solely on SNP effects will only
elucidate a portion of phenotypic variability. GxE analysis addresses the hypothesis
that, in all likelihood, a much larger fraction of variation in a phenotype can be
explained by genetic interaction with the environment. The ongoing improvement
in technology to perform GxE GWAS affords the opportunity to confirm this
supposition and identify novel factors that influence complex traits.
The environmental impact on complex phenotypes may not be limited,
however, to interactions between certain exposures and traditional ‘genes’, that is,
DNA sequences that code for mRNA which codes for proteins. The common disease
genetic and epigenetic (CDGE) hypothesis suggests an integrated approach that
177
considers how the genome and epigenome interact with each other and are
collectively influenced by the environment [22, 23]. CDGE argues that the most
common epigenetic tags, including DNA methylation and various post‐translational
histone modifications, influence DNA expression as they themselves are influenced
by environmental factors. For example, DNA methylation is affected by numerous
environmental factors, including heavy metal exposure [24], folate deficiency [25],
and diet of parents and grandparents [26]. Thus, epigenetic modifications could also
be associated with complex disease outcomes. Therefore, the impact of the
environment on a phenotype is not comprehensively explored without taking into
account the epigenome.
Epigenome‐wide association studies (EWAS) are limited by complicating
factors, including cell‐specific methylation patterns and cell heterogeneity within a
tissue [27]. However, with epigenetic data available in over 100 mouse tissues from
the ENCODE project [28], prioritization of positional candidates from GWAS and GxE
GWAS can utilize epigenetic data to strengthen the case for certain genes, or
alternatively, identify regions that appear as ‘gene deserts’ but may in fact contain
strong regulatory elements. Furthermore and of particular relevance, an EWAS
using methylation data in the liver was recently carried out with the HMDP [29],
which yielded two interesting observations. First, strong correlations were
identified between the degree of site‐specific methylation levels and numerous
clinical traits. Second, a locus harboring the methionine synthase reductase gene
178
(Mtrr) was shown to be associated with methylation levels of hundreds of CpGs
throughout the genome. Thus, incorporation of epigenetic data with phenotype and
DNA variation may provide additional biological insight into the genetic architecture
of complex traits.
Another potential confounding issue may be related to what Bjornsson et.
al. have suggested as epigenotyping [22]. For example, a candidate locus from
human GWAS that is significantly associated in one population, but is not replicated
in another, may reveal a variant/haplotype that is strongly modified by the regional
epigenome [22]. This notion could also extend to the HMDP, as demonstrated by
our studies. In our GWAS for hearing with 64 strains, we identified a statistically
significant association with a SNP in Nox3, a gene which we validated though
phenotypic profiling of Nox3 knockout mice. However, in a GWAS with a more
complete panel of 100 strains, the Nox3 locus was no longer statistically significant,
even though the minor allele frequency (MAF) of the SNP is approximately 20% in
both the 64 strain and the 100 strain analysis. It is feasible that, in a post‐noise
cochlea that is damaged and undergoing oxidative stress, the epigenetic profile, and
by extension gene expression, of some strains may have changed, thereby masking
the association of the Nox3 locus with hearing.
While we are only just beginning to access the full potential of GxE GWAS,
the wealth of publicly available epigenome data suggests that we will want to
additionally consider the substantial effect these modifications can have on
179
complex phenotypes. In particular, environment‐mediated changes in the
epigenetic profile can impact gene expression even in the absence of environment
interactions with naturally occurring genetic variation. As HMDP studies continue to
interrogate the environmental component, bioinformatic analyses and functional
validation efforts in the future should evaluate the epigenome at significant loci to
construct a more comprehensive picture of that region’s effect on the phenotype.

Summary
In this dissertation, I describe three applications of the HMDP to discover
novel genetic regulators for complex phenotypes. First, we performed GWAS for
three HSPC populations and validated Hopx as a positional candidate that regulates
short‐term HSC frequency. Second, we investigated variation in hearing by
performing GWAS for ABR at six different frequencies and identified a set of
frequency‐specific loci, as well as loci that are shared among similar frequencies.
Third, we performed another set of GWAS for ABR at the same frequencies as in the
prior study, but on animals that had been subjected to noise insult. These analyses
resulted in the identification and validation of a gene, Nox3, that affects NIHL.
Taken together, these studies demonstrate the power of the HMDP to explore a
wide range of phenotypes, and suggest future potential for the panel to be an
effective tool for analyses of other phenotypesthat have remained difficult to study.
For example, the GxE GWAS has been recognized for some time as a necessary
180
next‐generation analysis, as the majority of complex traits have a substantial
environmental component. However, few of these studies to date have achieved
success due to limited statistical power and inadequate control for environmental
factors. As studies with the HMDP are not limited by either of these concerns, use
of an appropriate modeling algorithm is the main factor that requires consideration.
Further, the role of the epigenome is inextricably linked with the environment and
genetic regulation, and the HMDP has already been utilized in an EWAS. In
conclusion, the versatility of the HMDP, in addition to its convenience and proven
success in identification of causal genes, marks the panel as a very attractive
platform for the study of complex traits.

181

REFERENCES

1. Rossi, L., et al., Less is more: unveiling the functional core of hematopoietic
stem cells through knockout mice. Cell Stem Cell, 2012. 11(3): p. 302‐17.
2. Girotto, G., et al., Hearing function and thresholds: a genome‐wide
association study in European isolated populations identifies new loci and
pathways. J Med Genet, 2011. 48(6): p. 369‐74.
3. Van Laer, L., et al., A genome‐wide association study for age‐related hearing
impairment in the Saami. Eur J Hum Genet, 2010. 18(6): p. 685‐93.
4. Tada, H., et al., Multiple associated variants increase the heritability
explained for plasma lipids and coronary artery disease. Circ Cardiovasc
Genet, 2014. 7(5): p. 583‐7.
5. Rich, I.N., Measurement of hematopoietic stem cell proliferation, self‐
renewal, and expansion potential. Methods Mol Biol, 2015. 1235: p. 7‐17.
6. Frati, C., et al., Resident cardiac stem cells. Curr Pharm Des, 2011. 17(30): p.
3252‐7.
7. Hou, J., et al., Cardiac stem cells and their roles in myocardial infarction.
Stem Cell Rev, 2013. 9(3): p. 326‐38.
8. Davis, D.R., et al., Isolation and expansion of functionally‐competent cardiac
progenitor cells directly from heart biopsies. J Mol Cell Cardiol, 2010. 49(2):
p. 312‐21.
9. Peng, A.W., et al., Integrating the biophysical and molecular mechanisms of
auditory hair cell mechanotransduction. Nat Commun, 2011. 2: p. 523.
10. Thomas, D., Gene‐‐environment‐wide association studies: emerging
approaches. Nat Rev Genet, 2010. 11(4): p. 259‐72.
11. Cheng, K.F. and J.Y. Lee, Assessing the joint effect of population stratification
and sample selection in studies of gene‐gene (environment) interactions.
BMC Genet, 2012. 13: p. 5.
182
12. Sul, J.H., et al., Accounting for population structure in gene‐by‐environment
interactions in genome‐wide association studies using mixed models. 2015,
University of California, Los Angeles: Unpublished manuscript.
13. Liao, S.Y., X. Lin, and D.C. Christiani, Gene‐environment interaction effects on
lung function‐ a genome‐wide association study within the Framingham
heart study. Environ Health, 2013. 12: p. 101.
14. Dai, X., et al., A genome‐wide association study for serum bilirubin levels and
gene‐environment interaction in a Chinese population. Genet Epidemiol,
2013. 37(3): p. 293‐300.
15. Zheng, J.S., et al., Genome‐wide contribution of genotype by environment
interaction to variation of diabetes‐related traits. PLoS One, 2013. 8(10): p.
e77442.
16. Hamza, T.H., et al., Genome‐wide gene‐environment study identifies
glutamate receptor gene GRIN2A as a Parkinson's disease modifier gene via
interaction with coffee. PLoS Genet, 2011. 7(8): p. e1002237.
17. Wei, S., et al., Genome‐wide gene‐environment interaction analysis for
asbestos exposure in lung cancer susceptibility. Carcinogenesis, 2012. 33(8):
p. 1531‐7.
18. Ohmen, J., et al., Genome‐wide association study for age‐related hearing
loss (AHL) in the mouse: a meta‐analysis. J Assoc Res Otolaryngol, 2014.
15(3): p. 335‐52.
19. Peters, L.L., et al., The mouse as a model for human biology: a resource
guide for complex trait analysis. Nat Rev Genet, 2007. 8(1): p. 58‐69.
20. Casci, T., Mouse genetics: Fruits of the Collaborative Cross. Nat Rev Genet,
2012. 13(4): p. 223‐223.
21. Threadgill, D.W. and G.A. Churchill, Ten years of the collaborative cross. G3
(Bethesda), 2012. 2(2): p. 153‐6.
22. Bjornsson, H.T., M.D. Fallin, and A.P. Feinberg, An integrated epigenetic and
genetic approach to common human disease. Trends Genet, 2004. 20(8): p.
350‐8.
23. Feinberg, A.P., Phenotypic plasticity and the epigenetics of human disease.
Nature, 2007. 447(7143): p. 433‐40.
183
24. Sutherland, J.E. and M. Costa, Epigenetics and the environment. Ann N Y
Acad Sci, 2003. 983: p. 151‐60.
25. Pufulete, M., et al., Folate status, genomic DNA hypomethylation, and risk of
colorectal adenoma and cancer: a case control study. Gastroenterology,
2003. 124(5): p. 1240‐8.
26. Kaati, G., L.O. Bygren, and S. Edvinsson, Cardiovascular and diabetes
mortality determined by nutrition during parents' and grandparents' slow
growth period. Eur J Hum Genet, 2002. 10(11): p. 682‐8.
27. Liang, L. and W.O. Cookson, Grasping nettles: cellular heterogeneity and
other confounders in epigenome‐wide association studies. Hum Mol Genet,
2014. 23(R1): p. R83‐8.
28. Yue, F., et al., A comparative encyclopedia of DNA elements in the mouse
genome. Nature, 2014. 515(7527): p. 355‐64.
29. Orozco, L.D., et al., Epigenome‐wide association of liver methylation patterns
and complex metabolic traits in mice. Cell Metab, 2015. 21(6): p. 905‐17.

184
REFERENCES AND BIBLIOGRAPHY

Albrecht I, Niesner U, Janke M, Menning A, Loddenkemper C, Kuhl AA, Lepenies I,
Lexberg MH, Westendorf K, Hradilkova K et al. 2010. Persistence of effector
memory Th1 cells is regulated by Hopx. Eur J Immunol 40(11): 2993‐3006.
Allayee H, Andalibi A, Mehrabian M. 2006. Using inbred mouse strains to identify
genes for complex diseases. Frontiers in bioscience : a journal and virtual
library 11: 1216‐1226.
Banfi B, Malgrange B, Knisz J, Steger K, Dubois‐Dauphin M, Krause KH. 2004. NOX3,
a superoxide‐generating NADPH oxidase of the inner ear. The Journal of
biological chemistry 279(44): 46065‐46072.
Baron M. 2001. The search for complex disease genes: fault by linkage or fault by
association? Molecular psychiatry 6(2): 143‐149.
Bedard K, Krause KH. 2007. The NOX family of ROS‐generating NADPH oxidases:
physiology and pathophysiology. Physiological reviews 87(1): 245‐313.
Bennett BJ, Farber CR, Orozco L, Kang HM, Ghazalpour A, Siemers N, Neubauer M,
Neuhaus I, Yordanova R, Guan B et al. 2010. A high‐resolution association
mapping panel for the dissection of complex traits in mice. Genome Res
20(2): 281‐290.
Bennett BJ, Orozco L, Kostem E, Erbilgin A, Dallinga M, Neuhaus I, Guan B, Wang X,
Eskin E, Lusis AJ. 2012. High‐resolution association mapping of
atherosclerosis loci in mice. Arterioscler Thromb Vasc Biol 32(8): 1790‐1798.
Bjornsson HT, Fallin MD, Feinberg AP. 2004. An integrated epigenetic and genetic
approach to common human disease. Trends in genetics : TIG 20(8): 350‐
358.
Blanchard TG, Yu F, Hsieh CL, Redline RW. 2003. Severe inflammation and reduced
bacteria load in murine helicobacter infection caused by lack of phagocyte
oxidase activity. The Journal of infectious diseases 187(10): 1609‐1615.
Boles NC, Lin KK, Lukov GL, Bowman TV, Baldridge MT, Goodell MA. 2011. CD48 on
hematopoietic progenitors regulates stem cells and suppresses tumor
formation. Blood 118(1): 80‐87.
Bond HM, Mesuraca M, Carbone E, Bonelli P, Agosti V, Amodio N, De Rosa G, Di
Nicola M, Gianni AM, Moore MA et al. 2004. Early hematopoietic zinc finger
protein (EHZF), the human homolog to mouse Evi3, is highly expressed in
primitive human hematopoietic cells. Blood 103(6): 2062‐2070.
Bonnet C, Louha M, Loundon N, Michalski N, Verpy E, Smagghe L, Hardelin JP,
185
Rouillon I, Jonard L, Couderc R et al. 2013. Biallelic nonsense mutations in
the otogelin‐like gene (OTOGL) in a child affected by mild to moderate
hearing impairment. Gene 527(2): 537‐540.
Brandes RP, Weissmann N, Schroder K. 2014. Nox family NADPH oxidases:
Molecular mechanisms of activation. Free radical biology & medicine 76:
208‐226.
Brown JR, Goldblatt D, Buddle J, Morton L, Thrasher AJ. 2003. Diminished
production of anti‐inflammatory mediators during neutrophil apoptosis and
macrophage phagocytosis in chronic granulomatous disease (CGD). Journal
of leukocyte biology 73(5): 591‐599.
Buchner, D.A. and Nadeau J.H., Contrasting genetic architectures in different mouse
reference populations used for studying complex traits. Genome Res, 2015.
25(6): p. 775‐791.
Busch K, Klapproth K, Barile M, Flossdorf M, Holland‐Letz T, Schlenner SM, Reth M,
Hofer T, Rodewald HR. 2015. Fundamental properties of unperturbed
haematopoiesis from stem cells in vivo. Nature 518(7540): 542‐546.
Bystrykh L, Weersing E, Dontje B, Sutton S, Pletcher MT, Wiltshire T, Su AI, Vellenga
E, Wang J, Manly KF et al. 2005. Uncovering regulatory pathways that affect
hematopoietic stem cell function using 'genetical genomics'. Nat Genet
37(3): 225‐232.
Casci T. 2012. Mouse genetics: Fruits of the Collaborative Cross. Nat Rev Genet
13(4): 223‐223.
Cervino AC, Darvasi A, Fallahi M, Mader CC, Tsinoremas NF. 2007. An integrated in
silico gene mapping strategy in inbred mice. Genetics 175(1): 321‐333.
Chambers SM, Boles NC, Lin K‐YK, Tierney MP, Bowman TV, Bradfute SB, Chen AJ,
Merchant AA, Sirin O, Weksberg DC et al. 2007. Hematopoietic Fingerprints:
An Expression Database of Stem Cells and Their Progeny. Cell Stem Cell 1(5):
578‐591.
Chen F, Kook H, Milewski R, Gitler AD, Lu MM, Li J, Nazarian R, Schnepp R, Jen K,
Biben C et al. 2002. Hop is an unusual homeobox gene that modulates
cardiac development. Cell 110(6): 713‐723.
Cheng G, Cao Z, Xu X, van Meir EG, Lambeth JD. 2001. Homologs of gp91phox:
cloning and tissue expression of Nox3, Nox4, and Nox5. Gene 269(1‐2): 131‐
140.
Cheng KF, Lee JY. 2012. Assessing the joint effect of population stratification and
sample selection in studies of gene‐gene (environment) interactions. BMC
genetics 13: 5.
186
Churchill GA Airey DC Allayee H Angel JM Attie AD Beatty J Beavis WD Belknap JK
Bennett B Berrettini W et al. 2004. The Collaborative Cross, a community
resource for the genetic analysis of complex traits. Nat Genet 36(11): 1133‐
1137.
Civelek M, Lusis AJ. 2014. Systems genetics approaches to understand complex
traits. Nat Rev Genet 15(1): 34‐48.
Cohen KS, Cheng S, Larson MG, Cupples LA, McCabe EL, Wang YA, Ngwa JS, Martin
RP, Klein RJ, Hashmi B et al. 2013. Circulating CD34+ progenitor cell
frequency is associated with clinical and genetic factors. Blood 121(8): e50‐
e56.
Collaborative Cross C. 2012. The genome architecture of the Collaborative Cross
mouse genetic reference population. Genetics 190(2): 389‐401.
Dai X, Wu C, He Y, Gui L, Zhou L, Guo H, Yuan J, Yang B, Li J, Deng Q et al. 2013. A
genome‐wide association study for serum bilirubin levels and gene‐
environment interaction in a Chinese population. Genetic epidemiology
37(3): 293‐300.
Davis DR, Kizana E, Terrovitis J, Barth AS, Zhang Y, Smith RR, Miake J, Marban E.
2010. Isolation and expansion of functionally‐competent cardiac progenitor
cells directly from heart biopsies. Journal of molecular and cellular
cardiology 49(2): 312‐321.
Davis RC, van Nas A, Bennett B, Orozco L, Pan C, Rau CD, Eskin E, Lusis AJ. 2013.
Genome‐wide association mapping of blood cell traits in mice. Mamm
Genome 24(3‐4): 105‐118.
Davis RL. 2003. Gradients of neurotrophins, ion channels, and tuning in the cochlea.
The Neuroscientist : a review journal bringing neurobiology, neurology and
psychiatry 9(5): 311‐316.
de Haan G, Bystrykh LV, Weersing E, Dontje B, Geiger H, Ivanova N, Lemischka IR,
Vellenga E, Van Zant G. 2002. A genetic and genomic analysis identifies a
cluster of genes associated with hematopoietic cell turnover. Blood 100(6):
2056‐2062.
de Haan G, Nijhof W, Van Zant G. 1997. Mouse Strain‐Dependent Changes in
Frequency and Proliferation of Hematopoietic Stem Cells During Aging:
Correlation Between Lifespan and Cycling Activity. Blood 89(5): 1543‐1550.
de Haan G, Van Zant G. 1997. Intrinsic and extrinsic control of hemopoietic stem cell
numbers: mapping of a stem cell gene. J Exp Med 186(4): 529‐536.
de Haan G, Van Zant G. 1999. Dynamic Changes in Mouse Hematopoietic Stem Cell
Numbers During Aging. Blood 93(10): 3294‐3301.
187
Dobie RA. 2008. The burdens of age‐related and occupational noise‐induced
hearing loss in the United States. Ear and hearing 29(4): 565‐577.
Erway LC, Shiau YW, Davis RR, Krieg EF. 1996. Genetics of age‐related hearing loss in
mice. III. Susceptibility of inbred and F1 hybrid strains to noise‐induced
hearing loss. Hearing research 93(1‐2): 181‐187.
Farber CR, Bennett BJ, Orozco L, Zou W, Lira A, Kostem E, Kang HM, Furlotte N,
Berberyan A, Ghazalpour A et al. 2011. Mouse genome‐wide association and
systems genetics identify Asxl2 as a regulator of bone mineral density and
osteoclastogenesis. PLoS Genet 7(4): e1002038.
Feinberg AP. 2007. Phenotypic plasticity and the epigenetics of human disease.
Nature 447(7143): 433‐440.
Flaherty JP, Fairfield HE, Spruce CA, McCarty CM, Bergstrom DE. 2011. Molecular
characterization of an allelic series of mutations in the mouse Nox3 gene.
Mamm Genome 22(3‐4): 156‐169.
Flint J, Eskin E. 2012. Genome‐wide association studies in mice. Nat Rev Genet
13(11): 807‐817.
Flint J, Mott R. 2008. Applying mouse complex‐trait resources to behavioural
genetics. Nature 456(7223): 724‐727.
Fransen E, Bonneux S, Corneveaux JJ, Schrauwen I, Di Berardino F, White CH,
Ohmen JD, Van de Heyning P, Ambrosetti U, Huentelman MJ et al. 2014.
Genome‐wide association analysis demonstrates the highly polygenic
character of age‐related hearing impairment. European journal of human
genetics : EJHG.
Frati C, Savi M, Graiani G, Lagrasta C, Cavalli S, Prezioso L, Rossetti P, Mangiaracina
C, Ferraro F, Madeddu D et al. 2011. Resident cardiac stem cells. Current
pharmaceutical design 17(30): 3252‐3257.
Frazer KA, Eskin E, Kang HM, Bogue MA, Hinds DA, Beilharz EJ, Gupta RV,
Montgomery J, Morenzoni MM, Nilsen GB et al. 2007. A sequence‐based
variation map of 8.27 million SNPs in inbred mouse strains. Nature
448(7157): 1050‐1053.
Friedman RA, Van Laer L, Huentelman MJ, Sheth SS, Van Eyken E, Corneveaux JJ,
Tembe WD, Halperin RF, Thorburn AQ, Thys S et al. 2009. GRM7 variants
confer susceptibility to age‐related hearing impairment. Human molecular
genetics 18(4): 785‐796.
Geiger H, True JM, de Haan G, Van Zant G. 2001. Age‐ and stage‐specific regulation
patterns in the hematopoietic stem cell hierarchy. Blood 98(10): 2966‐2972.
188
Gerrits A, Dykstra B, Otten M, Bystrykh L, de Haan G. 2008. Combining
transcriptional profiling and genetic linkage analysis to uncover gene
networks operating in hematopoietic stem cells and their progeny.
Immunogenetics 60(8): 411‐422.
Ghazalpour A, Bennett B, Petyuk VA, Orozco L, Hagopian R, Mungrue IN, Farber CR,
Sinsheimer J, Kang HM, Furlotte N et al. 2011. Comparative analysis of
proteome and transcriptome variation in mouse. PLoS Genet 7(6):
e1001393.
Ghazalpour A, Bennett BJ, Shih D, Che N, Orozco L, Pan C, Hagopian R, He A, Kayne
P, Yang WP et al. 2014. Genetic regulation of mouse liver metabolite levels.
Molecular systems biology 10(5): 730.
Ghazalpour A, Rau CD, Farber CR, Bennett BJ, Orozco LD, van Nas A, Pan C, Allayee
H, Beaven SW, Civelek M et al. 2012. Hybrid mouse diversity panel: a panel
of inbred mouse strains suitable for analysis of complex genetic traits.
Mamm Genome 23(9‐10): 680‐692.
Girotto G, Pirastu N, Sorice R, Biino G, Campbell H, d'Adamo AP, Hastie ND, Nutile T,
Polasek O, Portas L et al. 2011. Hearing function and thresholds: a genome‐
wide association study in European isolated populations identifies new loci
and pathways. Journal of medical genetics 48(6): 369‐374.
Girotto G, Vuckovic D, Buniello A, Lorente‐Canovas B, Lewis M, Gasparini P, Steel
KP. 2014. Expression and replication studies to identify new candidate genes
involved in normal hearing function. PloS one 9(1): e85352.
Hamza TH, Chen H, Hill‐Burns EM, Rhodes SL, Montimurro J, Kay DM, Tenesa A,
Kusel VI, Sheehan P, Eaaswarkhanth M et al. 2011. Genome‐wide gene‐
environment study identifies glutamate receptor gene GRIN2A as a
Parkinson's disease modifier gene via interaction with coffee. PLoS Genet
7(8): e1002237.
Hardy J, Singleton A. 2009. Genomewide association studies and human disease.
The New England journal of medicine 360(17): 1759‐1768.
Hartiala J, Bennett BJ, Tang WH, Wang Z, Stewart AF, Roberts R, McPherson R, Lusis
AJ, Hazen SL, Allayee H et al. 2014. Comparative genome‐wide association
studies in mice and humans for trimethylamine N‐oxide, a proatherogenic
metabolite of choline and L‐carnitine. Arterioscler Thromb Vasc Biol 34(6):
1307‐1313.
Heinonen‐Guzejev M, Vuorinen HS, Mussalo‐Rauhamaa H, Heikkila K, Koskenvuo M,
Kaprio J. 2005. Genetic component of noise sensitivity. Twin research and
human genetics : the official journal of the International Society for Twin
Studies 8(3): 245‐249.
189
Henckaerts E, Geiger H, Langer JC, Rebollo P, Van Zant G, Snoeck HW. 2002.
Genetically determined variation in the number of phenotypically defined
hematopoietic progenitor and stem cells and in their response to early‐
acting cytokines. Blood 99(11): 3947‐3954.
Henckaerts E, Langer JC, Snoeck HW. 2004. Quantitative genetic variation in the
hematopoietic stem cell and progenitor cell compartment and in lifespan
are closely linked at multiple loci in BXD recombinant inbred mice. Blood
104(2): 374‐379.
Hesse E, Kiviranta R, Wu M, Saito H, Yamana K, Correa D, Atfi A, Baron R. 2010. Zinc
finger protein 521, a new player in bone formation. Ann N Y Acad Sci 1192:
32‐37.
Hindorff LA, Junkins HA, Hall PN, Mehta JP, Manolio TA. 2014. A catalog of
published genome‐wide association studies. Available at:
http://www.genome.gov/gwastudies/.
Holme RH, Steel KP. 2004. Progressive hearing loss and increased susceptibility to
noise‐induced hearing loss in mice carrying a Cdh23 but not a Myo7a
mutation. Journal of the Association for Research in Otolaryngology : JARO
5(1): 66‐79.
Hope KJ, Cellot S, Ting SB, MacRae T, Mayotte N, Iscove NN, Sauvageau G. 2010. An
RNAi screen identifies Msi2 and Prox1 as having opposite roles in the
regulation of hematopoietic stem cell activity. Cell Stem Cell 7(1): 101‐113.
Hou J, Wang L, Jiang J, Zhou C, Guo T, Zheng S, Wang T. 2013. Cardiac stem cells and
their roles in myocardial infarction. Stem cell reviews 9(3): 326‐338.
Huang Q, Tang J. 2010. Age‐related hearing loss or presbycusis. European archives
of oto‐rhino‐laryngology : official journal of the European Federation of Oto‐
Rhino‐Laryngological Societies 267(8): 1179‐1191.
Hultqvist M, Olofsson P, Holmberg J, Backstrom BT, Tordsson J, Holmdahl R. 2004.
Enhanced autoimmunity, arthritis, and encephalomyelitis in mice with a
reduced oxidative burst due to a mutation in the Ncf1 gene. Proceedings of
the National Academy of Sciences of the United States of America 101(34):
12646‐12651.
Ishibashi K, Suzuki M, Sasaki S, Imai M. 2001. Identification of a new multigene four‐
transmembrane family (MS4A) related to CD20, HTm4 and beta subunit of
the high‐affinity IgE receptor. Gene 264(1): 87‐93.
Ito CY, Li CY, Bernstein A, Dick JE, Stanford WL. 2003. Hematopoietic stem cell and
progenitor defects in Sca‐1/Ly‐6A‐null mice. Blood 101(2): 517‐523.
Jawad M, Cole C, Zanker A, Giotopoulos G, Fitch S, Talbot CJ, Plumb M. 2008. QTL
190
analyses of lineage‐negative mouse bone marrow cells labeled with Sca‐1
and c‐Kit. Mamm Genome 19(3): 190‐198.
Kaati G, Bygren LO, Edvinsson S. 2002. Cardiovascular and diabetes mortality
determined by nutrition during parents' and grandparents' slow growth
period. European journal of human genetics : EJHG 10(11): 682‐688.
Kang HM, Zaitlen NA, Wade CM, Kirby A, Heckerman D, Daly MJ, Eskin E. 2008.
Efficient control of population structure in model organism association
mapping. Genetics 178(3): 1709‐1723.
Karlsson C, Baudet A, Miharada N, Soneji S, Gupta R, Magnusson M, Enver T,
Karlsson G, Larsson J. 2013. Identification of the chemokine CCL28 as a
growth and survival factor for human hematopoietic stem and progenitor
cells. Blood 121(19): 3838‐3842, S3831‐3815.
Keane TM, Goodstadt L, Danecek P, White MA, Wong K, Yalcin B, Heger A, Agam A,
Slater G, Goodson M et al. 2011. Mouse genomic variation and its effect on
phenotypes and gene regulation. Nature 477(7364): 289‐294.
Keenan JI, Peterson RA, 2nd, Hampton MB. 2005. NADPH oxidase involvement in
the pathology of Helicobacter pylori infection. Free radical biology &
medicine 38(9): 1188‐1196.
Kiel MJ, Yilmaz OH, Iwashita T, Yilmaz OH, Terhorst C, Morrison SJ. 2005. SLAM
family receptors distinguish hematopoietic stem and progenitor cells and
reveal endothelial niches for stem cells. Cell 121(7): 1109‐1121.
Kiel MJ, Yilmaz OH, Morrison SJ. 2008. CD150‐ cells are transiently reconstituting
multipotent progenitors with little or no stem cell activity. Blood 111(8):
4413‐4414; author reply 4414‐4415.
Kikuchi H, Hikage M, Miyashita H, Fukumoto M. 2000. NADPH oxidase subunit,
gp91(phox) homologue, preferentially expressed in human colon epithelial
cells. Gene 254(1‐2): 237‐243.
Kiviranta R, Yamana K, Saito H, Ho DK, Laine J, Tarkkonen K, Nieminen‐Pihala V,
Hesse E, Correa D, Maatta J et al. 2013. Coordinated transcriptional
regulation of bone homeostasis by Ebf1 and Zfp521 in both mesenchymal
and hematopoietic lineages. J Exp Med 210(5): 969‐985.
Komatsu J, Koyama H, Maeda N, Aratani Y. 2006. Earlier onset of neutrophil‐
mediated inflammation in the ultraviolet‐exposed skin of mice deficient in
myeloperoxidase and NADPH oxidase. Inflammation research : official
journal of the European Histamine Research Society [et al] 55(5): 200‐206.
Kook H, Epstein JA. 2003. Hopping to the beat. Hop regulation of cardiac gene
expression. Trends Cardiovasc Med 13(7): 261‐264.
191
Kozel PJ, Davis RR, Krieg EF, Shull GE, Erway LC. 2002. Deficiency in plasma
membrane calcium ATPase isoform 2 increases susceptibility to noise‐
induced hearing loss in mice. Hearing research 164(1‐2): 231‐239.
Kujawa SG, Liberman MC. 2009. Adding insult to injury: cochlear nerve
degeneration after "temporary" noise‐induced hearing loss. The Journal of
neuroscience : the official journal of the Society for Neuroscience 29(45):
14077‐14085.
Liang L, Cookson WO. 2014. Grasping nettles: cellular heterogeneity and other
confounders in epigenome‐wide association studies. Human molecular
genetics 23(R1): R83‐88.
Liang Y, Jansen M, Aronow B, Geiger H, Van Zant G. 2007. The quantitative trait
gene latexin influences the size of the hematopoietic stem cell population in
mice. Nat Genet 39(2): 178‐188.
Liao SY, Lin X, Christiani DC. 2013. Gene‐environment interaction effects on lung
function‐ a genome‐wide association study within the Framingham heart
study. Environmental health : a global access science source 12: 101.
Lindblad‐Toh K, Winchester E, Daly MJ, Wang DG, Hirschhorn JN, Laviolette JP,
Ardlie K, Reich DE, Robinson E, Sklar P et al. 2000. Large‐scale discovery and
genotyping of single‐nucleotide polymorphisms in the mouse. Nat Genet
24(4): 381‐386.
Lippert C, Listgarten J, Liu Y, Kadie CM, Davidson RI, Heckerman D. 2011. FaST linear
mixed models for genome‐wide association studies. Nature methods 8(10):
833‐835.
Listgarten J, Lippert C, Kadie CM, Davidson RI, Eskin E, Heckerman D. 2012.
Improved linear mixed models for genome‐wide association studies. Nature
methods 9(6): 525‐526.
Liu P, Vikis H, Lu Y, Wang D, You M. 2007. Large‐scale in silico mapping of complex
quantitative traits in inbred mice. PloS one 2(7): e651.
Liu P, Wang Y, Vikis H, Maciag A, Wang D, Lu Y, Liu Y, You M. 2006. Candidate lung
tumor susceptibility genes identified through whole‐genome association
analyses in inbred mice. Nat Genet 38(8): 888‐895.
Mahajan KN, Mitchell BS. 2003. Role of human Pso4 in mammalian DNA repair and
association with terminal deoxynucleotidyl transferase. Proceedings of the
National Academy of Sciences of the United States of America 100(19):
10746‐10751.
Mamczur P, Sok AJ, Rzechonek A, Rakus D. 2012. Cell cycle‐dependent expression
and subcellular localization of fructose 1,6‐bisphosphatase. Histochemistry
192
and cell biology 137(1): 121‐136.
Mann ZF, Kelley MW. 2011. Development of tonotopy in the auditory periphery.
Hearing research 276(1‐2): 2‐15.
Manolio TA, Brooks LD, Collins FS. 2008. A HapMap harvest of insights into the
genetics of common disease. The Journal of clinical investigation 118(5):
1590‐1605.
Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindorff LA, Hunter DJ, McCarthy MI,
Ramos EM, Cardon LR, Chakravarti A et al. 2009. Finding the missing
heritability of complex diseases. Nature 461(7265): 747‐753.
Morrison SJ, Qian D, Jerabek L, Thiel BA, Park IK, Ford PS, Kiel MJ, Schork NJ,
Weissman IL, Clarke MF. 2002. A genetic determinant that specifically
regulates the frequency of hematopoietic stem cells. J Immunol 168(2): 635‐
642.
Morrison SJ, Weissman IL. 1994. The long‐term repopulating subset of
hematopoietic stem cells is deterministic and isolatable by phenotype.
Immunity 1(8): 661‐673.
Mott R, Flint J. 2013. Dissecting quantitative traits in mice. Annual review of
genomics and human genetics 14: 421‐439.
Mukherjea D, Jajoo S, Kaur T, Sheehan KE, Ramkumar V, Rybak LP. 2010.
Transtympanic administration of short interfering (si)RNA for the NOX3
isoform of NADPH oxidase protects against cisplatin‐induced hearing loss in
the rat. Antioxidants & redox signaling 13(5): 589‐598.
Mukherjea D, Whitworth CA, Nandish S, Dunaway GA, Rybak LP, Ramkumar V.
2006. Expression of the kidney injury molecule 1 in the rat cochlea and
induction by cisplatin. Neuroscience 139(2): 733‐740.
Muller M, von Hunerbein K, Hoidis S, Smolders JW. 2005. A physiological place‐
frequency map of the cochlea in the CBA/J mouse. Hearing research 202(1‐
2): 63‐73.
Muller‐Sieburg CE, Riblet R. 1996. Genetic control of the frequency of
hematopoietic stem cells in mice: mapping of a candidate locus to
chromosome 1. J Exp Med 183(3): 1141‐1150.
Nadeau, J.H., Singer, J. B., Matin, A., Lander, E. S., Analysing complex genetic traits
with chromosome substitution strains. Nat Genet, 2000. 24(3): p. 221‐5.
Nadeau, J.H., Forejt, J., Takada, T., Shiroishi, T., Chromosome substitution strains:
gene discovery, functional analysis, and systems studies. Mamm Genome,
2012. 23(9‐10): p. 693‐705.
193
Oguro H, Ding L, Morrison SJ. 2013. SLAM family markers resolve functionally
distinct subpopulations of hematopoietic stem cells and multipotent
progenitors. Cell Stem Cell 13(1): 102‐116.
Ohlemiller KK, McFadden SL, Ding DL, Flood DG, Reaume AG, Hoffman EK, Scott RW,
Wright JS, Putcha GV, Salvi RJ. 1999. Targeted deletion of the cytosolic
Cu/Zn‐superoxide dismutase gene (Sod1) increases susceptibility to noise‐
induced hearing loss. Audiology & neuro‐otology 4(5): 237‐246.
Ohlemiller KK, McFadden SL, Ding DL, Lear PM, Ho YS. 2000. Targeted mutation of
the gene for cellular glutathione peroxidase (Gpx1) increases noise‐induced
hearing loss in mice. Journal of the Association for Research in
Otolaryngology : JARO 1(3): 243‐254.
Ohmen J, Kang EY, Li X, Joo JW, Hormozdiari F, Zheng QY, Davis RC, Lusis AJ, Eskin E,
Friedman RA. 2014. Genome‐wide association study for age‐related hearing
loss (AHL) in the mouse: a meta‐analysis. Journal of the Association for
Research in Otolaryngology : JARO 15(3): 335‐352.
Olofsson P, Holmberg J, Tordsson J, Lu S, Akerstrom B, Holmdahl R. 2003. Positional
identification of Ncf1 as a gene that regulates arthritis severity in rats. Nat
Genet 33(1): 25‐32.
Orford KW, Scadden DT. 2008. Deconstructing stem cell self‐renewal: genetic
insights into cell‐cycle regulation. Nat Rev Genet 9(2): 115‐128.
Orozco LD, Bennett BJ, Farber CR, Ghazalpour A, Pan C, Che N, Wen P, Qi HX,
Mutukulu A, Siemers N et al. 2012. Unraveling Inflammatory Responses
using Systems Genetics and Gene‐Environment Interactions in
Macrophages. Cell 151(3): 658‐670.
Orozco LD, Morselli M, Rubbi L, Guo W, Go J, Shi H, Lopez D, Furlotte NA, Bennett
BJ, Farber CR et al. 2015. Epigenome‐wide association of liver methylation
patterns and complex metabolic traits in mice. Cell Metab 21(6): 905‐917.
Paffenholz R, Bergstrom RA, Pasutto F, Wabnitz P, Munroe RJ, Jagla W, Heinzmann
U, Marquardt A, Bareiss A, Laufs J et al. 2004. Vestibular defects in head‐tilt
mice result from mutations in Nox3, encoding an NADPH oxidase. Genes &
development 18(5): 486‐491.
Park CC, Gale GD, de Jong S, Ghazalpour A, Bennett BJ, Farber CR, Langfelder P, Lin
A, Khan AH, Eskin E et al. 2011. Gene networks associated with conditional
fear in mice identified using a systems genetics approach. BMC Syst Biol 5:
43.
Parks BW, Nam E, Org E, Kostem E, Norheim F, Hui ST, Pan C, Civelek M, Rau CD,
Bennett BJ et al. 2013. Genetic control of obesity and gut microbiota
194
composition in response to high‐fat, high‐sucrose diet in mice. Cell Metab
17(1): 141‐152.
Peng AW, Salles FT, Pan B, Ricci AJ. 2011. Integrating the biophysical and molecular
mechanisms of auditory hair cell mechanotransduction. Nature
communications 2: 523.
Peters LL, Robledo RF, Bult CJ, Churchill GA, Paigen BJ, Svenson KL. 2007. The mouse
as a model for human biology: a resource guide for complex trait analysis.
Nat Rev Genet 8(1): 58‐69.
Pletcher MT, McClurg P, Batalov S, Su AI, Barnes SW, Lagler E, Korstanje R, Wang X,
Nusskern D, Bogue MA et al. 2004. Use of a dense single nucleotide
polymorphism map for in silico mapping in the mouse. PLoS biology 2(12):
e393.
Pufulete M, Al‐Ghnaniem R, Leather AJ, Appleby P, Gout S, Terry C, Emery PW,
Sanders TA. 2003. Folate status, genomic DNA hypomethylation, and risk of
colorectal adenoma and cancer: a case control study. Gastroenterology
124(5): 1240‐1248.
Purton LE, Scadden DT. 2007. Limiting factors in murine hematopoietic stem cell
assays. Cell Stem Cell 1(3): 263‐270.
Ragu C, Elain G, Mylonas E, Ottolenghi C, Cagnard N, Daegelen D, Passegue E,
Vainchenker W, Bernard OA, Penard‐Lacronique V. 2010. The transcription
factor Srf regulates hematopoietic stem cell adhesion. Blood 116(22): 4464‐
4473.
Ram R, Mehta M, Balmer L, Gatti DM, Morahan G. 2014. Rapid identification of
major‐effect genes using the collaborative cross. Genetics 198(1): 75‐86.
Rich IN. 2015. Measurement of hematopoietic stem cell proliferation, self‐renewal,
and expansion potential. Methods in molecular biology 1235: 7‐17.
Ringwald M, Iyer V, Mason JC, Stone KR, Tadepally HD, Kadin JA, Bult CJ, Eppig JT,
Oakley DJ, Briois S et al. 2011. The IKMC web portal: a central point of entry
to data and resources from the International Knockout Mouse Consortium.
Nucleic acids research 39(Database issue): D849‐855.
Rossi L, Lin KK, Boles NC, Yang L, King KY, Jeong M, Mayle A, Goodell MA. 2012. Less
is more: unveiling the functional core of hematopoietic stem cells through
knockout mice. Cell Stem Cell 11(3): 302‐317.
Ruben RJ. 1967. Development of the inner ear of the mouse: a radioautographic
study of terminal mitoses. Acta oto‐laryngologica: Suppl 220:221‐244.
Rybak LP, Mukherjea D, Jajoo S, Kaur T, Ramkumar V. 2012. siRNA‐mediated knock‐
195
down of NOX3: therapy for hearing loss? Cellular and molecular life sciences:
CMLS 69(14): 2429‐2434.
Sa, Q., Hart, E., Hill, A. E., Nadeau, J. H., Hoover‐Plow, J. L. Quantitative trait locus
analysis for hemostasis and thrombosis. Mamm Genome, 2008. 19(6): p.
406‐12.
Schroder K, Zhang M, Benkhoff S, Mieth A, Pliquett R, Kosowski J, Kruse C, Luedike
P, Michaelis UR, Weissmann N et al. 2012. Nox4 is a protective reactive
oxygen species generating vascular NADPH oxidase. Circulation research
110(9): 1217‐1225.
Shin CH, Liu ZP, Passier R, Zhang CL, Wang DZ, Harris TM, Yamagishi H, Richardson
JA, Childs G, Olson EN. 2002. Modulation of cardiac growth and
development by HOP, an unusual homeodomain protein. Cell 110(6): 725‐
735.
Skarnes WC, Rosen B, West AP, Koutsourakis M, Bushell W, Iyer V, Mujica AO,
Thomas M, Harrow J, Cox T et al. 2011. A conditional knockout resource for
the genome‐wide study of mouse gene function. Nature 474(7351): 337‐
342.
Sliwinska‐Kowalska M, Pawelczyk M. 2013. Contribution of genetic factors to noise‐
induced hearing loss: a human studies review. Mutation research 752(1): 61‐
65.
Snelgrove RJ, Edwards L, Rae AJ, Hussell T. 2006. An absence of reactive oxygen
species improves the resolution of lung influenza infection. Eur J Immunol
36(6): 1364‐1373.
Solberg Woods LC. 2014. QTL mapping in outbred populations: successes and
challenges. Physiological genomics 46(3): 81‐90.
Son EJ, Wu L, Yoon H, Kim S, Choi JY, Bok J. 2012. Developmental gene expression
profiling along the tonotopic axis of the mouse cochlea. PloS one 7(7):
e40735.
Spangrude GJ, Brooks DM. 1993. Mouse strain variability in the expression of the
hematopoietic stem cell antigen Ly‐6A/E by bone marrow cells. Blood
82(11): 3327‐3332.
Stucken EZ, Hong RS. 2014. Noise‐induced hearing loss: an occupational medicine
perspective. Current opinion in otolaryngology & head and neck surgery
22(5): 388‐393.
Sul JH, Yang W‐Y, Kostem E, Furlotte N, He D, Eskin E. 2015. Accounting for
population structure in gene‐by‐environment interactions in genome‐wide
association studies using mixed models. University of California, Los Angeles,
196
Unpublished manuscript.
Sutherland JE, Costa M. 2003. Epigenetics and the environment. Ann N Y Acad Sci
983: 151‐160.
Svenson KL, Gatti DM, Valdar W, Welsh CE, Cheng R, Chesler EJ, Palmer AA,
McMillan L, Churchill GA. 2012. High‐resolution genetic mapping using the
Mouse Diversity outbred population. Genetics 190(2): 437‐447.
Tada H, Won HH, Melander O, Yang J, Peloso GM, Kathiresan S. 2014. Multiple
associated variants increase the heritability explained for plasma lipids and
coronary artery disease. Circulation Cardiovascular genetics 7(5): 583‐587.
Takeda N, Jain R, Leboeuf MR, Padmanabhan A, Wang Q, Li L, Lu MM, Millar SE,
Epstein JA. 2013. Hopx expression defines a subset of multipotent hair
follicle stem cells and a progenitor population primed to give rise to K6+
niche cells. Development 140(8): 1655‐1664.
Takeda N, Jain R, LeBoeuf MR, Wang Q, Lu MM, Epstein JA. 2011. Interconversion
between intestinal stem cell populations in distinct niches. Science
334(6061): 1420‐1424.
Thomas D. 2010. Gene‐‐environment‐wide association studies: emerging
approaches. Nat Rev Genet 11(4): 259‐272.
Threadgill DW, Churchill GA. 2012. Ten years of the collaborative cross. G3 2(2):
153‐156.
Valdar W, Solberg LC, Gauguier D, Burnett S, Klenerman P, Cookson WO, Taylor MS,
Rawlins JN, Mott R, Flint J. 2006. Genome‐wide genetic association of
complex traits in heterogeneous stock mice. Nat Genet 38(8): 879‐887.
van der Harst P Zhang W Mateo Leach I Rendon A Verweij N Sehmi J Paul DS Elling U
Allayee H Li X et al. 2012. Seventy‐five genetic loci influencing the human
red blood cell. Nature 492(7429): 369‐375.
Van Laer L, Huyghe JR, Hannula S, Van Eyken E, Stephan DA, Maki‐Torkko E, Aikio P,
Fransen E, Lysholm‐Bernacchi A, Sorri M et al. 2010. A genome‐wide
association study for age‐related hearing impairment in the Saami. European
journal of human genetics : EJHG 18(6): 685‐693.
van Os R, Ausema A, Noach EJ, van Pelt K, Dontje BJ, Vellenga E, de Haan G. 2006.
Identification of quantitative trait loci regulating haematopoietic parameters
in B6AKRF2 mice. Br J Haematol 132(1): 80‐90.
Vlajkovic SM, Lin SC, Wong AC, Wackrow B, Thorne PR. 2013. Noise‐induced
changes in expression levels of NADPH oxidases in the cochlea. Hearing
research 304: 145‐152.
197
Wei S, Wang LE, McHugh MK, Han Y, Xiong M, Amos CI, Spitz MR, Wei QW. 2012.
Genome‐wide gene‐environment interaction analysis for asbestos exposure
in lung cancer susceptibility. Carcinogenesis 33(8): 1531‐1537.
Welter D, MacArthur J, Morales J, Burdett T, Hall P, Junkins H, Klemm A, Flicek P,
Manolio T, Hindorff L et al. 2014. The NHGRI GWAS Catalog, a curated
resource of SNP‐trait associations. Nucleic acids research 42(Database
issue): D1001‐1006.
White CH, Ohmen JD, Sheth S, Zebboudj AF, McHugh RK, Hoffman LF, Lusis AJ, Davis
RC, Friedman RA. 2009. Genome‐wide screening for genetic loci associated
with noise‐induced hearing loss. Mamm Genome 20(4): 207‐213.
White JK, Gerdin AK, Karp NA, Ryder E, Buljan M, Bussell JN, Salisbury J, Clare S,
Ingham NJ, Podrini C et al. 2013. Genome‐wide generation and systematic
phenotyping of knockout mice reveals new roles for many genes. Cell
154(2): 452‐464.
Wiltshire T, Pletcher MT, Batalov S, Barnes SW, Tarantino LM, Cooke MP, Wu H,
Smylie K, Santrosyan A, Copeland NG et al. 2003. Genome‐wide single‐
nucleotide polymorphism analysis defines haplotype patterns in mouse.
Proceedings of the National Academy of Sciences of the United States of
America 100(6): 3380‐3385.
Yalcin B, Fullerton J, Miller S, Keays DA, Brady S, Bhomra A, Jefferson A, Volpi E,
Copley RR, Flint J et al. 2004. Unexpected complexity in the haplotypes of
commonly used inbred strains of laboratory mice. Proceedings of the
National Academy of Sciences of the United States of America 101(26): 9734‐
9739.
Yamashita K, Katoh H, Watanabe M. 2013. The homeobox only protein homeobox
(HOPX) and colorectal cancer. Int J Mol Sci 14(12): 23231‐23243.
Yang H, Ding Y, Hutchins LN, Szatkiewicz J, Bell TA, Paigen BJ, Graber JH, de Villena
FP, Churchill GA. 2009. A customized and versatile high‐density genotyping
array for the mouse. Nature methods 6(9): 663‐666.
Yang H, Wang JR, Didion JP, Buus RJ, Bell TA, Welsh CE, Bonhomme F, Yu AH,
Nachman MW, Pialek J et al. 2011. Subspecific origin and haplotype diversity
in the laboratory mouse. Nat Genet 43(7): 648‐655.
Yankaskas K. 2013. Prelude: noise‐induced tinnitus and hearing loss in the military.
Hearing research 295: 3‐8.
Yue F Cheng Y Breschi A Vierstra J Wu W Ryba T Sandstrom R Ma Z Davis C Pope BD
et al. 2014. A comparative encyclopedia of DNA elements in the mouse
genome. Nature 515(7527): 355‐364.
198
Zhang W, Korstanje R, Thaisz J, Staedtler F, Harttman N, Xu L, Feng M, Yanas L, Yang
H, Valdar W et al. 2012. Genome‐wide association mapping of quantitative
traits in outbred mice. G3 2(2): 167‐174.
Zhao X, Jones SM, Yamoah EN, Lundberg YW. 2008. Otoconin‐90 deletion leads to
imbalance but normal hearing: a comparison with other otoconia mutants.
Neuroscience 153(1): 289‐299.
Zheng JS, Arnett DK, Lee YC, Shen J, Parnell LD, Smith CE, Richardson K, Li D, Borecki
IB, Ordovas JM et al. 2013. Genome‐wide contribution of genotype by
environment interaction to variation of diabetes‐related traits. PloS one
8(10): e77442.
Zuk O, Hechter E, Sunyaev SR, Lander ES. 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.
199

Abstract (if available)

Abstract Genome‐wide studies in humans focused on the discovery of genetic mechanisms involved in common, complex phenotypes have enjoyed a moderate degree of success, but are limited by a number of factors. These include statistical power, feasibility of functional validation of candidate genes, and ability to collect relevant biological tissue for the trait. These issues can be addressed by performing similar analyses in mouse models, namely, by using a platform termed the Hybrid Mouse Diversity Panel (HMDP) that was developed for such studies. The HMDP has been used in a range of genome‐wide association studies (GWAS) for complex traits, and in this dissertation, I describe three additional applications for the panel. Here, we have validated novel genes regulating hematopoietic stem cell frequency and noise induced hearing loss, and identified a number of loci that are significantly associated with hearing.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Functional characterization of colorectal cancer GWAS loci

PDF

Understanding acute lymphoblastic leukemia in different ethnic groups in the United States

PDF

The cancer stem-like phenotype: therapeutics, phenotypic plasticity and mechanistic studies

PDF

Role of the bone marrow niche components in B cell malignancies

PDF

Identification and characterization of cancer-associated enhancers

PDF

Influence of genetic background on the effect of traffic-related air pollution on obesity phenotypes

PDF

The role of endoplasmic reticulum chaperones in regulating hematopoietic stem cells and hematological malignancies

PDF

Investigating the role of STAT3 in mouse and rat embryonic stem cell self-renewal and differentiation

PDF

Understanding the genetic architecture of complex traits

PDF

Functional characterization of colon cancer risk-associated enhancers: connecting risk loci to risk genes

PDF

Exploring stem cell pluripotency through long range chromosome interactions

PDF

Genetics and the environment: evaluating the role of noncoding RNA in autism spectrum disorder

PDF

Integrated genomic & epigenomic analyses of glioblastoma multiforme: Methods development and application

PDF

Identification and characterization of adult stem cells in the oral cavity

PDF

Role of STAT3 phosphorylation in mouse embryonic stem cell self-renewal and differentiation

PDF

Genetic studies of inflammation and cardiovascular disease

PDF

Novel roles for Maf1 in embryonic stem cell differentiation and adipogenesis

PDF

Integrative genomic and epigenomic analysis of human cancer

PDF

Genetic and dietary determinants of nonalcoholic fatty liver disease in Hispanic children

PDF

Regional localization and regulation of hematopoietic stem cells in the bone marrow stem cell niche