Genetic Resistance to Diet-Induced Obesity in Mice

GENETIC RESISTANCE TO DIET-INDUCED OBESITY IN MICE

LINDSAY CATHERINE BURRAGE

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Thesis Advisors: Dr. Joseph H. Nadeau Dr. Colleen M. Croniger

Department of Genetics

CASE WESTERN RESERVE UNIVERSITY

August, 2006

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

______

candidate for the Ph.D. degree *.

(signed)______(chair of the committee)

______

(date) ______

*We also certify that written approval has been obtained for any proprietary material contained therein.

TABLE OF CONTENTS

LIST OF TABLES...... 4 LIST OF FIGURES...... 8 ACKNOWLEDGEMENTS ...... 12 LIST OF ABBREVIATIONS...... 14 ABSTRACT...... 19 CHAPTER I: INTRODUCTION AND SUMMARY OF RESEARCH AIMS. 21 A. Obesity: An introduction...... 22 1. Historical perspective...... 22 2. Definition and prevalence of human obesity...... 22 3. Causes of human obesity...... 26 4. Pathological consequences of obesity...... 28 5. Physiologic control of energy balance and body weight. . . 32 6. Treatment of human obesity...... 35 B. Genetics of obesity...... 36 1. An introduction to obesity genetics...... 36 2. Genetic forms of obesity...... 37 C. Methods for investigating the genetics of obesity...... 42 1. Genetic studies of obesity in humans...... 42 2. Genetic studies of obesity in mouse...... 43 D. C57BL/6J and A/J inbred strains: Mouse models for obesity. . . . 52 1. Differential susceptibility to diet-induced obesity in C57BL/6J and A/J male mice...... 52 2. Energy metabolism in C57BL/6J and A/J male mice. . . . 54 3. Genetic studies of obesity in C57BL/6J and A/J inbred strains...... 55 4. C57BL/6J and A/J: Advantages for complex trait studies. . 56 5. B6 –ChrA CSSs accelerate QTL mapping studies in C57BL/6J and A/J strains...... 57 E. Summary and research aims...... 63

CHAPTER II: GENETIC DISSECTION OF OBESITY IN CHROMOSOME SUBSTITUTION STRAINS...... 66 A. Introduction...... 67 B. Materials and Methods...... 69 C. Results...... 78 D. Discussion...... 110

CHAPTER III: GENETIC DISSECTION OF OBESITY RESISTANCE ON A/J-DERIVED CHROMOSOME 6...... 118 A. Introduction...... 119 B. Materials and Methods...... 123 C. Results...... 129 D. Discussion...... 147

CHAPTER IV: PHENOTYPIC DISSECTION OF RESISTANCE TO DIET-INDUCED OBESITY IN B6-CHR 6A CHROMOSOME SUBSTITUTION STRAIN MICE...... 153 A. Introduction...... 154 B. Materials and Methods...... 156 C. Results...... 160 D. Discussion...... 181

CHAPTER V: DISCUSSION AND FUTURE DIRECTIONS...... 188 A. Discussion...... 189 1. Multiple genes and gene interactions influence obesity resistance...... 189 2. CSSs provide models for studying obesity resistance and related traits...... 192 3. Evaluation of CSS intercrosses as a method for QTL localization ...... 193 B. Future Directions...... 196 1. How do we identify obesity resistance genes detected in CSSs?...... 196 2. How do we identify obesity resistance genes which did not produce peaks in the CSS intercrosses?...... 198 3. What mechanism explains the resistance to diet-induced obesity in CSSs and congenic strains?...... 202

APPENDICES...... 205 A. Appendix I: Genetic markers used in CSS intercross mapping studies...... 205 B. Appendix II: Correlations among traits in CSS intercross progeny...... 212 B. Appendix III: Multiple QTL analysis in CSS intercross progeny. 233 C. Appendix IV: Genetic and phenotypic analyses of Obrq2. . . . . 242

REFERENCES...... 264

LIST OF TABLES

CHAPTER II

II-1. Composition of high-fat and low-fat diets...... 72

II-2. Three-way ANOVA tables for C57BL/6J and A/J fed diets with varied fat

and carbohydrate composition...... 81

II-3. Initial HFSC diet CSS survey...... 83

II-4. LFCC diet CSS survey...... 85

II-5. HFSC vs. LFCC diet CSS survey analysis...... 88

II-6. Comparison of replicate HFSC diet CSS surveys...... 91

II-7. HFSC diet replicate CSS survey #1...... 92

II-8. HFSC diet replicate CSS survey #2...... 93

II-9. Trait correlations in B6-ChrA CSS F2 crosses...... 96

II-10. LOD scores for B6-ChrA F2 genome scan with p values adjusted for traits

analyzed...... 97

II-11. LOD scores from B6-ChrA F2 genome scan with p values adjusted for

number of crosses and number of traits analyzed...... 99

II-12. Support intervals for significant and suggestive QTLs...... 100

II-13. QTL inheritance patterns...... 103

CHAPTER III

III-1. Markers used for congenic panel construction...... 126

III-2. FW in the HFSC diet congenic strain survey...... 135

III-3. Final BMI for the HFSC diet congenic strain survey...... 136

III-4. 62.9-A and 62-BL congenic strain replicate analyses...... 137 6

III-5. 62-BL reciprocal cross analysis...... 146

CHAPTER IV

IV-1. HFSC diet consumption in C57BL/6J, A/J, and B6-Chr 6A males...... 167

IV-2. Blood chemistry and liver triglycerides in C57BL/6J (B6), A/J, and

B6-Chr 6A at 85 days of age (50 days of high-fat diet consumption). . . . 169

IV-3. Blood chemistry and liver triglycerides in C57BL/6J (B6), A/J, and

B6-Chr 6A at 135 days of age (100 days of high-fat diet consumption). . . 170

IV-4. Metabolic phenotyping in 62-B congenic strain...... 175

IV-5. Sterol measurements in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and 62-B

congenic strain...... 177

IV-6. Amino acid measurements in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and

62-B congenic strain...... 179

IV-7. Acylcarnitine profile in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and 62-B

congenic strain...... 180

APPENDIX III

AIII-1. Two dimensional genome scan in B6-ChrA CSS F2 crosses...... 236

APPENDIX IV

AIV-1. Composition of LabDiet 5010...... 246

AIV-2. FW and BMI for 92-A vs. 62-BL after 100 days of 5010 or HFSC diet

consumption...... 253 7

AIV-3. HFSC diet consumption in 92-A vs. 62-BL males...... 256

AIV-4. Blood chemistry in 92-A vs. 62-BL male mice fed the HFSC diet for 28

or 100 days...... 257

AIV-5. 62-BL subcongenic HFSC diet survey...... 262

LIST OF FIGURES

CHAPTER I

I-1. Energy balance...... 24

I-2. Leptin and the melanocortin pathway...... 34

I-3. B6-ChrA CSS panel construction...... 58

I-4. QTL mapping with CSSs...... 60

CHAPTER II

II-1. Time course for body weight studies...... 73

II-2. C57BL/6J (B6) and A/J weight gain on diets that differ in fat and

carbohydrate composition...... 79

II-3. C57BL/6J (B6) and A/J body weight (MW, FW, EWG, FWG, and WG)

when fed diets that differ in fat and carbohydrate composition...... 80

II-4. Correlations between replicate HFSC diet survey traits...... 89

II-5. B6-ChrA CSS whole genome scan...... 101

II-6. Comparison of CSS surveys and CSS whole genome scan analysis ...... 104

II-7. FW for B6-ChrA CSS F2 progeny vs. C57BL/6J (B6)...... 106

II-8. Maternal inheritance was not detected in B6-MitoA ...... 109

CHAPTER III

III-1. Congenic panel derived from B6-Chr 6A...... 124

III-2. FW and BMI in the F1 [B6-Chr 6A (CSS-6) x C57BL/6J (B6)] male mice. 131

III-3. FW and BMI in the HFSC diet congenic strain survey...... 134

III-4. 62-BL congenic strain HFSC diet replicate analysis...... 138 10

III-5. Obrq1 critical interval...... 141

III-6. Obrq2 critical interval...... 143

III-7. Obrq3 critical interval...... 144

III-8. Summary of QTLs discovered on chromosome 6 ...... 145

CHAPTER IV

IV-1. Gonadal fat pad weights and BMI in C57BL/6J (B6) relative to A/J and

B6-Chr 6A (CSS-6) fed the HFSC diet for 100 days...... 161

IV-2. Pearson’s correlation coefficients for fat pad weight vs. final body weight

and fat pad weight vs. BMI in B6-Chr 6A males fed the HFSC diet. . . . . 162

IV-3. HFSC diet consumption in C57BL/6J (B6), A/J, and B6-Chr 6A (CSS-6)

males...... 164

IV-4. B6-Chr 6A (CSS-6) is resistant to the development of fatty liver after 50

days of HFSC diet consumption...... 172

IV-5. B6-Chr 6A (CSS-6) is resistant to the development of fatty liver after 100

days of HFSC diet consumption...... 173

IV-6. Summary of adiposity, food intake, blood chemistry, and liver triglyceride

analyses in C57BL/6J (B6) and B6-Chr 6A male mice fed the HFSC diet

for 100 days...... 183

IV-7. Summary of adiposity, food intake, blood chemistry, and liver triglyceride

analyses in C57BL/6J (B6) and 62-B male mice fed the HFSC diet for 100

days...... 185

CHAPTER V

V-1. Methods for investigating the mechanism of obesity resistance...... 204

APPENDIX III

AIII-1. Two dimensional genome scan in the B6-ChrA F2 crosses...... 240

APPENDIX IV

AIV-1. 62-B subcongenic panel...... 249

AIV-2. 92-A vs. 62-BL growth curves on 5010 and HFSC diets...... 252

AIV-3. HFSC diet consumption in 92-A vs. 62-BL male mice...... 255

AIV-4. 62-BL subcongenic HFSC diet survey...... 261

AIV-5. F2 male progeny derived from B6-Chr 6A and 62-BL mice...... 263

ACKNOWLEDGEMENTS

My interest in the genetics of metabolism led me to the laboratory of Dr. Joseph

Nadeau in the spring of 2000. Since then, Dr. Nadeau has taught me to ask questions, pursue answers, and think and work independently. During the course of training to become an independent scientist, working in Dr. Nadeau’s laboratory has taught me to make my own scientific decisions, to take scientific risks, to enjoy my own success, and to “live with” the consequences of my mistakes.

I spent the summer of 2000 exploring the genetics of PEPCK in Dr. Richard

Hanson’s laboratory under the guidance of Dr. Colleen Croniger. Dr. Croniger was a rotation mentor, a collaborator, and eventually a co-mentor on my thesis work. Dr.

Croniger’s patience, dedication, and sincere belief in my abilities motivated me in the final stages of my thesis work. Dr. Croniger is an amazing role model and I am grateful that she agreed to serve as a co-mentor for this project.

I could never have completed this work without the support of all of the former and current members of the Nadeau lab. In particular, I am grateful for the guidance, mentorship, and advice provided by Dr. David Sinasac and the friendship and collaboration of Annie Hill. I am also grateful for the technical assistance provided by two summer students, Nicole Nadeau and Christine Jimenez, and the assistance and support provided by the staff of the genetics department. I am also indebted to Kevin

Jimenez and Lonnie Thomas for their wonderful care of our mouse colony.

My thesis committee, Dr. Matthew Warman, Dr. Mark Adams, Dr. Richard

Hanson, and Dr. Arthur Zinn, provided endless guidance and support for this work.

Furthermore, I am indebted to my collaborators, Dr. Mark Daly, Andrew Kirby, Dr. Karl 13

Broman, Dr. Gary Churchill, and Dr. Keith Shockley. In addition, I am grateful to Dr.

Clifford Harding and the MSTP for support and guidance and to my clinical tutorial mentors, Dr. James Leslie and Dr. Debra Leizman, for allowing me to spend time in their clinics. In particular, in Dr. Leizman’s medicine clinic, I was exposed to obesity on a regular basis and was constantly reminded of the significance of this work.

My introduction to the field of genetics occurred in the summer of 1998 and I am forever indebted to the late Drs. Emmanuel Shapira and James Miller for introducing me to the genetics of metabolism and to the members of the Hayward Genetics Center for serving as my first genetics teachers. I am always grateful to my undergraduate thesis mentor, Dr. Hans Andersson, an excellent teacher and mentor, for his encouragement to pursue this PhD and for his continued advice and guidance along the way.

I could never have survived my PhD without the support of several wonderful friends in Cleveland including Usha Narayanan, Jonathan Moseley, Emma Larkin, Davis

Ryman, Dolly Padovani, and Jason Heaney. I also must acknowledge my dear friend,

Tanya Rege, who has provided endless support and encouragement during the PhD process and during the last 14 years of our friendship.

Most importantly, I acknowledge the support and guidance provided by my family. As educators, my parents have always challenged me to pursue knowledge and motivated me to pursue my dreams.

Lastly, my thesis work is dedicated to the memory of my grandfather, Irby T.

Baudouin, Jr., whose belief in me provided motivation during the challenges of completing this thesis and to the hundreds who lost their lives in my beloved hometown of New Orleans, Louisiana during Hurricane Katrina. 14

LIST OF ABBREVIATIONS

ACTH adrenocorticotropic hormone

AGRP agouti-related protein

ALT alanine aminotransferase

α-MSH α-melanocyte stimulating hormone

ANOVA analysis of variance

AST aspartate aminotransferase

ATP adenosine 5'-triphosphate

ATP III Adult Treatment Panel III

B6-Chr #A C57BL/6J-Chr #A/NaJ; # refers to substituted chromosome

BAC bacterial artificial chromosome

BMI body mass index

bp base pair

BUN blood urea nitrogen

cc cubic centimeter (milliliter)

CCK cholecystokinin

cM centimorgan

cm centimeter

CO2 carbon dioxide

CPE carboxypeptidase E

CSS chromosome substitution strain

CWRU Case Western Reserve University df degrees of freedom dl deciliter 16

DNA deoxyribonucleic acid dNTP deoxynucleotide triphosphate

ELISA enzyme-linked immunosorbent assay

EWG mean weight gained in first half of diet study (~first 55 days)

F1 first filial generation

F2 intercross progeny, second filial generation

FW final body weight (after 100 days of diet consumption)

FWG mean weight gained in second half of the diet study (~ second 45 days)

H&E hematoxylin and eosin

HDL high density lipoprotein

HFCC high-fat, complex carbohydrate

HFSC high-fat, simple carbohydrate

IW initial body weight (at 35 days of age, prior to diet study)

GLP-1 glucagon-like peptide kcal kilocalorie kg kilogram

KOH potassium hydroxide

LDL low density lipoprotein

LFCC low-fat, complex carbohydrate

LFSC low-fat, simple carbohydrate

LOD logarithm of odds m meter

Mb megabase of DNA 17

MC1R melanocortin-1 receptor

MC3R melanocortin-3 receptor

MC4R melanocortin-4 receptor

mg milligram

MgCl2 magnesium chloride

mM millimolar

mmHg millimeters of mercury

mmol/L millimoles per liter

MW mid-point body weight (after ~55 days of diet consumption) nd not determined

NEAT non-exercise associated (or activity) thermogenesis

NHANES National Health and Nutrition Examination Survey nM nanomolar

NPY neuropeptide Y

Obrq obesity resistance QTL

OXM oxyntomodulin

PCR polymerase chain reaction

POMC pro-opiomelanocortin

PP pancreatic peptide

PYY 3-36 peptide YY

QTL quantitative trait locus

SNP single nucleotide polymorphism std dev standard deviation 18

Taq Thermus aquaticus

UCPs uncoupling proteins

UCP-1 uncoupling protein-1

UCP-2 uncoupling protein-2

U.S. United States of America

µl microliter

µM micromolar

VLDL very low density lipoprotein

WG mean weight gained per day across 100 day diet study

Genetic Resistance to Diet-Induced Obesity in Mice

Abstract

LINDSAY C. BURRAGE

Obesity is a major public health concern in the 21st century. Approximately one-

third of U.S. adults are obese, with the prevalence of obesity and related conditions like

hypercholesterolemia, hypertension, and diabetes mellitus, rising at an alarming rate.

Although hereditary factors are known to influence obesity, genetic studies of human

obesity are complicated by numerous environmental factors such as diet and activity level that also influence body weight. An alternative approach using inbred strains of mice which differ in their susceptibility to high-fat, diet-induced obesity, can be useful for dissecting the genetics of this highly complex trait.

To investigate the genetics of high-fat, diet-induced obesity in the obesity susceptible C57BL/6J inbred strain and obesity resistant A/J inbred strain, we used the

B6-ChrA chromosome substitution strains (CSSs), a panel of inbred strains with individual A/J-derived chromosomes substituted onto the C57BL/6J genetic background.

A CSS body weight survey identified 13 A/J chromosomes that reproducibly confer resistance to high-fat, diet-induced obesity, suggesting that at least 13 obesity resistance 20

genes were detected. To determine the number and location of resistance genes on A/J

chromosomes, intercross progeny derived from each CSS were analyzed. The intercross

studies detected resistance QTLs on only three A/J chromosomes and surprisingly

detected obesity promoting QTLs on two A/J chromosomes. In addition, more detailed

investigation of obesity resistance on A/J chromosome 6 with congenic strains derived

from the B6-Chr 6A strain revealed evidence for at least three genes influencing obesity

resistance. Collectively, our studies lend support to the theory that genetic resistance to

high-fat, diet-induced obesity in our model is complex, even when narrowing our studies from the whole genome to a single chromosome level. Furthermore, phenotyping studies of B6-Chr 6A and a congenic strain derived from it also revealed resistance to several

metabolic syndrome phenotypes in addition to obesity. Thus, these CSSs and congenic

strains derived from them will enable the genetic dissection of obesity and metabolic

syndrome, and simultaneous genetic and phenotypic studies using these strains will lead

to the discovery of many obesity resistance genes and possibly novel therapeutic targets

for obesity.

CHAPTER I: INTRODUCTION AND SUMMARY OF RESEARCH AIMS

A. OBESITY: AN INTRODUCTION

1. Historical perspective

Obesity is a major public health problem in the 21st century. The high prevalence

of obesity and the associated morbidity and mortality dominate the headlines of

newspapers, magazines, and television. Despite its recent pervasiveness, obesity is not a

modern medical condition. According to George Bray, who has written extensive

reviews on the history of obesity, the earliest depictions of obesity are statues of obese

females discovered in Europe that are believed to have been created 25,000 years ago

(BRAY 1990). An early acknowledgement of the morbidity and mortality linked to obesity is in the works of the ancient physician Hippocrates who wrote that obese individuals “are apt to die earlier than those who are slender” (HIPPOCRATES 1939). Both

Hippocrates and Galen provide early indications that diet and exercise are therapeutic

strategies for treating obesity (BRAY 1990; CHRISTOPOULOU-ALETRA and

PAPAVRAMIDOU 2004; PAPAVRAMIDOU et al. 2004). Some of the first full papers on

obesity, according to Bray, were written in the 16th and 17th century, and in 1727,

Thomas Short wrote “I believe no age did ever afford more instances of corpulency than

our own” (SHORT 1727). Two-hundred and seventy-five years later, obesity experts are

making similar observations in an age where obesity is considered a worldwide epidemic.

2. Definition and prevalence of human obesity

The vocabulary describing obesity has evolved through history. For example,

terms for obesity have included “simple gluttony,” “corpulence,” “fat,” “polysarcia,”

“stoutliness” and “portliness.” Many of these terms have a negative connotation due to 23

the social stigma often associated with obesity. According to The American Heritage

Dictionary, the formal definition of obese is “extremely fat” or “corpulent”’ (1969).

The excessive accumulation of body fat in obese patients results from

disturbances in energy balance (Figure I-1). According to the basic laws of

thermodynamics, when energy consumption balances energy utilization, no excess fat is

stored. In contrast, when energy consumption exceeds energy utilization, the excess

energy is stored in the form of body fat or adipose tissue, and obesity results. This body

fat is stored in many areas of the human body, but the largest fat depots are usually in

subcutaneous, retroperitoneal, and visceral locations.

Clinically, the definition of obesity varies. The most commonly used measure of

obesity is BMI (body mass index), which corrects body weight for height. BMI is

calculated by dividing body weight in kilograms by height in meters squared (HEALTH

1998; WHO 1997). Both the National Heart, Lung, and Blood Institute and the World

2 Health Organization define obesity as a BMI greater than 30 kg/m (HEALTH 1998; WHO

2 1997). A BMI between 25 and 30 kg/m defines overweight (HEALTH 1998; WHO

1997). One disadvantage of BMI measurements is that BMI does not take into account fat mass vs. lean muscle mass, and some individuals with high muscle mass, such as athletes, are misclassified as obese using these definitions. Other measures of obesity include estimations of total body fat obtained using bioimpedance, dual x-ray absorptiometry, and various imaging modalities (WILLETT et al. 1999).

Bioimpedance uses electrical conductivity to estimate total fat mass. The ability

Figure I-1. Energy balance. A. When energy intake balances expenditure, no weight is gained or lost. B. Alternatively, if intake exceeds expenditure, weight gain occurs. C. If expenditure exceeds intake, weight loss occurs. Individual variation in the metabolic efficiency of fuel conversion into energy may alter the relationship between expenditure and intake (“fulcrum” of the balance).

Body Weight

Intake Expenditure

Gain Loss

Body Weight Expenditure

Intake Gain Loss

Intake Body Weight

Gain Loss Expenditure

of body tissues to conduct electrical current depends on the content of free electrolytes,

which is positively correlated with tissue water content. Because fat does not contain

much water, the fat mass can be calculated using total body weight and the estimate of

total body water obtained from the impedance measurements. Impedance measurements

are highly influenced by a variety of factors (e.g. temperature) that affect electrolyte

concentrations. Alternatively, dual-energy x-ray absorptiometry (DEXA) is used to

measure fat mass. DEXA passes two photons of varied energy levels through body

tissues. Body fat percentage is estimated based on the attenuation patterns of these

photons. To obtain even more accurate measures of fat mass, advanced imaging

methods, such as computed tomography (CT) and magnetic resonance imaging (MRI),

are used. Unfortunately, all of these methods require expensive equipment and expertise

that is not ordinarily available in a typical physician’s office.

Skin-fold thickness and waist circumference are much cheaper, inexpensive

measures of obesity. Skin-fold thickness is measured by pinching the skin to create a fold that can be measured with calipers and thus, estimates subcutaneous adipose tissue

mass. Alternatively, measurements of waist circumference estimate both subcutaneous

and visceral adipose tissue mass. Skin-fold thickness and waist circumference

measurements must be obtained using standardized protocols for maximal utility.

Overall, BMI is the most widely used measure of obesity because of the low cost

and ease of measurement and because height and weight measurements do not require

careful protocol standardization. The most recent data from the National Health and

Nutrition Examination Survey (NHANES), which used BMI to measure obesity,

indicated that approximately 65.7% of U.S. adults are overweight or obese and that 26

30.6% of U.S. adults are obese (HEDLEY et al. 2004). Relative to the early 1960’s when

the prevalence of obesity in U.S. adults was only 13.4%, the current prevalence

represents a considerable rise (FLEGAL et al. 2002). Likewise, childhood obesity is also

rising with 31.0% of U.S. children currently overweight or at risk for becoming

overweight (HEDLEY et al. 2004). The epidemic of obesity is not unique to the U.S.

population, and the World Health Organization has described obesity as a “global

epidemic” (WHO 1997).

3. Causes of human obesity

a. Environmental influences

Because obesity results from disturbances in energy balance, any factor that

influences energy intake or energy utilization contributes to obesity. The caloric density

of food and the quantity of food consumed are suspected to be large contributors to the

development of obesity (HEALTH 1998; HILL and PETERS 1998). In recent years, the

increased availability of food and the popularity, palatability, and cheaper cost of high-

fat, energy-dense foods (DREWNOWSKI and SPECTER 2004) are probably large

contributors to increased caloric consumption.

Although obesity derives from the Latin term “obesus,” which literally translates

to mean an individual who has become overweight through eating (American Heritage,

1969), diet and food consumption are only one cause of obesity. Decreased energy

consumption through body activity also increases risk for obesity (HEALTH 1998; HILL and PETERS 1998). Barriers to physical activity (e.g. safety) and decreased requirements

for exercise as a consequence of technologic advances, such as automobiles, computers, 27

and electronic games, contribute to decreased energy expenditure. Furthermore, non-

exercise activity thermogenesis (NEAT), which encompasses the movements required for

daily living (that are not considered exercise), such as sitting, standing, and fidgeting,

also contributes to total energy expenditure and is an intense area of obesity prevention

research (LEVINE 2004).

b. Genetic influences

Obesity is a classic example of a complex trait because both genetic and

environmental factors influence body weight. Twin studies provided some of the earliest

estimations of the heritability or proportion of the total body weight variance explained

by genetic factors. Although there are a variety of statistical methods that are used for

twin studies, the basic principle involves comparing the correlation of body weight between monozygotic (“identical”) and dizygotic (“fraternal”) twins. Because monozygotic twins are genetically identical, a trait with a strong genetic component will

have higher correlations in monozygotic vs. dizygotic twins and thus, a high heritability

estimate. Alternatively, if a trait is influenced mostly by environmental factors, no

differences between monozygotic and dizygotic twins will be observed, and the

heritability estimate will be low. One of the earliest and largest twin studies of obesity concluded that approximately 80% of variation in body weight is due to genetic factors

(STUNKARD et al. 1986). A second study controlled for environmental factors by

analyzing twins raised in the same household vs. twins raised separately and determined that the heritability of BMI is approximately 70% (STUNKARD et al. 1990). Since then,

additional studies have provided similar results (MAES et al. 1997). Consequently, body 28

weight appears to be strongly influenced by genetic factors. The genetics of obesity will

be discussed more thoroughly in a later section.

4. Pathological consequences of obesity a. Morbidity and mortality associated with obesity

The rising prevalence of obesity is a public health concern because of the high morbidity and mortality associated with elevated body weight. Approximately 15.2% of deaths in 2000 were attributable to risk factors for obesity, including poor diet and decreased activity (MOKDAD et al. 2004; MOKDAD et al. 2005). These risk factors rival

only smoking with regard to mortality risk (MOKDAD et al. 2004; MOKDAD et al. 2005).

b. Clinical conditions associated with obesity

The long list of medical conditions associated with obesity contributes to the increased morbidity and mortality associated with the condition. For instance, hypertension, diabetes, stroke, coronary artery disease, dyslipidemia, congestive heart failure, gallstones, sleep apnea, osteoarthritis, and even some types of cancer have been associated with obesity (HEALTH 1998). Furthermore, several of these conditions define

metabolic syndrome, a risk factor for cardiovascular disease. Lastly, the social stigma

associated with obesity can profoundly affect the psychological health of some obese individuals (HEALTH 1998).

29 c. Metabolic syndrome

Metabolic syndrome, also referred to as insulin resistance syndrome or syndrome

X, is probably the most well-described clinical condition associated with obesity.

Metabolic syndrome is defined by elevated body weight, dyslipidemia, hypertension, and insulin resistance (ALBERTI and ZIMMET 1998; BALKAU and CHARLES 1999; EXPERT

PANEL ON DETECTION 2001; REAVEN 1988). This constellation of clinical features is associated with an increased risk for cardiovascular disease and type II diabetes mellitus.

The various components of metabolic syndrome and their association with obesity are described below. It is important to note that even without the other features of metabolic syndrome, an elevated BMI may be an independent risk factor for cardiovascular disease

(YAN et al. 2006).

i. Insulin resistance and type II diabetes mellitus

Insulin is a hormone produced by the beta cells of the pancreas in response

to high blood glucose levels. Although the etiology is unclear, obese patients

have an increased risk for developing resistance to insulin. Relative to non-

insulin resistant patients, the tissue or cellular response to insulin is impaired in

insulin resistant individuals, so they must produce more insulin to maintain

normal blood glucose levels. Consequently, for the diagnosis of metabolic

syndrome, insulin resistance is defined by elevated plasma glucose (>6.1 mmol/L)

or elevated fasting insulin levels (ALBERTI and ZIMMET 1998; BALKAU and

CHARLES 1999; EXPERT PANEL ON DETECTION 2001). Insulin resistant patients 30

are at increased risk for the development of type II diabetes mellitus, which

occurs when insulin production declines.

ii. Dyslipidemia

Dyslipidemia refers to an unfavorable profile of plasma lipoproteins that is

associated with increased risk for heart disease. The plasma lipoproteins transport

cholesterol and triglyceride in the blood and consist of HDL, LDL, and VLDL

cholesterol and chylomicrons. The cholesterol transported by lipoproteins is derived from the diet and synthesized by the liver. Dietary cholesterol is transported from the gut to the liver by chylomicrons. Then, LDL (“bad”) cholesterol transports cholesterol from the liver to peripheral tissues. In contrast,

HDL (“good”) cholesterol transports cholesterol from peripheral tissues back to

the liver. Similarly, chylomicrons also transport dietary triglycerides to the

peripheral tissues, whereas VLDL transports triglycerides from the liver to the

periphery.

According to the clinical criteria for metabolic syndrome, dyslipidemia is

defined by reduced levels of HDL (“good”) cholesterol (<0.9-1.3 mmol/L

depending on sex and definition used) or elevated plasma triglycerides (>1.7 or

2.0 mmol/L, depending on the definition used) (ALBERTI and ZIMMET 1998;

BALKAU and CHARLES 1999; EXPERT PANEL ON DETECTION 2001).

Approximately 30-40% of obese patients have low HDL cholesterol compared to

9-16% of non-obese or non-overweight controls (BROWN et al. 2000). Obesity is 31

also associated with hypertriglyceridemia (DENKE et al. 1993; DENKE et al. 1994).

Dyslipidemia is a major risk factor for coronary heart disease.

iii. Hypertension

Hypertension or elevated blood pressure is defined by a systolic blood pressure greater than 140 mm Hg (135 mmHg in ATP III definition) or a diastolic blood pressure greater than 90 mmHg (85 mmHg in ATP III definition) (ALBERTI and ZIMMET 1998; BALKAU and CHARLES 1999; EXPERT PANEL ON DETECTION

2001). Epidemiologic studies have demonstrated that approximately 40% of obese individuals (BMI >30 kg/m2) are hypertensive compared to approximately

15% of non-obese or non-overweight controls (BROWN et al. 2000). The exact

pathophysiology explaining the relationship between elevated blood pressure and

obesity is unknown. Hypertension is a well-established risk factor for heart

disease, renal disease, and stroke.

iv. Non-alcoholic fatty liver

Although non-alcoholic fatty liver is not a component of the formal

definition of metabolic syndrome, it is strongly associated with the phenotype.

Non-alcoholic fatty liver occurs when fat accumulates in the liver (“steatosis”).

In addition to fat deposition, inflammation (“steatohepatitis”), fibrosis, and necrosis may be observed. Approximately 29% of severely obese patients have

steatosis compared to 7.1% of non-obese individuals (WANLESS and LENTZ

1990). Likewise, 18.5% of obese patients have the more severe form, 32

steatohepatitis, as compared to only 2.7% of unaffected control patients

(WANLESS and LENTZ 1990). Non-alcoholic fatty liver disease may progress to

more advanced liver diseases such as hepatic cirrhosis.

5. Physiologic control of energy balance and body weight

Despite the high prevalence of obesity and the long list of associated phenotypes,

the exact physiologic and pathologic mechanisms that underlie obesity are unknown.

Both genetic and environmental factors contribute to body weight and energy balance,

but the mechanisms through which these factors act are unclear. Evidence suggests that

genetic and environmental factors affect body weight by influencing energy balance, a

complex process that requires the integration of signals from a variety of organ systems

including the central nervous system, adipose tissue, the endocrine system, and the

digestive system. Much of our understanding of energy balance has developed from the

study of rodent models of obesity.

a. Leptin and the central control of food intake/energy balance

Leptin is a hormone produced by adipose tissue that normally circulates in the

plasma at levels that correlate with adipose tissue mass (CONSIDINE et al. 1996; ZHANG et al. 1994). Leptin influences food intake and body weight by crossing the blood brain barrier and interacting with the melanocortin pathway, a central pathway known to be highly involved in energy balance. When leptin levels in the brain rise, gene expression of pro-opiomelanocortin (POMC), a prohormone, also increases in the hypothalamus

(MIZUNO et al. 1998; SCHWARTZ et al. 1997; THORNTON et al. 1997). α-Melanocyte 33

stimulating hormone (α-MSH), a product of POMC, activates the melanocortin-3

receptor (MC3R) and the melanocortin-4 receptor (MC4R), which results in decreased

food intake and decreased adipose tissue deposition (CHEN et al. 2000; FAN et al. 1997;

HUSZAR et al. 1997; TSUJII and BRAY 1989). In addition to activating POMC neurons,

leptin inhibits the production of neuropeptide Y (NPY) (SCHWARTZ et al. 1996a;

SCHWARTZ et al. 1996b; STEPHENS et al. 1995), a stimulator of food intake (CLARK et al.

1984), reduced energy expenditure via decreased brown fat thermogenesis (BILLINGTON et al. 1991), increased white adipose deposition (BILLINGTON et al. 1991), and weight

gain (STANLEY et al. 1986). Leptin also inhibits the production of agouti-related protein

(AGRP) (MIZUNO and MOBBS 1999; WILSON et al. 1999), an MC3R and MC4R antagonist (OLLMANN et al. 1997). Consequently, when leptin levels rise, food intake decreases but energy expenditure increases. In contrast, decreases in plasma leptin levels produce the opposite results. Insulin appears to produce similar effects as leptin on food

intake (CHAVEZ et al. 1995; WOODS et al. 1979), body weight (CHAVEZ et al. 1995;

WOODS et al. 1979), and the melanocortin/NPY pathway (BENOIT et al. 2002; SCHWARTZ et al. 1991; SCHWARTZ et al. 1992). The leptin/melanocortin pathway is depicted in

Figure I-2.

b. Peripheral signals that influence food intake/energy balance

Leptin and insulin are not the only peripheral factors that influence body weight.

Various peripheral hormones derived from many organ systems play a role in energy

balance and highlight the complexity of this system. For instance, several hormones

Figure I-2. Leptin and the melanocortin pathway. Increased plasma leptin leads to decreased food intake and increased energy utilization whereas decreased plasma leptin increases food intake and decreases energy utilization (SCHWARTZ et al. 2000). Modified from Schwartz et al., 2000.

Decreased fat mass Increased fat mass

Decreased plasma leptin Increased plasma leptin To hypothalamus To hypothalamus

Increased POMC release Increased Increased neuropeptide Y AGRP release release Increased α-MSH release

Binding of α-MSH to MC3R and MC4R Increased food intake Decreased energy utilization Decreased food intake Increased parasympathetic tone Increased energy utilization Increased sympathetic tone

Legend: Activating pathway α-MSH = alpha-melanocyte stimulating hormone Inhibiting pathway MC4R = Melanocortin-4 receptor

AGRP = Agouti-related protein POMC = Pro-opiomelanocortin MC3R = Melanocortin-3 receptor

derived from the gastrointestinal system, including ghrelin (KOJIMA et al. 1999; TSCHOP et al. 2000), obestatin (ZHANG et al. 2005), peptide YY (PYY 3-36) (BATTERHAM et al.

2004; BATTERHAM et al. 2002; TSCHOP et al. 2004), oxyntomodulin (OXM) (DAKIN et al. 2001; DAKIN et al. 2004), glucagon-like peptide (GLP-1) (TANG-CHRISTENSEN et al.

1996; TURTON et al. 1996), and cholecystokinin (CCK) (DELLA-FERA and BAILE 1979;

GIBBS et al. 1973; KISSILEFF et al. 1981; MCLAUGHLIN et al. 1985; REIDELBERGER and

SOLOMON 1986; WEST et al. 1984) influence food intake and weight gain. Furthermore,

in addition to insulin, the pancreas secretes pancreatic polypeptide (PP) (ASAKAWA et al.

2003), which has been associated with food consumption and other functions. Similarly,

in addition to leptin, the adipose tissue secretes several hormones including resistin

(HOLCOMB et al. 2000; KIM et al. 2001; STEPPAN et al. 2001) and adiponectin (HU et al.

1996; MAEDA et al. 1996; NAKANO et al. 1996; SCHERER et al. 1995), and serum levels

of these hormones vary in obese vs. lean patients (ARITA et al. 1999; DEGAWA-

YAMAUCHI et al. 2003). In many cases, these peripheral factors are suspected to interact

with key metabolic pathways, such as the melanocortin pathway, in the central nervous

system to produce their effects, but several of these hormones also have peripheral

actions. Disruptions in these pathways may contribute to the development of obesity.

6. Treatment of human obesity

Historically and still today, the most common treatments for obesity have

involved decreased food consumption and increased exercise. To prevent obesity, it has

been estimated that the modification of energy balance by only 100 kilocalories daily

through decreased caloric consumption and increased energy expenditure is sufficient to 36

prevent weight gain in most individuals (HILL et al. 2003). In contrast, to treat obesity

and promote weight loss, the goal of dietary intervention is to reduce caloric intake by

500-1000 kilocalories per day (HEALTH 1998). Simultaneously, increased activity level through exercise enhances the efficacy of dietary therapy (HEALTH 1998). Behavior

therapy is also recommended for some obese patients (HEALTH 1998).

Only two weight loss drugs, orlistat and sibutramine, are currently FDA-

approved for the treatment of obesity, but both drugs have side effects that must be

considered prior to initiating treatment (DEWALD et al. 2006). These medications

should be used in addition to traditional diet and exercise therapy (DEWALD et al. 2006).

Lastly, bariatric or weight-loss surgery is recommended only in very severe obesity cases

(HEALTH 1998). A clearer understanding of the causes of obesity may lead to the

discovery of new therapies for the condition.

B. GENETICS OF OBESITY

1. Introduction to obesity genetics

In 1760, Malcolm Flemyng wrote “Not that all corpulent persons are great eaters;

or all thin persons spare feeders. We daily see instances of the contrary. Tho’ a

voracious appetite be one cause of Corpulency, it is not the only cause; . . .” (FLEMYNG

1760). Obviously, as early as the 1700’s, it was clear that environmental factors are not the sole cause of obesity. Flemyng’s comments were profound for his time, and he provided possibly the first suggestion that genetics or familial factors contribute to obesity (BRAY 1990). Despite Flemyng’s observations, it would be many years before 37

the genetics of obesity was intensively studied. Even today, the genetic factors

contributing to obesity are not clearly understood.

2. Genetic forms of obesity

Genetic studies of obesity divide the condition into two forms. First, the monogenic forms of obesity are due to single gene defects and are very rare, profound, early-onset forms of the condition that follow the laws of simple Mendelian inheritance.

Syndromic forms of obesity are a subtype of monogenic obesity, and in these cases, obesity is only one component of a large constellation of other phenotypic features. The second form of obesity is multifactorial because both genetic and environmental factors contribute to obesity susceptibility. Unlike monogenic obesity, many genetic factors contribute to the multifactorial forms, and thus, these forms are described as polygenic.

a. Monogenic obesity

The first monogenic (non-syndromic) form of obesity discovered was congenital

leptin deficiency (CLEMENT et al. 1998; STROBEL et al. 1998). Leptin deficient patients,

who have mutations in the gene encoding leptin, have severe early-onset obesity, hyperphagia, hyperinsulinemia, hypothalamic hypogonadism, T-cell abnormalities, and advanced bone age (CLEMENT et al. 1998; FAROOQI et al. 2002; STROBEL et al. 1998).

Likewise, several members of a family with similar phenotype but with elevated levels of plasma leptin were discovered to have a mutation in the leptin receptor gene (CLEMENT et al. 1998). Mutations in the genes encoding leptin or the leptin receptor have been discovered in only a very small number of severely obese patients. 38

POMC, another component of the leptin/melanocortin pathway, is also defective

in a small subset of severely obese patients. Patients with mutations in POMC have

early onset obesity, adrenal insufficiency, and red hair pigmentation (KRUDE et al. 1998;

KRUDE et al. 2003; KRUDE et al. 1999). These phenotypes are due to a deficiency of

ACTH and α-MSH, two products of the POMC protein (KRUDE et al. 1998; KRUDE et al.

2003; KRUDE et al. 1999).

Mutations in the prohormone convertase 1 gene, which encodes a protein that

cleaves prohormones and neuropeptides into their active products, have also been

discovered in a small subset of obesity patients (JACKSON et al. 2003; JACKSON et al.

1997; O'RAHILLY et al. 1995). Patients with mutations in the prohormone convertase

gene exhibit extreme early-onset obesity, abnormal glucose homeostasis,

hypocortisolemia, hypogonadotrophic hypogonadism, malabsorption in the small

intestine, and increased plasma levels of both proinsulin and POMC (JACKSON et al.

2003; JACKSON et al. 1997; O'RAHILLY et al. 1995).

Approximately 6% of a population of severe, early-onset obesity patients have mutations in the MC4R making MC4R mutations the most common form of monogenic obesity (FAROOQI et al. 2003). MC4R mutations are also associated with increased lean mass, hyperinsulinemia, hyperphagia, and increased linear growth. (FAROOQI et al. 2003;

VAISSE et al. 1998; YEO et al. 1998). Although the MC4R mutations are the first

examples of dominant obesity mutations, the phenotype appears to be modified by the

environment and other genetic factors because some unaffected heterozygous patients

have been described (FAROOQI et al. 2003). 39

An important lesson from studies of monogenic forms of obesity is that single

gene mutations account for only a small percentage of even the most severely affected

obesity patients. Although investigations of severe, early-onset obesity will identify additional monogenic forms of obesity, most obesity patients are suspected to have more complex, multifactorial forms of the condition.

b. Obesity syndromes

Obesity is also a clinical feature of several Mendelian syndromes or disorders.

The most well-known forms of syndromic obesity include Prader-Willi syndrome,

Alstrom syndrome, Albright hereditary osteodystrophy, and Bardet-Biedl syndrome

(FAROOQI and O'RAHILLY 2005). In these disorders, obesity is one component of a set of

wide-ranging phenotypes, and the genetic defects that produce the phenotypes vary from

chromosomal abnormalities that involve several genes to mutations in one or a few

closely related genes (FAROOQI and O'RAHILLY 2005).

c. Multifactorial obesity

Investigations of polygenic forms of obesity have linked obesity and obesity-

related traits to nearly every human chromosome (PERUSSE et al. 2005). Despite these

efforts, no mutations or polymorphisms have been unambiguously proven to cause

typical forms of obesity. The detection and identification of genes responsible for polygenic forms of obesity is challenging because of the large number of genes that are involved and the many environmental factors that also influence body weight. The genes 40

that contribute to polygenic obesity are usually referred to as susceptibility genes, and

many of these genes may require a particular environment to produce their effects.

d. Genetic and environmental interactions

The recent rise in the prevalence of obesity in the United States and worldwide is

obviously not explained by genetic susceptibility alone because the genetics of the

population cannot change so drastically in such a short time. Instead, the environment

has evolved over the past few decades. For example, the availability of energy dense

foods has increased while the need for physical activity has decreased. Despite exposure

to similar environments, individuals do not uniformly develop obesity. Consequently,

genetic factors that influence an individual’s physiologic response to these environmental

factors may play an important role in determining an individual’s susceptibility to obesity

and thus, may explain the “global epidemic” of obesity (WHO 1997).

The Pima Indians, a Native American population in Arizona that migrated from

Mexico, are a well-described example of the strong influence of gene-diet interactions in

the development of obesity (RAVUSSIN et al. 1994). The Pima Indians living in Arizona

are exposed to energy-dense foods and a decreased need for physical labor and have an

extremely high prevalence of obesity (RAVUSSIN et al. 1994). In comparison, the Pima

Indians still living in Mexico, who follow a traditional diet, have a much lower

prevalence of the condition (RAVUSSIN et al. 1994). Obviously, the Pima Indians have a

strong genetic susceptibility to obesity, but obesity is more likely to develop when individuals with the genetic susceptibility are exposed to an obesity-promoting

environment. 41

In another example, a twin study investigating the effects of long-term overfeeding demonstrated that some individuals are more susceptible than others to obesity when consuming excess calories (BOUCHARD et al. 1990). In this example,

exposure to the same obesity-promoting environment (through over-feeding) was obviously not sufficient for equivalent weight gain in all individuals. Instead, genetic factors contribute to the extent of weight gain during exposure to an obesity-promoting environment. Because environmental factors are so difficult to control in human studies, the influence of gene-environment interactions may complicate traditional genetic mapping studies of polygenic forms of human obesity and may partly explain why genes contributing to typical forms of obesity are so challenging to detect and identify.

In 1962, Neel proposed the “thrifty genotype” hypothesis, which explains why diabetes mellitus exists in human populations and which provides an evolutionary explanation for gene-environment interactions in obesity-related diseases (NEEL 1962).

Neel suggested that genetic factors that promote diabetes and related diseases have

historically interacted with the environment and may have actually provided a survival

advantage in the past. For example, he suggested that the ability to store excess energy in

the form of adipose tissue may have been advantageous during times of decreased food

availability, such as starvation (NEEL 1962). According to this hypothesis, over time, positive selection may have preserved genes that promote fat deposition in the human genome (NEEL 1962). Unfortunately, with increased food availability in modern times,

these same genes that may have previously provided survival advantage now confer

increased morbidity and mortality and may contribute to the current obesity epidemic

(NEEL 1962). Since Neel first described his hypothesis, the term “thrifty gene” has been 42

extrapolated to include any genetic factor that would promote survival during famine and,

therefore, includes not only genes that promote efficient fuel utilization and storage but also genes that promote hyperphagia or increased appetite, decreased physical activity, and behavioral traits that promote food hording (PRENTICE et al. 2005).

C. METHODS FOR INVESTIGATING THE GENETICS OF OBESITY

1. Genetic studies of obesity in humans

Human genetic studies of typical, polygenic forms of obesity generally utilize one

of two approaches: candidate gene association or whole genome linkage studies. With

candidate gene approaches, genes suspected to play a role in obesity susceptibility or

genes in a region of the genome associated with obesity (see below) are selected. Then,

polymorphisms within the gene are identified and analyzed in a population of obese and

control patients to test whether the polymorphism or group of polymorphisms (haplotype) are significantly associated with the obesity phenotype.

Unlike candidate gene association studies, a list of potential obesity-related

genes is not required to pursue whole genome linkage. For whole genome linkage, large

populations of families or pairs of siblings are identified and genetic markers spanning

the genome are genotyped in study participants. Although the details of the various

statistical methods for mapping quantitative trait loci (QTLs), or genes influencing

complex traits using the linkage approach will not be discussed in detail, the overall

objective is to determine if there is increased allele sharing (“linkage”) at a particular

marker among affected vs. unaffected family members. 43

Together, candidate gene association and linkage analyses have identified many

QTLs (PERUSSE et al. 2005). Although a few of these regions have been replicated in one or more populations, in most cases, the associations and linkages have not been replicated. The difficulty of replication may be due to genetic heterogeneity, phenotypic measurement variation, and environmental variation.

The study of traits such as obesity is complex because of the many environmental factors that may obscure the phenotype. For example, a high calorie diet may promote obesity regardless of the genetic constitution. Likewise, a low calorie diet and exercise may promote leanness despite genetic susceptibility to obesity. Consequently, many

investigators focus on early-onset or morbid obesity because these forms of obesity may be less strongly influenced by diet and activity (BELL et al. 2005). Because of the

complexity of human genetic studies of typical forms of obesity, many investigators have

used mouse models for the study of the genetics of obesity.

2. Genetic studies of obesity in mouse

The simple, fixed genetics of inbred mouse strains and the ability to control environmental factors such as diet make mouse studies of obesity less complicated than human studies. As in humans, two forms of genetic obesity exist in mouse. The simplest mouse models of obesity are the monogenic mouse mutants, which include the naturally occurring obesity mutants and genetically engineered knock-out or transgenic models. In contrast, inbred strains provide useful models for the more common, polygenic forms of obesity because of the naturally occurring variation in body weight among them.

44 a. Spontaneous mouse mutants led to the discovery of the first obesity genes

The ob, db, Ay, tub, and fat mouse models, which develop extreme diet- independent obesity, are the well-characterized spontaneous, monogenic forms of mouse obesity. These mice developed spontaneous mutations that produced early-onset, severe obesity relative to unaffected littermates even when fed a standard mouse diet. The discovery and analysis of these mutants have led to the identification of the first obesity genes, provided the earliest mouse models for obesity, and revolutionized the obesity field.

i. ob: Studies of the ob (obese) mouse led to the discovery of the hormone leptin

(ZHANG et al. 1994). ob mice have a nonsense mutation in Lep (ZHANG et al.

1994), the gene encoding leptin, that produces severe early onset obesity,

hyperglycemia, hyperinsulinemia, hyperphagia, reduced body temperature, and

infertility due to leptin deficiency (COLEMAN 1982; INGALLS et al. 1950). The

phenotype of the ob/ob mouse is similar to the phenotype of the leptin deficient

patients.

ii. db: The db/db (diabetes) mice display a similar phenotype to the ob/ob mice

(COLEMAN 1982; HUMMEL et al. 1966). After the leptin gene was cloned, studies

revealed that the gene responsible for the db phenotype encodes the leptin

receptor (CHEN et al. 1996; TARTAGLIA et al. 1995). db/db mice have a 106

nucleotide insertion that disrupts normal splicing and produces a premature stop

codon in one form of the leptin receptor (Ob-Rb) (CHEN et al. 1996). The

resulting protein is a hypothalamic leptin receptor without a large portion of the 45

cytoplasmic domain that does not transmit signals from the leptin protein (CHEN et al. 1996; LEE et al. 1996). As with the ob/ob mouse, the phenotype of db/db

mice is similar to the phenotype of patients with leptin receptor deficiency.

iii. Obese yellow and viable yellow mice (Ay, Avy, etc.) The obese yellow and

viable yellow mice harbor mutations that result in overexpression of the agouti

gene. Ay and Avy mice have an autosomal dominant phenotype that includes

yellow coat color (BULTMAN et al. 1992; CUENOT 1905), late onset obesity (8-17

months of age) (CASTLE 1941; DANFORTH 1927; DICKIE and WOOLLEY 1946;

HESTON and VLAHAKIS 1961), hyperphagia (DICKERSON and GOWEN 1947;

FRIGERI et al. 1988; YEN et al. 1984), hyperinsulinemia (FRIGERI et al. 1983;

GILL and YEN 1991), and an increased susceptibility to certain tumors (HESTON and DERINGER 1947; HESTON and VLAHAKIS 1961; HESTON and VLAHAKIS 1968;

y VLAHAKIS and HESTON 1963). The phenotype associated with A is due to a large deletion that places the agouti protein, a melanocortin-1 receptor (MC1-R) and

MC4-R antagonist, under the control of a ubiquitously expressed promoter

(BULTMAN et al. 1992; LU et al. 1994; MICHAUD et al. 1993; MILLER et al. 1993).

Consequently, the MC4-R is blocked and the yellow obese phenotype results.

Likewise, in Avy mice, an intracisternal A particle is inserted into the agouti gene

and leads to ubiquitous overexpression of the gene (DUHL et al. 1994).

iv. fat: fat/fat mice display obesity, hyperglycemia, early-onset

hyperproinsulinemia, and infertility (COLEMAN and EICHER 1990; NAGGERT et 46

al. 1995). fat/fat mice have a missense mutation in the gene encoding

carboxypeptidase E (CPE) leading to decreased activity of CPE, an enzyme

involved in the sorting and processing of prohormones (BERMAN et al. 2001;

COOL et al. 1997; NAGGERT et al. 1995). Hormones known to be processed by

CPE include proinsulin, POMC, prodynorphin (the precursor of several neuroendocrine hormones involved in feeding regulation), pro-melanin- concentrating hormone, pro-neurotensin, pro-gastrin, pro-glucagon, and other neuronal and endocrine pro-hormones (BERMAN et al. 2001; FRIIS-HANSEN et al.

2001; NAGGERT et al. 1995; ROVERE et al. 1996; UDUPI et al. 1997).

Consequently, complex alterations in feeding behavior, energy balance, and glucose homeostasis are believed to influence weight gain in these mice because many of these neuroendocrine hormones play important roles in these processes.

v. tub: The tub mutation is associated with autosomal recessive late-onset obesity, retinal and cochlear degeneration, insulin resistance, and infertility

(COLEMAN and EICHER 1990; OHLEMILLER et al. 1995). The tub gene, which is

expressed in various regions of the nervous system including the hypothalamus, encodes a transcription regulator with a highly conserved C-terminus domain that

is disrupted in the mutant protein by a splice donor site mutation (CHUNG et al.

1996). Studies suggest that the disruption of the C-terminal domain leads to an unstable protein product (STUBDAL et al. 2000). Although the exact role of TUB in obesity development has not been established, TUB appears to be a transcriptional regulator that is activated by G-coupled protein receptor signaling 47

(SANTAGATA et al. 2001). Although the target genes of TUB have not been

identified, alterations in the expression of NPY and POMC, two components of

the melanocortin pathway, have been detected in tubby mice suggesting that the

melanocortin pathway may be altered in these mice (GUAN et al. 1998).

vi. Fat aussie: The fat aussie mouse, the most recently discovered spontaneous

obesity mouse mutant, develops late-onset obesity (by 120 days of age),

hyperphagia, type 2 diabetes mellitus, hypercholesterolemia, infertility, and

hearing loss (ARSOV et al. 2006). Genetic studies of the fat aussie mouse

identified an 11bp deletion in the Alms1 (Alstrom syndrome 1) gene that produces

a premature stop codon (ARSOV et al. 2006). The human ortholog of Alms1 had

previously been associated with a similar phenotype in humans (Alström

syndrome) (COLLIN et al. 2002; HEARN et al. 2002), and the encoded protein is a

component of the basal bodies in cilia and centrosomes (HEARN et al. 2005). The

exact mechanism of obesity in the fat aussie mouse or in Alström syndrome

patients is unclear.

b. Single gene knock-out and transgenic models of obesity

Genetically engineered knock-out and transgenic mice provide additional models

for monogenic forms of human obesity. Many knock-out and transgenic models for both

diet-independent and -dependent obesity have been generated (BROCKMANN and BEVOVA

2002; PERUSSE et al. 2005). Knock-out and transgenic models have extreme

perturbations in single genes and are useful for studying single genes and pathways 48

involved in obesity. Investigations of these mouse models, which are too numerous to

list and describe here, have linked genes involved in a variety of pathways, ranging from

the melanocortin pathway to hormone biology and glucose metabolism, with obesity

(BUTLER and CONE 2001; TSCHOP and HEIMAN 2001).

c. Mouse models of polygenic obesity

Although single gene models provide insight into rare monogenic forms of

obesity and into pathways involved in energy balance, polygenic mouse models of obesity provide better resources for the genetic study of typical, multifactorial forms of

human obesity. The obesity in polygenic mouse models is due to large numbers of genes and in some cases, composition of the diet. These models can be used for the detection and identification of obesity QTLs. QTLs are regions of the genome that have

been statistically associated with a complex trait, such as obesity, and which most likely

contain one or more genes that influence the trait. The identification of these QTLs in mice may provide important insight into human obesity QTLs, obesity pathways, and possibly new therapeutic targets for obesity. Furthermore, studies of multifactorial mouse models of obesity have revealed the importance of gene interactions in the

development of obesity.

i. Mouse obesity QTLs:

Mapping crosses with a wide variety of inbred mouse strains have been

used for obesity QTL discovery (BROCKMANN and BEVOVA 2002; PERUSSE et al.

2005). Inbred strains are useful because they vary considerably in body weight 49

and because of the complex nature of obesity in these strains. Some inbred

strains, like LG/J and SM/J, have been selected for large and small body weight respectively during inbred strain development (GOODALE 1938; MACARTHUR

1944) and serve as models for diet-independent obesity or large size. In contrast, several inbred strains, which were not selected for a particular body size during inbred strain development, differ in body weight when fed a high-fat diet.

Examples of strain pairs used for genetic studies of diet-induced obesity include

C57BL/6J and A/J (SURWIT et al. 1988), SWR/J and ARK/J (WEST et al. 1994a;

WEST et al. 1994b; YORK et al. 1997), and CAST/Ei and C57BL/6J (YORK et al.

1996). Strains with differential susceptibility to high-fat, diet-induced obesity are excellent models for the typical forms of human obesity because both dietary and genetic factors influence body weight in these strains.

To dissect the genetics of obesity in inbred strains with varying body weight, crosses between the strains are made to generate F1 mice. The F1 mice are then intercrossed to generate F2 progeny or backcrossed to one of the parental inbred strains to generate backcross progeny. Whole genome scan mapping studies usually require the collection of hundreds to thousands of F2 or backcross progeny. Once the mice are collected, molecular markers are genotyped in each individual (or individuals at the extreme ends of the phenotypic distribution), body weight or obesity-related phenotypes are measured and then, statistical methods are used to test for linkage between the phenotype and genotypes. The results of mapping studies are presented as LOD scores, which measure the likelihood that a gene affecting the trait resides in a particular region of the 50

genome. Regions with significant LOD scores are selected for fine-mapping studies to localize the QTL with the goal of eventually identifying the underlying gene.

Conventional QTL studies in mice have linked obesity or obesity-related phenotypes to every chromosome except chromosome Y (BROCKMANN and

BEVOVA 2002; PERUSSE et al. 2005). As of 2002, a total of 160 QTLs had been

identified with the highest numbers on chromosomes 1, 11, and 7, and many chromosomes appear to have multiple obesity or obesity-related QTLs

(Brockmann and Bevova 2002).

Other approaches for mapping QTLs have been pursued in the mouse. For

instance, recombinant inbred strains, which are generated by crossing two inbred

strains followed by twenty generations of inbreeding (SILVER 1995), have been

used for mapping obesity QTLs. In particular, recombinant inbred lines derived

from LG/J and SM/J were used to map QTLs associated with obesity (CHEVERUD et al. 2004). Likewise, chromosome substitution strains (CSSs), which are discussed in a later section, have also provided important models for the dissection of complex traits such as obesity (SINGER et al. 2004).

Despite the large number of obesity QTLs detected in the mouse, very few

obesity QTLs have been associated definitively with the causative gene. Using a

combination of mapping crosses, gene expression profiles, and a knock-out mouse

model, 5-lipoxygenase, an enzyme involved in leukotriene biosynthesis, was

recently discovered to be underlying a QTL increasing fat pad mass in an F2 cross

between C57BL/6J and DBA/2J inbred mice (MEHRABIAN et al. 2005). This 51

study illustrates the need for novel, multi-dimensional approaches for QTL

detection and successful gene identification.

ii. Gene interactions and obesity

In addition to the existence of many genes that influence body weight, increasing evidence indicates that interactions between genes also influence obesity in the mouse (WARDEN et al. 2004). Some of the earliest evidence for genetic interactions arose from the simplest mouse models of obesity because the phenotype of single gene obesity mutations varies depending on the genetic

background on which the mutation is placed (WARDEN et al. 2004). For example,

when the db or ob mutation is placed on the C57BL/6J background, the mice are

obese and hyperinsulinemic with beta cell hyperplasia (COLEMAN and HUMMEL

1973; HUMMEL et al. 1972). In contrast, when the same mutations are placed on

the C57BL/Ks genetic background, the mice are obese with severely decreased

insulin production and early death (COLEMAN and HUMMEL 1973; HUMMEL et al.

1972). Obviously, variation in other genes in the C57BL/6J vs. C57BL/Ks

genetic backgrounds must modulate the phenotype.

Genetic mapping studies using more complex, polygenic models of

obesity have also provided evidence for gene interactions (BROCKMANN et al.

2000; CHEVERUD et al. 2001; WARDEN et al. 2004; YI et al. 2004; YI et al. 2006).

In these examples, two QTLs produce a greater effect than would be predicted if the QTLs were acting independently and additively. Methods for discovering and analyzing gene interactions in mapping data are still emerging and will most 52

likely be utilized more extensively in the future (WARDEN et al. 2004). The

existence of gene interactions complicates traditional mapping analysis and may

partially explain why so few obesity QTLs are detected in human and mouse

mapping studies.

D. C57BL/6J AND A/J INBRED STRAINS: MOUSE MODELS FOR OBESITY

As stated previously, the C57BL/6J and A/J inbred mouse strains are ideal mouse models for obesity studies because the strains differ in their susceptibility to high-fat, diet-induced obesity (BLACK et al. 1998; COLLINS et al. 2004; REBUFFE-SCRIVE et al.

1993; SURWIT et al. 1988). When fed a high-fat, simple carbohydrate (HFSC) diet

(containing sucrose), C57BL/6J, but not A/J, male mice develop obesity (SURWIT et al.

1988), but on a low-fat diet, the body weights are more similar (BLACK et al. 1998).

Consequently, a combination of genetic and dietary factors explains the obesity in

C57BL/6J, and this strain provides an important model for typical, polygenic forms of diet-induced human obesity.

1. Differential susceptibility to diet-induced obesity in C57BL/6J and A/J male mice

The body weight variation in C57BL/6J and A/J male mice fed the high-fat diet has been demonstrated to be due to differences in fat deposition. For example, C57BL/6J males fed the high-fat diet display an increased percentage of body fat and larger fat pads relative to the A/J males (BLACK et al. 1998; REBUFFE-SCRIVE et al. 1993). Although both strains increase adipocyte cell size when fed the high-fat diet, only C57BL/6J males 53

also increase adipocyte cell number indicating that adipocyte hyperplasia is associated

with weight gain in this strain (BLACK et al. 1998; REBUFFE-SCRIVE et al. 1993).

The cause of obesity in C57BL/6J male mice is not understood. C57BL/6J males do not have decreased motor activity relative to A/J males when fed the high-fat diet

(BROWNLOW et al. 1996). The role of hyperphagia is unclear with several studies

detecting no differences in food intake in C57BL/6J vs. A/J males (BROWNLOW et al.

1996; SURWIT et al. 1995; WATSON et al. 2000), and only one study detecting slightly

increased food intake in C57BL/6J vs. A/J males (PAREKH et al. 1998). Because motor activity and hyperphagia do not seem to explain the obesity, fundamental differences in energy metabolism may explain the differential susceptibility to obesity.

In addition to obesity, C57BL/6J males also develop hyperglycemia (REBUFFE-

SCRIVE et al. 1993; SURWIT et al. 1995; SURWIT et al. 1988), hyperinsulinemia

(REBUFFE-SCRIVE et al. 1993; SURWIT et al. 1995; SURWIT et al. 1988),

hypercholesterolemia (REBUFFE-SCRIVE et al. 1993), and hypertension (MILLS et al.

1993) when consuming the high-fat diet This constellation of phenotypes is remarkably similar to metabolic syndrome, which is associated with obesity in human populations

(ALBERTI and ZIMMET 1998; BALKAU and CHARLES 1999; EXPERT PANEL ON DETECTION

2001; REAVEN 1988). Consequently, C57BL/6J males are models for both diet-induced

obesity and metabolic syndrome (COLLINS et al. 2004). The obesity, hyperglycemia, and

hyperinsulinemia are completely reversible when the high-fat diet is replaced with a low- fat diet (PAREKH et al. 1998).

2. Energy metabolism in C57BL/6J and A/J male mice

Differences in plasma leptin (SURWIT et al. 1997; WATSON et al. 2000), body

temperature (SURWIT et al. 1997), and uncoupling protein gene expression (SURWIT et al.

1997; WATSON et al. 2000) have also been associated with the differential susceptibility

of C57BL/6J and A/J males to diet-induced obesity. For example, when fed the high-fat

diet for two to four weeks, A/J males have significantly higher plasma leptin

concentrations relative to C57BL/6J males (SURWIT et al. 1997; WATSON et al. 2000).

In contrast, the plasma leptin concentrations of C57BL/6J males are the same or higher

than that of A/J males after 4-8 months of the diet when obesity is present (SURWIT et al.

1997; WATSON et al. 2000). When adipose mass is taken into account, A/J males appear

to produce more leptin per unit adipose tissue than C57BL/6J mice (WATSON et al.

2000). Intraperitoneal leptin treatment does not rescue the obesity in C57BL/6J mice

suggesting that the obesity phenotype is not due to differences in leptin concentration or

leptin production (SURWIT et al. 2000). Furthermore, resistance to intraperitoneal leptin

administration has been demonstrated in C57BL/6J but not A/J mice after 4-8 weeks on a

high-fat diet even though intracerebroventricular leptin produced a response in C57BL/6J

mice at the same time points (PRPIC et al. 2003). These results suggest that the HFSC

diet may decrease the ability of C57BL/6J mice to transport leptin across the blood-brain

barrier (PRPIC et al. 2003).

Studies of body temperature also provide evidence suggesting that differences in

energy metabolism explain the obesity in C57BL/6J, but not A/J males. Despite similar

body temperature when fed a low-fat diet, the A/J males have an increased body temperature relative to C57BL/6J males when fed a high-fat diet (SURWIT et al. 1997). 55

These differences in body temperature suggest that A/J males may metabolize fuel less

efficiently than C57BL/6J. One potential explanation for the elevated body temperature in A/J males is increased activity of uncoupling proteins. Uncoupling proteins (UCPs) are mitochondrial membrane proteins that uncouple electron transport from oxidative phosphorylation during the process of mitochondrial respiration. UCPs decrease the efficiency of energy production because they promote the production of heat rather than

ATP, the normal product of respiration. Increased expression of both UCP-1 and UCP-2 has been detected in both white and brown adipose tissue of the A/J males fed the high- fat diet providing further support for the hypothesis that inefficient energy utilization in

A/J males may promote resistance to obesity (SURWIT et al. 1997; WATSON et al. 2000).

Overall, despite these clues, the mechanism of obesity development in C57BL/6J and

obesity resistance in A/J males remains unclear.

3. Genetic studies of obesity in C57BL/6J and A/J inbred strains

Although the differential susceptibility of C57BL/6J and A/J male mice to diet-

induced obesity and metabolic syndrome is obviously due to a combination of genetic

and dietary factors, few genetic studies of obesity in these strains have been performed.

Only one genetic mapping study of body weight in segregating populations derived from

C57BL/6J and A/J males is published in the literature (Zhang and Gershenfeld 2003).

This study, which utilizes both backcross and F2 progeny, was performed using a

standard diet (17 kcal% fat), and only two significant QTLs were detected (Zhang and

Gershenfeld 2003). The small number of QTLs detected is probably due to minimal trait

variation associated with the low-fat diet and the limited power of F2 and backcross 56

approaches for detecting QTLs associated with a trait as complex as body weight. Neither

F2 nor backcross studies of body weight in these strains on the high-fat diet have been

reported in the literature.

4. C57BL/6J and A/J: Advantages for complex trait studies

The C57BL/6J and A/J inbred strains are useful models for genetic studies of

complex traits, such as diet-induced obesity, for many reasons. First, these two inbred

strains are inexpensive and easily available from Jackson Laboratory

(http://www.jax.org/). In addition, several important resources are available for genetic

investigations of these strains including chromosome substitution strains (CSSs)

(discussed below) (NADEAU et al. 2000; SINGER et al. 2004) and recombinant inbred

strains (MARSHALL et al. 1992; NESBITT and SKAMENE 1984), which can be used for

QTL mapping and studies of phenotype correlations, respectively (HOIT et al. 2002). In

addition, a panel of recombinant congenic strains derived from A/J and C57BL/6J has

been generated and may be useful for studies of complex traits (FORTIN et al. 2001).

Furthermore, many microsatellite markers and single nucleotide polymorphisms (SNPs) have been identified in this strain pair and are useful for mapping studies

(http://www.ncbi.nlm.nih.gov/projects/SNP/;http://www.informatics.jax.org/; http://www.broad.mit.edu/personal/claire/MouseHapMap/Inbred.htm). Lastly, once

QTLs are discovered in these strains, the existence of bacterial artificial chromosome

(BAC) libraries (http://bacpac.chori.org/mouse27.htm), and the existence of many knock- out and transgenic models on the C57BL/6J background may assist in discovering and proving the identity of the causative gene. 57

5. B6-ChrA CSSs accelerate QTL mapping studies in C57BL/6J and A/J strains

a. Definition of CSSs

The B6-ChrA CSSs, the first complete panel of mammalian CSSs, consists of 22

strains in which individual A/J chromosomes (19 autosomes, 2 sex chromosomes, and

mitochondrial genome) replace the equivalent chromosome on the C57BL/6J background

A (NADEAU et al. 2000; SINGER et al. 2004). Briefly, the B6-Chr CSS panel was

constructed by crossing C57BL/6J (recipient genome) and A/J (donor genome) mice to

generate F1 progeny. The F1 progeny were backcrossed to C57BL/6J. After each

backcross generation, only mice with a non-recombinant A/J chromosome were selected

(by genotyping) for further backcrosses. After at least 10 backcross generations, mice

which inherited the same non-recombinant A/J chromosome were selected (by

genotyping) and intercrossed to homozygose the A/J chromosome (NADEAU et al. 2000;

SINGER et al. 2004). Modifications of these procedures were used to construct the mitochondrial and sex chromosomal substitution strains. CSS panel construction is illustrated in Figure I-3. Chromosome substitution strains have subsequently been constructed for the study of complex traits in rats (COWLEY et al. 2004a; COWLEY et al.

2004b; MALEK et al. 2006).

Figure I-3. B6-ChrA CSS panel construction. (NADEAU et al. 2000; SINGER et al. 2004)

X C57BL/6J (B6) A/J … … … … F1 Backcross for 10 generations … …

Intercross N10+ to homozygose N10 the A/J chromosome … …

B6-Chr 1A. . .19A, XA, YA, mitoA … …

b. Advantages of CSSs

CSSs are useful for complex trait studies because they simplify QTL detection

(Figure I-4) (NADEAU et al. 2000; SINGER et al. 2004). For instance, to detect

chromosomes with QTLs affecting a trait of interest, such as obesity, the phenotypes of

the individual CSSs and the host strain (in this case, C57BL/6J) are analyzed (NADEAU et al. 2000; SINGER et al. 2004). If a CSS has a phenotypic value that is significantly

different from C57BL/6J then a QTL probably exists on the substituted A/J chromosome

(NADEAU et al. 2000; SINGER et al. 2004). Therefore, no genotyping or crosses are

required to detect QTLs using CSSs (NADEAU et al. 2000; SINGER et al. 2004).

CSSs also have many statistical advantages. Because the whole genome is not

segregating in CSS populations, the background noise from other QTLs is reduced, and

the proportion of the total phenotypic variance due to any single QTL is larger relative to

a traditional F2 cross (BELKNAP 2003; SINGER et al. 2004). For instance, for a trait with a

heritability of 0.40, Belknap demonstrated that a QTL which accounts for 6% of the trait

variance in an F2 cross will account for 12% (complete dominance) to 18% (no

dominance assumed) of the variance in a CSS population (Belknap 2003). A second

advantage of CSSs relative to traditional crosses is that the LOD thresholds for

significant and suggestive QTLs are several-fold lower using CSSs (BELKNAP 2003;

SINGER et al. 2004). Finally, at least 37% fewer mice are needed to detect a QTL using

the CSS vs. F2 cross (SINGER et al. 2004).

Figure 1-4. QTL mapping with CSSs. Comparisons of the phenotypes of CSSs and C57BL/6J (B6) reveal which chromosomes have genes affecting the trait of interest.

A A B6 B6-Chr 1 B6-Chr 19 A/J … … … … … … , … … , …

Detect genes by comparing B6 and CSS phenotypes: = = No Gene Detected Gene Trait Value Trait

B6 B6-Chr 1A B6-Chr 19A A/J

Analyses of 150 traits ranging from anxiety to metabolism in the B6-ChrA CSSs detected 53 QTLs and highlighted the advantages of the CSS approach for QTL detection

A (SINGER et al. 2004; SINGER et al. 2005). In separate analyses, the B6-Chr CSSs have

been used to map QTLs associated with pubertal timing (KREWSON et al. 2004), prepulse

inhibition (PETRYSHEN et al. 2005), and airway hyperresponsiveness (ACKERMAN et al.

2005). Even before the B6-Chr 6A CSSs were constructed, a CSS derived from MOLF/Ei and 129/Sv was used to map a QTL associated with testicular tumorigenesis (MATIN et al. 1999). Although no significant or even suggestive QTLs were detected for testicular tumorigenesis on chromosome 19 in a cross between the 129/Sv-Ter/+ and MOLF/Ei strains, a significant QTL was detected using a CSS with MOLF/Ei chromosome 19 on the 129/Sv background (MATIN et al. 1999). Consequently, CSSs are much more

powerful for detecting QTLs in the mouse as compared to traditional approaches using

F2 and backcross populations.

In addition to the statistical advantages, follow-up studies to fine map QTLs can

be pursued using F2 or backcrosses derived from individual CSSs. F2 progeny derived

from each CSS, in which only one chromosome segregates, can be used to localize QTLs.

Likewise, congenic strains spanning the entire chromosome or even a small region of the

chromosome can be constructed in three to four generations which is much less time-

consuming than the 10+ generations required to construct congenic lines from the

original C57BL/6J and A/J parental strains (YOUNGREN et al. 2003). Lastly, unlike F2 or

backcross progeny which are genetically unique, each individual CSS is a fixed inbred

strain and a renewable resource. Consequently, as genetic studies are performed to

narrow individual QTLs, additional phenotypic studies can be simultaneously performed 62

in the original CSS (or congenic strains generated using CSSs) to further refine the

phenotype of interest and provide clues to the candidate gene.

Once QTLs are fine-mapped, the availability of polymorphism data and BAC

transgenic libraries derived from C57BL/6J and A/J (as discussed previously) will be

useful for gene discovery. Furthermore, because many knock-out mouse models exist on

the C57BL/6J background, combination knock-out and transgenic approaches have the

potential for providing proof for QTL localization and for functional studies of the

candidate gene or genes.

c. CSS obesity studies

To date, the only QTLs associated with high-fat, diet-induced obesity in

A C57BL/6J and A/J were detected in studies using B6-Chr CSSs (SINGER et al. 2004).

When the B6-ChrA CSSs were fed the HFSC diet, 17 CSSs were resistant to the diet- induced obesity (SINGER et al. 2004). Therefore, 17 obesity resistance QTLs were

detected. Furthermore, the cumulative effect of the 17 resistance QTLs far exceeds the

weight difference between C57BL/6J and AJ indicating that gene interactions must

contribute to obesity resistance in the A/J males (SINGER et al. 2004). Consequently, the

B6-ChrA CSSs provide a unique model for dissecting the genetics of diet-induced obesity

resistance and will serve as the model used in the research described in this thesis.

E. SUMMARY AND RESEARCH AIMS

Obesity is a major public health problem in the United States and worldwide.

Despite the high prevalence of obesity, the exact pathophysiology of the condition is not

well-understood. The increased availability of food and decreased activity levels in

modern society are environmental factors that presumably contribute to the rising

prevalence of obesity, but genetic factors also play a significant role. Unfortunately,

despite many genetics studies of obesity, no genes have been clearly associated with

complex, typical forms of the condition. The discovery of obesity genes is important

because it may lead to a greater understanding of the physiology of obesity and to the

discovery of novel therapeutic strategies for obesity. To achieve these goals, we must

first develop and improve methods for obesity gene discovery.

Because of the complexity of human obesity studies, we used inbred mouse

strains as models for studying human obesity. In particular, we are interested in the

genetics of the differential susceptibility of C57BL/6J and A/J inbred male mice to high-

fat, diet-induced obesity. The central goal of this work is to utilize the B6-ChrA CSSs to investigate the genetics of obesity resistance at a whole genome and single chromosome level in these two inbred strains.

To dissect the genetics of obesity in the CSSs, two independent, replicate CSS

HFSC diet surveys were performed. These surveys revealed that 13 A/J-derived chromosomes reproducibly conferred obesity resistance. Although these surveys indicate which chromosomes have obesity resistance QTLs, they do not provide clues to the number of QTLs present or to QTL localization on individual chromosomes. In Chapter

2, the number and location of QTLs on individual A/J chromosomes was estimated using 64

a CSS whole genome scan. To this end, F2 progeny were derived from each CSS, and

mapping studies were performed. Interestingly, obesity resistance QTLs were discovered

and localized on three A/J chromosomes with evidence for at least two QTLs on

chromosome 6. Moreover, A/J-derived obesity-promoting QTLs were detected on two

A/J chromosomes. Therefore, single, strong QTLs were not detected on most

chromosomes that reproducibly conferred resistance in the HFSC diet CSS surveys.

Consequently, obesity resistance is characterized by complexity even on single

chromosomes.

To further investigate the genetics of obesity resistance on a single chromosome,

we generated and screened a panel of congenic strains derived from B6-Chr 6A (Chapter

3.) The congenic panel survey revealed two obesity resistance QTLs (Obrq1 and Obrq2).

In addition, the congenic panel revealed a third QTL that acts as a suppressor of one of the obesity resistance QTLs (Obrq3). Consequently, the congenic strains provided further evidence indicating that the genetics of obesity resistance is complex on both a whole genome and chromosome level.

In addition to providing tools for studying the genetics of obesity resistance, the

CSSs and congenic strains derived from them also provide resources for investigating the physiology and phenotypes related to obesity. Chapter 4 discusses physiologic and metabolic studies used to refine the phenotype of B6-Chr 6A and a congenic strain

containing Obrq2. Our investigations revealed that B6-Chr 6A is resistant to several

aspects of metabolic syndrome including hypercholesterolemia, hyperglycemia, and

elevated liver triglycerides in addition to obesity. Likewise, the obesity resistant 65 congenic strain had a similar phenotype. Consequently, our models of obesity resistance are also models of resistance to the metabolic syndrome.

Overall, these studies demonstrate that the genetics of obesity is highly complex in our model with multiple genes and possibly gene interactions influencing the trait.

Moreover, this work suggests that the B6-ChrA CSSs might be important tools for the discovery of genes, pathways and mechanisms involved in obesity resistance and possibly for the discovery of novel, therapeutic strategies for the condition.

CHAPTER II: GENETIC DISSECTION OF OBESITY IN CHROMOSOME SUBSTITUTION STRAINS

The work described in this chapter is the result of a collaborative effort among many investigators. The C57BL/6J and A/J parental strain analyses were performed by the candidate in collaboration with Annie Hill. The CSS surveys were performed by Annie Hill and Dr. David Sinasac. The mouse husbandry and body weight measurements for the CSS intercrosses were performed by the candidate in collaboration with Annie Hill. Nicole Nadeau and Christine Jimenez, two undergraduate students, also assisted with body weight measurements. The mitochondrial reciprocal cross study was designed by the candidate, but the data were gathered by Annie Hill. Dr. Mark Daly and Andrew Kirby performed SNP genotyping for the intercrosses, and Dr. Karl Broman assisted with the statistical analyses for the parental strains, CSS surveys, and intercross studies. Dr. Karl Broman wrote the section describing the statistical threshold determination. All other work was performed by the candidate. 67

A. INTRODUCTION

Over the past two decades, more than 1200 genes associated with monogenic

disease have been cloned (BOTSTEIN and RISCH 2003). In contrast, very few genes

associated with polygenic diseases, such as obesity, have been discovered. Unlike

Mendelian disorders which are caused by relatively rare, single gene mutations with large

phenotypic effects, complex disease phenotypes are believed to be influenced by

mutations or SNPs in many genes and by many environmental factors. Because of the

phenotypic complexity, methods previously used for mapping genes associated with

single gene disorders have not been as successful for mapping complex disease genes.

For example, genomewide linkage scans for obesity and obesity-related traits in many

human populations have detected only a single significant QTL (COMUZZIE et al. 1997;

DENG et al. 2002; HAGER et al. 1998) even though obesity is obviously not a monogenic

disorder. Furthermore, obesity-related QTLs are only rarely confirmed in a second

human population (BELL et al. 2005), and few, if any, genes have been unambiguously

associated with complex, polygenic forms of human obesity.

Although variation in diet and exercise levels could explain the challenges of

detecting obesity QTLs in humans, similar problems hinder QTL detection in animal models where environmental factors are more easily controlled. For example, only two

significant QTLs were discovered in a genetic analysis of body weight (regular diet) in

514 intercross and 223 backcross progeny derived from C57BL/6J and A/J (ZHANG and

GERSHENFELD 2003). Similar investigations of diet-induced obesity have produced

comparable results (ISHIMORI et al. 2004; REED et al. 2003; YORK et al. 1996).

Consequently, novel approaches that increase the power for QTL detection must be 68

pursued in both humans and animal models because detecting QTLs is the first and

perhaps most important, rate-limiting step in studies of multigenic traits.

The B6-ChrA mouse chromosome substitution strains (CSSs) are a novel model

for investigating the genetics of the differential susceptibility to diet-induced obesity in

the C57BL/6J and A/J inbred strains (NADEAU et al. 2000; SINGER et al. 2004). A CSS

survey using the HFSC diet demonstrated that 17 CSSs had lower body weight relative to

C57BL/6J (SINGER et al. 2004). Therefore, the CSS high-fat diet survey detected at least

17 genes that influence body weight in the A/J strain (SINGER et al. 2004). Furthermore, the CSS model detected more QTLs than traditional approaches (SINGER et al. 2004).

Although many body weight-related QTLs were detected in the CSS survey, it is

unclear whether these QTLs affect growth or body size regardless of diet consumed or

whether they specifically confer resistance to high-fat, diet-induced obesity.

Distinguishing between these two possibilities is important because the identification of

QTLs specific for high-fat, diet-induced obesity may lead to the identification of novel obesity genes and therapeutic strategies for diet-induced obesity. Further studies of the

CSSs using a low-fat diet should begin to address these questions.

Another question posed by the CSS surveys is whether single QTLs or the cumulative action of several QTLs produces the decreased body weight conferred by individual A/J chromosomes. To determine the number and location of QTLs conferred

by each substituted chromosome, an intercross (F2) strategy can be used. In F2 crosses

with CSSs, only one chromosome segregates because the other chromosomes are fixed.

Consequently, compared to traditional F2 crosses in which “noise” from other

segregating QTLs may obscure QTL detection, the F2 crosses derived from CSSs provide 69 a simpler, more powerful model for QTL detection. For example, in studies of body weight, total plasma cholesterol, and anxiety, fewer than 100 F2 or backcross progeny derived from individual CSSs were sufficient to localize significant QTLs on several chromosomes (SINGER et al. 2004; SINGER et al. 2005).

The present study investigates both the dietary specificity and the genetic complexity of the decreased body weight conferred by several A/J-derived chromosomes in the CSSs. First, a low-fat diet CSS survey and replicate high-fat diet CSS surveys demonstrated that many CSSs are models specifically for resistance to high-fat, diet- induced obesity and that the resistance associated with individual A/J chromosomes is reproducible. Next, to test whether we could determine the number and location of resistance QTLs on each A/J chromosome and to test whether we could identify QTLs that were not detected in the surveys, we analyzed intercross progeny derived from each

CSS. Surprisingly, the intercrosses revealed A/J-derived, resistance QTLs on only three chromosomes. In addition, two A/J-derived, obesity promoting QTLs were detected.

Furthermore, the crosses provided evidence indicating that more than one QTL probably explains the resistance conferred by some A/J chromosomes, in particular chromosome 6.

Thus, even in CSSs, obesity resistance is a highly complex trait.

B. MATERIALS AND METHODS

1. Mouse husbandry: C57BL/6J and A/J males for all studies except replicate CSS survey #2 were obtained from the Jackson Laboratory (Bar Harbor, ME) at 4 weeks of age. For replicate CSS survey #2, colonies were established at CWRU using C57BL/6J and A/J from the Jackson Laboratory (Bar Harbor, ME), and male mice generated from 70 these colonies were used for the analysis. All B6-ChrA CSSs were raised at CWRU. F2 males (n = 78 to 93) were derived from intercrosses involving B6-ChrA CSS (except B6-

Chr YA and B6-Chr MitoA) and C57BL/6J. The F2 crosses were made in one direction

(CSS female x C57BL/6J male) to control for possible grand-parental effects. For the reciprocal B6-Chr MitoA crosses, B6-Chr MitoA female x C57BL/6J male and C57BL/6J female x B6-Chr MitoA male crosses were made, and male progeny were collected. All mice were weaned at three to four weeks of age and raised on LabDiet 5010 autoclavable rodent diet (LabDiet, Richmond, IN) ad libitum. Animals were housed in micro-isolator cages with 12:12 hour light:dark cycle.

2. Diet studies: For all diet studies, the mice were introduced to the study diet (described below) at approximately 35 days of age, fed ad libitum for approximately 100 days, and weighed at two-week intervals. In CSS replicate study #2, the mice were maintained on the diet for approximately 112 days.

a. Diets: For all diet studies, one of the following four diets was used: high-fat and simple carbohydrate (HFSC, Research Diets D12331, 58 kcal% fat, 26 kcal% carbohydrate – sucrose and maltodextrin, 16 kcal% protein), high-fat and complex carbohydrate (HFCC, Research Diets D12330, 58 kcal% fat, 26 kcal% carbohydrate – cornstarch and maltodextrin, 16 kcal% protein), low-fat and simple carbohydrate (LFSC,

Research Diets D12329, 11 kcal% fat, 73 kcal% carbohydrate – sucrose and maltodextrin, 16 kcal% protein), and low-fat and complex carbohydrate diets (LFCC,

Research Diets D12328, 11 kcal% fat, 73 kcal% carbohydrate – cornstarch and 71

maltodextrin, 16 kcal% protein). Table II-1 provides detailed information about the composition of each diet.

b. Phenotype measures: The following traits were measured: IW (weight at ~35 days of age), MW (weight at ~90 days of age), FW (weight at ~135 days of age), BMI (FW/(final nasoanal length)2), EWG (mean weight gain per day for first ~55 days), FWG (mean

weight gain per day for second ~45 days), and WG (mean weight gain per day). The

traits are shown in Figure II-1.

3. Genotyping:

a. B6-ChrA CSS F2 genome scan: Tail tissue was used for DNA isolation. Uniformly

spaced microsatellite markers and SNPs (226 markers total) were used for the whole

genome scan analysis (average inter-marker interval: 11 Mb; greatest inter-marker

distance: 40 Mb). For microsatellite markers, tail tissue was digested in proteinase K

(Invitrogen, Carlsbad, CA) in 1X PCR buffer (Invitrogen, Carlsbad, CA) overnight at

55ºC. The enzyme was inactivated at 100ºC for one hour prior to using the DNA for

genotyping studies. For the microsatellite markers, the reactions were performed in 1X

PCR buffer, 2 mM MgCl2 0.3mM of each dNTP, 0.04 units of Taq polymerase/µL and

0.4 - 0.5 µM of forward and reverse primers in 25 µL total reaction volume. The reaction

conditions were: 94°C for 2 minutes, 94°C for 1 minute, 60°C for 1 minute, 72°C for 1 minute, repetition of steps 2-4 34 times, and 72°C for 5 minutes. For microsatellite

markers, polyacrylamide gel (6%) electrophoresis was used to separate the products, and

001.5 100.0 1000.1 5558.5 5.56 Total Total kcal/gram ies of diets is referredof to as ies

100.0

1000.1 5558.5 5.56 ew Burnswick, Thisew ser NJ). Total Total kcal/gram

uke University). uke University).

FD&C Red Dye #40 0.05 0 obtained from Research Diets (N 100.0 for Richard Surwit (D for Richard 1366.6 5557 4.07 Total Total kcal/gram

se they were originally formulated originally they were se 100.0

4.07 gm% kcal% gm% kcal% gm% kcal% gm% kcal% 1366.6 5557

Total Total kcal/gram

D12328 (LFCC) D12329 (LFSC) D12330 (HFCC) D12331 (HFSC) the Surwit diet series becau series the Surwit diet Table II-1. Compositionhigh-fat Diets were diets. of and low-fat Ingredients:Casein, 80 MeshDL-MethionineMaltodextrin10 CornStarch Sucrose gm% 228 kcal%SoybeanOil 912Coconut Oil, Hydrogenated 2 170 Ingredients:Mineral S10001 Mix Casein, 80 Mesh 40 0 680 Sodium Bicarbonate 835Potassium Citrate, 1 H20 360 3340 DL-Methionine Maltodextrin 10Vitamin V10001 Mix 0 25 Coconut Bitartrate Oil,Choline Hydrogenated Corn Starch 40 10.5 4 gm% 225 228 #5 Dye Yellow FD&C 0 0 0 0 kcal% 40 Soybean 10 Oil 912 Sucrose Mineral S10001 Mix Sodium Bicarbonate 170 Potassium Citrate, 1 H20 2 360 40 Ingredients: 0.1 Casein, 80 2Mesh Coconut Oil, 680 Hydrogenated 0 Vitamin V10001 Mix 0 0 0 Maltodextrin 10 4 10.5 Bitartrate Choline 40 #1 Dye Blue FD&C 333.5 DL-Methionine 0 25 3001.5 gm% 228 0 835 0 Corn Starch 10 0 225 Coconut Oil, Hydrogenated kcal% Potassium Citrate, 1 H20 3340 Bicarbonate Sodium 912 Mineral S10001 Mix 0.1 Soybean Oil 40 333.5 2 Sucrose 170 Casein, 80 Mesh Ingredients: 3 Vitamin V10001 Mix 0 680 2 0 4 10.5 #1 Dye Blue FD&C Maltodextrin10 175 40 Bitartrate Choline 0 0 228 gm% 0 700 10 25 DL-Methionine 0 Potassium Citrate, 1 H20 kcal% 912 Sodium Bicarbonate CornStarch 0.05 225 Mineral S10001 Mix 0 40 170 2 SoybeanOil Vitamin V10001 Mix 4 0 680 0 10.5 2 #40 Dye Red FD&C 0 40 Sucrose 0 0 0 Bitartrate Choline 10 0 0 25 0.1 0 40 225 0 2 175 700 0 ProteinCarbohydrateFat 74.3 16.8 73.1 16.4 Carbohydrate 4.8 Protein 10.5 Fat 74.3 73.1 16.8 Carbohydrate 16.4 4.8 Protein 10.5 35.5 Fat 25.5 Carbohydrate 23 16.4 35.5 Protein 35.8 25.5 58 Fat 23 16.4 35.8 58 73

Figure II-1. Time course for body weight studies. The various traits used in the body weight studies are depicted. IW = initial weight, MW = midpoint weight, FW=final weight, BMI=body mass index, EWG = mean weight gain per day in early half of study, FWG= mean weight gain per day in final half of study, WG = mean weight gain per day, IA = initial age, MA = middle age, FA=final age.

IW MW FW, BMI

(MW-IW) (FW-MW) EWG = FWG = (MA-IA) (FA-MA)

(FW-IW) (grams) weight Body WG = (FA-IA)

035 91 135

IA MA FA

Age (days)

the products were visualized under ultraviolet light with ethidium bromide (Gel Doc

2000, Bio-Rad, Hercules, CA).

For SNPs, DNA was isolated (Qiagen DNeasy Kit), and the SNPs were genotyped

using the Sequenom MassArray 7K system and MALDI-TOF MS (matrix-assisted laser

desorption/ionization time-of-flight mass spectrometry). For each marker scored with

mass spectrometry, genotypes were scored independently with both the Sequenom

software (version 3.3) and also with a second method developed by Dr. Joel Hirschhorn

(unpublished). Generally, only markers in which 90% of the typed individuals had

concordant, positive (not "no call") genotypes by both methods were used in the analysis.

Among the concordant calls of usable markers, we required the distribution of genotypes

at each locus to be consistent with the expected distribution of genotypes in an F2 cross,

which was evaluated with a chi-square test at p=0.0016, the Lander-Kruglyak threshold

for an intercross (free model) (LANDER and KRUGLYAK 1995). All markers and p values

derived from the chi-square analysis are provided in Appendix I.

b. B6-Chr MitoA: To confirm that the B6-Chr MitoA strain had A/J-derived mitochondria, a mitochondrial polymorphism in CoIII (position 9348, C57BL/6J=G and

A/J=A) (BAYONA-BAFALUY et al. 2003), a subunit of the mitochondrial cytochrome c

oxidase enzyme, was genotyped in C57BL/6J, A/J, and B6-Chr MitoA mice (tail tissue) using the same procedure described in the previous section. The following primers were used: CGAAACCACATAAATCAAGCCC (forward) and

CTCTCTTCTGGGTTTATTCAGA (reverse). PCR products (10 µL) were digested with

Asp I (Roche Diagnostics, Mannheim, Germany) using 5 units of enzyme in 1X dilution 75

buffer B. Asp I digests the C57BL/6J-derived allele. The products from the digestion

were separated using agarose gel electrophoresis (2%) and visualized under ultraviolet

light with ethidium bromide (Gel Doc 2000, Bio-Rad, Hercules, CA).

4. Statistical analyses: Unless otherwise noted, all statistical analyses were performed

with the statistical software R (IHAKA and GENTLEMAN 1996). QTL analysis were

performed with R/QTL (BROMAN et al. 2003), an add-on package for R.

a. C57BL/6J and A/J diet studies: A standard three-way ANOVA was used to test the

effects of strain, dietary fat content, and dietary carbohydrate composition on FW, MW,

WG, EWG, and FWG in C57BL/6J and A/J males.

b. B6-ChrA CSS LFCC and HFSC diet surveys: For each diet, IW, MW, FW, EWG,

FWG, and WG from A/J and each CSS were compared to C57BL/6J using an unpaired t- test. To test whether the response of a CSS to the high-fat diet was different from the

response of C57BL/6J to the high-fat diet, the difference in the mean trait value for a CSS

on the high-fat vs. low-fat diets was compared to that of C57BL/6J. P values, adjusted for the search across the 22 chromosomes and across the six traits, were determined using a permutation test, with 100,000 permutation replicates. Individuals were permuted with respect to their strain assignments, and the same t statistics used with the observed data were calculated with the permuted data; the maximum t statistic across strains and phenotypes was recorded for each replicate. The adjusted p value for a particular strain 76

and phenotype was the proportion of permutation replicates at which the maximal t

statistic (maximized across strains and phenotypes) exceeded the observed t statistic.

c. B6-ChrA CSS HFSC diet replicate surveys: Pearson’s correlation coefficients were calculated to evaluate the correlation between the mean IW, FW, and BMI from replicate survey #1 and replicate survey #2 (Statistica Software, version 6.0). For each of the two surveys, we assessed significance of the differences in IW, FW, and BMI, between each

CSS and the C57BL/6J strain as described for the HFSC and LFCC surveys. Differences

between the two replicate surveys were identified by the same approach used to assess

the difference between the high- and low-fat diets: an unpaired t test was used to compare

the CSS – C57BL/6J differences between the two replicates. P values, adjusted for the

search across the 22 chromosomes and across three phenotypes, were again calculated by

a permutation test, using 100,000 permutation replicates.

d. Single QTL analysis of F2 crosses: Mice that lost >10% of their body weight in a two-

week interval (suggestive of illness) were removed from the analysis. In the F2 progeny,

the traits that were analyzed included IW, MW, FW, BMI, WG, EWG, and FWG.

Pearson’s correlation coefficients were calculated for each trait pair in each F2 cross.

The FW of each F2 cross was compared to C57BL/6J using an unpaired t-test followed

by 20,000 permutations of the data set to determine significance thresholds that adjust for

the search across chromosomes.

An F test was used to compare the variance from each F2 population to C57BL/6J

(Graphpad Prism, version 3.02). To further investigate the variance, all C57BL/6J mice 77

(rather than only the contemporaneous sample) were pooled (an unpaired t-test indicated

that the samples did not differ). Then, the variance from this pooled sample was

compared to the variance of the intercross progeny derived from chromosomes on which

QTLs were expected or detected. For all variance analyses, the statistical threshold used

was p<0.0025 to account for multiple testing (0.0025=0.05/20 strains tested).

QTL analyses were performed using standard interval mapping (LANDER and

BOTSTEIN 1989). Age was used as a covariate for all body weight and BMI analyses.

Confidence intervals for each QTL were estimated by calculating 1.5 LOD support

intervals. To establish significance thresholds that account for the search across 20

chromosomes and seven traits, a permutation test was used, with 100,000 permutation

replicates. For this analysis, we used Mijk to indicate the maximum LOD score on

chromosome i with phenotype j in permutation replicate k, and Mik to indicate the

maximum LOD score across phenotypes of the Mijk. For a single chromosome, 1 – α

quantile of the Mik (across k) is a significance threshold adjusted for the phenotypes

analyzed but not for the number of chromosomes tested (no genome scan adjustment).

Because length of the chromosome varies, the genome-scan-adjusted thresholds were

also adjusted for chromosome length. For the genome scan adjustment, Li denotes the

genetic length of chromosome i, and L indicates the total length of all chromosomes. We

Li/L use the 1 – (1 – α) quantile of the Mik (across permutation replicates, k) as the threshold for chromosome i, adjusting for the number of phenotypes analyzed and the

search across the genome. We used the accepted thresholds for suggestive (α=0.63) and

significant (α=0.05) QTLs (LANDER and KRUGLYAK 1995).

e. B6-MitoA reciprocal crosses: An unpaired t-test was used to compare FW of male offspring from the reciprocal crosses (Graphpad Prism, version 3.0).

C. RESULTS

1. Genetic susceptibility to high-fat, diet-induced obesity in C57BL/6J males

To test whether the weight differences in C57BL/6J and A/J males are due to the

high-fat or carbohydrate content of the diet, male mice from both strains were fed one of

four diets that differ in fat and carbohydrate composition, and FW was analyzed. Using

the three-way ANOVA for FW, a strong strain effect (p < 2.0 X 10-16) was observed with

C57BL/6J males heavier than A/J males. Thus, the C57BL/6J males are genetically susceptible to increased FW regardless of the dietary composition. In addition, a strong fat, but not carbohydrate, effect was observed (p < 2.0 X 10-16). Therefore, regardless of the strain, mice fed the high-fat diets were heavier than those fed low-fat diets. A strong interaction between strain and fat, but not strain and carbohydrate, was also observed for

FW (p < 2.0 X 10-16). Consequently, although C57BL/6J males were heavier than A/J

males on all four diets, the differences are exaggerated on the high-fat vs. low-fat diets.

Thus, C57BL/6J males were genetically susceptible to high-fat, diet-induced obesity.

Similar results were obtained for MW, EWG, FWG, and WG except that significant interactions between fat and carbohydrate content were also detected for EWG and WG, but these effects were not nearly as strong as the strain, fat, and strain-fat interaction effects observed (Figure II-2 and II-3; Table II-2). These results were similar to those obtained in previous studies (SURWIT et al. 1995).

Figure II-2. C57BL/6J (B6) and A/J weight gain on diets that differ in fat and carbohydrate composition. Approximately 24-30 male mice per strain were raised on each of four diets and weighed at two week intervals. The mean body weight for each strain-diet combination at each time point is plotted. (Abbreviations: HFSC = high-fat, simple carbohydrate diet; HFCC = high-fat, complex carbohydrate diet; LFSC = low-fat, simple carbohydrate diet; LFCC = low-fat, complex carbohydrate diet)

B6-HFSC 55 B6- HFCC B6-LFCC 50 B6-LFSC AJ-HFSC 45 AJ-HFCC AJ-LFCC AJ-LFSC 40

Weight (grams) Weight 25

0 0 102030405060708090100110120130140150 Age (days)

Figure II-3. C57BL/6J (B6) and A/J body weight (MW, FW, EWG, FWG, and WG) when fed diets that differ in fat and carbohydrate composition. Abbreviations: HFSC = high-fat, simple carbohydrate diet; HFCC = high-fat, complex carbohydrate diet; LFSC = low-fat, simple carbohydrate diet; LFCC = low-fat, complex carbohydrate diet. Twenty-four to thirty mice were used per strain.

50 60

50 40

40 30

Mid-weight (grams) 20

Final body weight (grams) 10 10

0 0 B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC Strain and diet Strain and diet 0.4 0.5

)

2 0.4 0.3

0.3 0.2 0.2

Weight gain (grams/day) gain Weight

0.1 Body mass index (grams/cm 0.1

0.0 0.0 B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC Strain and diet Strain and diet

0.5 0.4

0.4 0.3

0.3 0.2

0.2 0.1

Final weight gain (grams/day) gain weight Final Early weight gain(grams/day) 0.0 0.1

0.0 -0.1 B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC B6-HFSC B6-HFCC B6-LFSC B6-LFCC AJ-HFSC AJ-HFCC AJ-LFSC AJ-LFCC Strain and diet Strain and diet

Table II-2. Three-way ANOVA tables for C57BL/6J and A/J fed diets with varied fat and carbohydrate composition.

Mid Weight df Sum Sq Mean Sq. F value Pr(>F) Strain 1 1132.41 1132.41 138.9408 < 2.2e-16 Fat 1 1983.20 1983.20 243.3272 < 2.2e-16 Carbohydrate 1 0.74 0.74 0.0902 0.76412 Strain:Fat 1 240.52 240.52 29.5101 1.25E-07 Strain:Carbohydrate 1 4.00 4.00 0.4911 0.48403 Fat:Carbohydrate 1 2.03 2.03 0.2487 0.6184 Strain:Fat:Carbohydrate 1 35.56 35.56 4.3626 0.03768 Residuals 267 2176.14 8.15

Final Weight df Sum Sq Mean Sq. F value Pr(>F) Strain 1 4874.8 4874.8 336.9329 < 2e-16 Fat 1 5375.8 5375.8 371.5609 < 2e-16 Carbohydate 1 10.7 10.7 0.7385 0.39091 Strain:Fat 1 1381.5 1381.5 95.4838 < 2e-16 Strain:Carbohydrate 1 3.9 3.9 0.2669 0.60586 Fat:Carbohydrate 1 69.1 69.1 4.7748 0.02975 Strain:Fat:Sugar 1 64.2 64.2 4.4399 0.03604 Residuals 267 3863 14.5

Early Weight Gain df Sum Sq Mean Sq. F value Pr(>F) Strain 1 0.37673 0.37673 217.1116 < 2.2e-16 Fat 1 0.96209 0.96209 554.4582 < 2.2e-16 Carbohydrate 1 0.00577 0.00577 3.3258 0.06932 Strain:Fat 1 0.14022 0.14022 80.8096 < 2.2e-16 Strain:Carbohydrate 1 0.00074 0.00074 0.4267 0.51417 Fat:Carbohydrate 1 0.04996 0.04996 28.7946 1.75E-07 Strain:Fat:Carbohydrate 1 0.00605 0.00605 3.487 0.06295 Residuals 267 0.46329 0.00174

Weight Gain df Sum Sq Mean Sq. F value Pr(>F) Strain 1 0.47108 0.47108 423.3437 < 2.2e-16 Fat 1 0.67753 0.67753 608.8731 < 2.2e-16 Carbohydrate 1 0.00122 0.00122 1.0981 0.2956 Strain:Fat 1 0.1658 0.16580 148.9996 < 2.2e-16 Strain:Carbohydrate 1 0.00237 0.00237 2.1271 0.1459 Fat:Carbohydrate 1 0.03 0.03000 26.9558 4.13E-07 Strain:Fat:Carbohydrate 1 0.00182 0.00182 1.6367 0.2019 Residuals 267 0.29711 0.00111

Final Weight Gain df Sum Sq Mean Sq. F value Pr(>F) Strain 1 0.65062 0.65062 350.5718 < 2.2e-16 Fat 1 0.39274 0.39274 211.6208 < 2.2e-16 Carbohydrate 1 0.00015 0.00015 0.0831 0.773338 Strain:Fat 1 0.22079 0.22079 118.9693 < 2.2e-16 Strain:Carbohydrate 1 0.01901 0.01901 10.2424 0.001538 Fat:Carbohydrate 1 0.01439 0.01439 7.7512 0.005751 Strain:Fat:Carbohydrate 1 0.00006 0.00006 0.0326 0.856947 Residuals 267 0.49552 0.00186 82

2. Obesity resistance in CSSs is specific to the high-fat diet

Because the C57BL/6J males are susceptible to elevated body weight regardless

of the diet, we hypothesized that the QTLs detected in the CSS survey may be either

independent or dependent of the diet consumed (Table II-3). To test this hypothesis, we

analyzed the CSSs on the LFCC diet (Table II-4). The LFCC diet has similar

composition as our maintenance diet, and we (see above) and others have demonstrated

that this diet does not induce the extreme obesity observed in C57BL/6J males fed the

HFSC diet (SURWIT et al. 1995).

To test whether the reduced FW in the HFSC survey was the result of the diet, we

calculated the effect of the HFSC diet on FW. For example, the effect of the high-fat diet

can be calculated by comparing the FW for each strain on the high-fat vs. low-fat diets.

If the difference in the FW on the two diets was similar between a CSS and C57BL/6J,

we concluded that there was no diet effect for that CSS and that the decreased FW may

be due to differences in body weight independent of the diet. Alternatively, if the

difference in the FW attained on the two diets varied between a CSS and C57BL/6J, we

concluded that the diet effect was different between the strains and that the reduced FW

was probably due to resistance to high-fat, diet-induced obesity.

Of the 17 CSSs with decreased FW in the HFCC survey, we detected a significant

diet effect in 10 CSSs. Thus, relative to C57BL/6J, these 10 CSSs appear resistant to

HFSC diet-induced obesity. In contrast, the effect of the high-fat diet was not

significantly different from C57BL/6J in the remaining seven CSSs. Thus, in these seven

mber of the A/J

00 days.00 The traits ndard deviation, nd=not nd=not deviation, ndard

Males from each CSS were fed the HFSC dietfed thefrom for Males approximatelywere each CSS 1

A/J 24 17.90 1.51 nd 27.46 2.67 nd 31.62 3.82 nd Strain n Mean Dev Std Value P Mean Dev Std value P Mean Std Dev value P CSS-1CSS-2 17CSS-3 12CSS-4 16CSS-5 20 13.73CSS-6 20 22.24CSS-7 12 20.26CSS-8 18 13.72CSS-9 2.74 20 16.42 1.90 17 18.50 <0.0001 2.63 18.62 0.0164 2.77 17.70 0.9262 4.12 <0.0001 23.93 1.83 34.83 0.5012 3.17 32.34 3.13 1 28.96 1.79 27.49 1 <0.0001 1 4.68 32.59 3.45 0.9394 2.58 3.13 28.76 33.61 27.77 1 <0.0001 0.018 3.50 28.34 41.59 1 2.79 31.58 3.62 31.56 2.18 43.55 5.21 2.89 0.028 <0.0001 1 0.0006 3.11 37.40 1 4.19 5.19 <0.0001 30.47 32.93 <0.0001 33.17 40.60 1 4.28 2.69 4.25 0.0206 3.55 <0.0001 <0.0001 5.50 <0.0001 0.9994 CSS-Y 14 16.55 3.33 0.8022 28.69 2.15 0.012 34.82 4.38 0.0001 CSS-X 14 19.03 1.88 1 36.12 5.54 0.1929 47.52 4.65 0.1556 CSS-10CSS-11 13CSS-12 16CSS-13 12CSS-14 20 17.37CSS-15 17 19.88CSS-16 17 14.74CSS-17 20 19.35CSS-18 2.79 15.24 17CSS-19 1.92 10 15.42 2.39 16.89 16 0.9984 1 2.17 16.06 0.0112 3.89 20.03 2.89 18.09 0.0194 1 3.51 0.0384 28.94 1.42 0.9125 28.55 30.45 1.58 0.286 1.82 0.9993 29.83 31.83 30.91 2.98 31.40 1 3.16 3.70 0.0168 27.73 27.32 0.0147 0.7596 2.37 1.99 2.47 0.1911 3.17 35.40 34.41 0.8806 33.77 35.16 1.28 0.9979 1 1.86 35.56 0.0001 0.0006 5.50 36.04 3.43 3.40 7.02 36.73 0.0002 34.35 <0.0001 0.9992 33.49 3.95 0.0004 35.36 4.56 0.0002 4.42 1.86 0.0004 43.79 2.57 0.0038 4.25 <0.0001 <0.0001 0.0036 6.79 1 CSS-Mito 23 15.22 2.36 0.0062 30.78 1.95 0.7446 36.21 3.47 0.0003 C57BL/6J 20 18.54 1.39 nd 32.89 3.86 nd 42.71 5.97 nd HFSC DietHFSC (grams) IW MW (grams) (grams) FW from each CSS were compared to C57BL/6J using an unpaired t-test, and the p values are listed. Each CSS is indicated by the nu the by indicated is Each CSS listed. are p values and the t-test, unpaired an using C57BL/6J to compared were CSS each from chromosome is substituted that onto thebackground. C57BL/6J Statistical thresholdused is p<0.05. Abbreviations: std dev = sta determined. Table II-3A. Initial CSS survey: body HFSC diet traits. weight 84

mber of the A/J

00 days. The traits traits The days. 00 ndard deviation, nd=not nd=not deviation, ndard

for approximately 1 for approximately

re fed the HFSC diet re fed

weight gain traits. Males from each CSS we weight gain traits. Males from each CSS

A/J 24 0.14 0.03 nd 0.17 0.04 nd 0.10 0.04 nd Strain n Mean Std Dev value P MeanDev Std value P Mean Dev Std value P CSS-1CSS-2 17CSS-3 12CSS-4 16CSS-5 20CSS-6 0.27 20CSS-7 0.21 12CSS-8 0.12 18CSS-9 0.18 20 0.05 0.20 17 0.04 0.14 0.9509 0.05 0.12 0.8279 0.03 0.16 <0.0001 0.05 0.19 0.0011 0.03 0.0414 0.03 0.37 <0.0001 0.20 0.04 <0.0001 0.18 0.04 <0.0001 0.21 0.0138 0.27 0.18 0.09 0.17 0.04 0.06 0.0072 0.21 0.0023 0.04 <0.0001 0.22 0.06 0.03 0.0003 0.16 0.05 <0.0001 0.22 0.07 0.05 1 <0.0001 0.05 0.0008 0.13 0.06 0.09 0.004 0.05 0.07 0.06 0.9251 0.10 0.10 <0.0001 0.06 1 0.04 0.17 0.0557 0.05 0.0001 0.07 0.04 <0.0001 0.0001 <0.0001 0.05 0.9817 CSS-Y 14 0.17 0.02 0.0001 0.24 0.05 0.543 0.11 0.05 0.0021 CSS-X 14 0.28 0.04 0.6613 0.31 0.05 1 0.25 0.06 0.5976 CSS-10CSS-11 13CSS-12 16CSS-13 12CSS-14 20 0.18CSS-15 17 0.16CSS-16 17 0.20CSS-17 20 0.17CSS-18 17 0.06 0.19CSS-19 10 0.05 0.18 0.0027 16 0.06 0.20 <0.0001 0.04 0.16 0.1029 0.03 0.17 <0.0001 0.03 0.26 0.0117 0.05 0.23 0.18 0.0004 0.03 0.1113 0.04 0.21 <0.0001 0.20 0.05 0.0002 0.26 0.07 0.25 0.05 1 0.25 0.2223 0.24 0.08 <0.0001 0.04 0.17 0.0063 0.07 0.0001 0.05 0.11 0.15 0.9126 0.07 0.27 0.04 0.6584 0.18 0.3849 0.03 0.12 0.406 0.08 <0.0001 0.12 0.07 0.06 0.05 0.0023 0.4154 0.11 0.14 0.06 0.10 0.9997 0.16 0.06 0.0038 1 0.02 0.0051 0.06 <0.0001 0.03 0.25 0.06 0.2038 0.0001 0.9881 0.09 0.4766 CSS-Mito 23 0.22 0.04 0.9178 0.29 0.06 1 0.13 0.05 0.0184 C57BL/6J 20 0.25 0.05 nd 0.29 0.06 nd 0.20 0.06 nd HFSC Diet WG (grams/day) EWG (grams/day) FWG (grams/day) Table II-3A. Initial HFSC diet CSS survey: Table II-3A. Initial HFSC diet CSS survey: from each CSS were compared to C57BL/6J using an unpaired t-test, and the p values are listed. Each CSS is indicated by the nu the by indicated is Each CSS listed. are p values and the t-test, unpaired an using C57BL/6J to compared were CSS each from chromosome is substituted that onto thebackground. C57BL/6J Statistical thresholdused is p<0.05. Abbreviations: std dev = sta determined. 85

mber of the A/J of the A/J mber ays.traits The

andard deviation, nd=not nd=not deviation, andard

approximately 100 d

fed the LFCC diet for the fed

ight traits. Male mice from each CSS were ight traits.

A/J 29 18.60 1.52 nd 24.55 1.79 nd 26.35 2.13 nd Strain n Mean Std Dev P Value Mean Std Dev P value Mean Std Dev P value CSS-1CSS-2CSS-3 17CSS-4 18CSS-5 14CSS-6 18 16.20CSS-7 20 15.13CSS-8 17 12.59CSS-9 15 14.87 3.79 18 15.68 3.49 20 16.75 0.0207 2.17 17.64 0.0001 2.25 <0.0001 18.61 2.89 <0.0001 16.16 4.12 0.0008 28.29 3.18 0.1481 26.12 2.33 23.71 0.9203 3.07 24.11 0.0087 1 25.84 1.86 26.07 1.58 0.90 27.67 0.8778 2.34 0.9967 <0.0001 24.46 1.07 0.0001 26.60 1.92 0.8308 30.53 2.52 0.9934 26.34 28.62 2.54 26.56 1 1.65 0.0006 29.34 2.67 28.24 1.33 2.04 1 <0.0001 2.47 0.0349 1 28.99 26.66 2.12 <0.0001 2.12 0.4705 29.22 0.0060 2.05 2.35 <0.0001 0.2742 2.01 0.3938 CSS-Y 21 17.70 2.34 0.8458 26.79 2.06 1 29.05 2.21 0.1617 CSS-X 19 18.83 1.54 1 26.42 1.25 1 28.97 1.87 0.1459 CSS-10CSS-11CSS-12 20CSS-13 21CSS-14 17CSS-15 20 16.07CSS-16 18 17.43CSS-17 19 15.56CSS-18 13 15.20 3.26CSS-19 20 18.43 3.32 18 19.75 0.0061 3.51 13 16.59 0.5743 2.12 16.86 0.0011 2.56 <0.0001 16.91 3.94 20.28 24.23 1.42 1 25.97 1.43 1 0.1791 25.64 3.93 26.08 0.1419 2.24 0.2127 1.91 1.69 1 26.31 0.0001 26.90 2.18 2.42 28.69 25.46 0.9370 25.89 0.6362 0.9874 25.77 1.72 1.40 28.05 27.92 2.08 1.74 28.54 2.92 28.88 1 0.2888 1 0.284 0.9165 2.09 1.77 1.72 <0.0001 1.79 2.41 28.15 0.999 0.0002 31.43 28.09 29.40 29.41 0.0286 0.0856 30.80 2.41 2.77 1.40 3.56 1.23 0.0015 0.0017 0.6282 0.8413 1 2.54 1 CSS-Mito 23 18.93 1.92 1 28.55 1.70 0.3529 31.65 2.04 1 C57BL/6J 28 19.45 0.86 nd 27.06 1.27 nd 31.01 1.71 nd

LFCC Diet IW (grams) MW (grams) FW (grams) Table II-4A. LFCC diet CSS survey: body we LFCC diet CSS survey: Table II-4A. from each CSS were compared to C57BL/6J using an unpaired t-test, and the p values are listed. Each CSS is indicated by the nu the by indicated is CSS Each are listed. values p the and t-test, unpaired an using C57BL/6J to compared were CSS each from chromosome that is substituted onto the C57BL/6Jbackground. Statistical threshold used is p<0.05.Abbreviations: std dev = st determined. 86

ays.traits from The

of the A/J chromosome that of , nd=not determined. determined. nd=not ,

r approximately 100 d

Dev value P Mean Std Dev value P fed the LFCC diet fo fed the

d. Each CSS is indicated by the number is indicated by CSS d. Each

oldused p<0.05. is Abbreviations:dev std = standarddeviation

unpaired t-test, and the p values are liste t-test, unpaired gain traits. Male mice from each CSS were

A/J 29 0.08 0.02 nd 0.11 0.03 nd 0.04 0.05 nd Strain n Mean Std Dev value P Mean Std CSS-1CSS-2CSS-3 17CSS-4 18CSS-5 14CSS-6 18 0.15CSS-7 20 0.14CSS-8 17 0.14CSS-9 15 0.12 18 0.04 0.14 20 0.03 0.12 0.1351 0.02 0.11 0.8654 0.03 0.11 0.9448 0.04 0.10 0.03 0.5113 1 0.23 0.04 0.20 0.02 1 0.20 0.03 1 1 0.9982 0.18 0.06 0.16 0.05 <0.0001 0.04 0.18 0.0009 0.14 0.17 0.0079 0.06 0.15 0.04 0.05 0.0975 0.05 0.06 1 0.05 0.08 0.06 0.1022 0.04 0.5272 0.08 0.04 1 1 0.02 0.0007 0.02 0.05 0.06 0.0099 0.03 0.0542 0.05 0.05 0.06 1 0.02 0.03 0.04 0.0002 0.0199 <0.0001 0.03 0.03 0.0002 0.0125 CSS-Y 21 0.11 0.02 1 0.16 0.04 0.9628 0.05 0.02 0.0001 CSS-X 19 0.10 0.02 0.9234 0.13 0.03 1 0.06 0.02 0.0100 CSS-10CSS-11CSS-12 20CSS-13 21CSS-14 17CSS-15 20 0.09CSS-16 18 0.10CSS-17 19 0.13CSS-18 13 0.13CSS-19 20 0.04 0.10 18 0.03 0.12 0.3777 13 0.02 0.12 0.9941 0.03 0.11 0.02 0.13 0.9962 1 0.02 0.11 0.7034 0.14 0.02 0.15 0.03 1 0.03 1 0.18 0.02 1 0.14 0.19 0.07 1 0.04 1 0.17 1 0.05 0.18 1 0.04 0.05 0.15 0.0502 0.16 0.0081 0.14 1 0.03 0.03 0.03 0.4528 0.04 0.06 0.04 0.3198 0.05 0.06 0.04 1 0.9978 0.04 0.03 0.06 1 0.02 0.02 0.05 <0.0001 0.02 <0.0001 0.0288 0.08 0.0201 0.03 0.06 0.03 <0.0001 0.02 0.06 0.1114 0.0201 0.03 0.03 1 0.0239 0.03 0.2160 CSS-Mito 23 0.13 0.03 0.9999 0.17 0.04 0.3232 0.07 0.03 0.5971 C57BL/6J 28 0.12 0.02 nd 0.14 0.02 nd 0.09 0.03 nd each CSS were compared to C57BL/6J using an using C57BL/6J to compared were each CSS Table II-4B. LFCC diet CSS survey: weight LFCC diet CSS survey: Table II-4B. is substitutedonto C57BL/6Jbackground. the Statistical thresh

LFCC Diet WG (grams/day) EWG (grams/day) FWG (grams/day) 87 strains, the decreased FW may be due to a genetic predisposition for decreased body weight or body size independent of diet. Overall, because most of the QTLs detected were specific to the high-fat diet (Table II-5), the CSSs are a valid model for studying the genetics of high-fat, diet-induced obesity resistance.

3. 13 A/J chromosomes reproducibly confer resistance to diet-induced obesity

Many environmental factors such as diet lot and season of the year could potentially influence body weight in the CSSs. Thus, two, independent, replicate high-fat diet CSS surveys were performed to test whether the A/J chromosomes reproducibly confer resistance to obesity on the HFSC diet despite variation in these environmental factors. Although IW was poorly correlated (-0.22) across the replicate surveys, FW and

BMI were highly correlated between the two surveys (FW correlation coefficient: 0.90,

BMI correlation coefficient: 0.91, Figure II-4). Moreover, except for chromosome X, the effect (BMI and FW) conferred by each A/J chromosome relative to C57BL/6J did not significantly differ between the two surveys (Table II-6). Therefore, the replicate studies were highly correlated and hence, reproducible.

Relative to C57BL/6J, thirteen A/J chromosomes (3, 4, 5, 6, 7, 8, 10, 12, 13, 14,

16, 17, and Y) reproducibly conferred decreased FW in both surveys (Tables II-7 and II-

8). To test whether the decreased FW was associated with decreased adiposity, we measured BMI (an estimate of adiposity) in all strains. In both surveys, the 13 CSSs with reduced FW (except for B6-Chr YA/J in replicate #2) also had a significantly reduced

BMI relative to C57BL/6J indicating that the FW differences were likely due to decreased adiposity and hence, obesity resistance. 88

ckground. Statistical were compared to the compared to the were FWG FWG

(grams/day) T statisticvalue P trait value on LFCC) and

EWG

(grams/day) T statistic value P

its in each CSS on the HFSC vs. LFCC diets HFSC vs. LFCC its in each CSS on the

WG WG

(grams/day) T statistic value P

analysis. The differences between the tra the between differences The analysis.

IW (grams) T statistic value P (grams) MW T statistic value P FW (grams) T statistic value P

Table II-5. HFSC vs. LFCC diet CSS survey diet CSS HFSC vs. LFCC Table II-5. difference in C57BL/6J using an unpaired t-test. The difference between the mean trait values (mean trait value on HFSC – mean – mean HFSC on value trait (mean values trait mean the between The difference t-test. an unpaired using C57BL/6J in difference valuespthe are listed. Eachusing CSS is indicated number ofA/J chromosome the the substituted is that onto ba C57BL/6J the deviation. standard = dev std Abbreviations: p<0.05. is used threshold A/J -0.70 nd nd 2.92 nd nd 5.27 nd nd 0.06 nd nd 0.06 nd nd 0.06 nd nd Strain CSS-1CSS-2 -2.47CSS-3 7.11CSS-4 7.67CSS-5 1.262 -1.15CSS-6 -6.182 0.73CSS-7 -6.688 1.76CSS-8 0.203 0.98CSS-9 <0.0001 1 -1.394 -0.91 <0.0001 -2.041 7.77 -1.511 6.22 1 0.9999 0.000 5.25 6.54 0.9185 -7.191 0.9995 -0.324 6.75 <0.0001 0.490 3.37 2.70 1 -0.617 0.10 1 -0.832 9.14 2.217 2.585 1 1 1.74 4.949 14.93 1 -2.952 0.8206 0.5312 11.06 5.22 0.0004 3.682 -1.96 0.2663 5.02 4.69 8.07 1.48 0.41 0.0377 3.98 13.94 0.9478 4.40 4.23 2.43 3.95 0.0133 6.46 1 -1.46 0.08 0.0027 0.0053 0.6594 <0.0001 -0.02 5.11 0.9998 0.07 0.02 0.12 3.012 0.01 0.06 0.0002 0.09 8.708 0.2328 3.789 6.144 0.05 0.146 6.990 4.462 <0.0001 2.544 <0.0001 0.0263 0.00 <0.0001 0.0021 5.029 -0.01 1 0.5648 0.00 0.05 6.278 0.00 0.0002 0.09 7.061 0.07 0.14 <0.0001 6.263 <0.0001 4.614 6.639 0.06 2.926 3.790 0.17 <0.0001 0.568 <0.0001 -0.01 0.0011 0.2814 4.410 0.0261 0.04 0.03 -2.599 0.07 1 5.347 0.02 0.0027 0.11 0.5192 3.002 0.0001 3.607 1.615 0.04 0.11 4.270 -0.338 0.2377 0.0475 0.9977 0.0046 3.127 -0.118 1 0.1775 1 CSS-XCSS-Y 0.20 -1.16 -0.885 0.199 1 1 9.70 1.89 -3.324 3.427 0.1063 0.0801 18.55 5.77 -4.31 3.78 0.0039 0.0271 0.18 0.06 -3.143 4.071 0.1707 0.0097 0.17 0.08 -0.661 3.181 1 0.1548 0.19 0.06 -3.975 2.333 0.0137 0.7359 CSS-10CSS-11 1.29CSS-12 2.45CSS-13 -0.83CSS-14 -1.747 4.14CSS-15 -2.779 -3.19CSS-16 -0.064 -4.33CSS-17 0.9906 -4.286 0.30CSS-18 0.3791 1.859 -0.80CSS-19 2.813 3.11 1 6.22 0.0044 -0.976 -2.20 2.97 0.9747 -0.093 0.355 -2.992 4.83 -0.331 0.991 2.91 3.52 1 2.547 1 0.2435 3.15 0.917 1 0.5624 2.409 1 2.038 4.50 1.43 2.27 2.382 7.49 0.6764 1 0.9202 9.39 6.36 0.6987 1.161 3.527 5.24 3.183 7.48 2.74 1.44 7.16 2.92 0.0602 -0.436 0.1542 1 0.406 3.90 2.72 0.9998 5.96 3.04 5.70 5.34 1 0.0176 0.4224 7.32 0.06 0.09 0.2204 <0.0001 3.37 4.16 0.07 12.99 0.10 2.74 0.06 0.04 0.094 4.148 2.493 0.007 -0.78 0.4043 3.444 2.100 0.0072 0.04 0.6072 4.470 5.859 0.05 1 0.08 0.0762 0.8901 <0.0001 0.03 0.0021 0.09 4.979 4.818 3.006 0.15 0.01 0.12 0.02 0.0003 5.485 0.08 2.669 0.0006 0.2358 -1.474 <0.0001 5.776 1.660 0.4625 6.211 0.01 3.501 0.09 <0.0001 0.06 0.9962 0.10 0.9998 <0.0001 0.07 0.0648 5.562 0.12 2.905 0.08 0.06 3.866 0.13 0.233 0.00 <0.0001 1.556 0.2947 0.0197 -0.512 1.482 2.481 0.08 1.206 0.9989 1 5.351 0.04 0.09 1 0.9997 0.6167 1 0.0001 1.227 3.404 0.945 1 0.19 0.0855 1 -3.654 0.0412 CSS-Mito -3.71 2.460 0.6344 2.24 3.408 0.0846 4.56 4.95 0.0004 0.09 2.463 0.6316 0.12 1.800 0.9851 0.06 2.575 0.5391 C57BL/6J -0.91 nd nd 5.83 nd nd 11.70 nd nd 0.13 nd nd 0.16 nd nd 0.11 nd nd 89

Figure II-4. Correlations between replicate HFSC diet survey traits. The IW, FW, and BMI are plotted for each CSS (indicated by the number of the substituted chromosome), C57BL/6J (B6), and A/J from each replicate survey. Please note that the data points for chromosome 3 and 13 and chromosomes 5 and 14 are overlapped in the BMI graphic. The solid line represents the best-fit line for the data, and the dotted lines represent the 95% confidence intervals for the best-fit line.

A. Initial weight

22 B6 15 11 12 21 9

2 1 X 20 Y 7 10 14 6 16 19

4 19 17 8 18 3 5 18 13 CSS Replicate #2 IW(grams)

17 A/J

16 15.0 15.5 16.0 16.5 17.0 17.5 18.0 18.5 19.0 19.5 CSS Replicate #1 IW (grams)

B. Final weight

50 X 9 48 2 1 18 B6 46

44 15 11 42 19 Y 1214 40 16 5 38 10 8 13 4 36 17 6 7 34 3 CSS Replicate #2 FW (grams)

CSS #2 Replicate FW (grams) 32

30 A/J

28 28 30 32 34 36 38 40 42 44 46 48 50 CSSCSS Replicate Replicate #1 #1 FW FW (grams) (grams)

Figure II-4 (continued

C. Body mass index

0.44 X 18 9 0.42 2

)

2 B6 0.40 )

2 19 11 0.38 Y 15 12 0.36 14 16 5

8 0.34 4 10 6 3

Replicate #2 BMI (grams/cm 7 13 0.32 17 CSS Replicate #2 BMI (grams/cm BMI #2 Replicate CSS 0.30

A/J 0.28 0.28 0.30 0.32 0.34 0.36 0.38 0.40 0.42

2 CSS ReplicateReplicate #1 #1 BMI BMI (grams/cm (grams/cm) 2)

viations: nd=not nd=not viations:

red. Permutations of the the of Permutations red. ) T statistic value P 2 Each CSS is indicated using using CSS is indicated Each

SC replicate survey #1 vs. #2 were compa SC #1 vs. #2 were survey replicate

7BL/6J background. Statistical threshold used is p<0.05. Abbre . The trait difference and p value for each trait is presented. trait is presented. for each value p and difference trait . The

et CSS surveys.values TheHF in the trait

A/J 1.44 nd nd 0.39 nd nd 0.02 nd nd Strain (grams) IW T statistic value P FW (grams) T statistic value P (grams/cm BMI CSS-1CSS-2CSS-3CSS-4 -4.34CSS-5 -5.09CSS-6 0.10CSS-7 -1.09 0.95CSS-8 -1.50 1.776CSS-9 -1.66 -3.207 -2.35 -2.470 0.8685 1 -1.44 -1.972 0.0484 -4.82 -1.818 0.3304 -1.033 -2.87 0.7293 -34.59 -1.958 0.8418 -0.95 -1.00 1.665 -1.36 0.691 1 0.7406 -0.616 -1.40 -0.144 0.9250 -0.156 0.028 -1.02 1 1 0.048 -2.88 1 -4.22 1 1 -0.134 1 0.788 -0.01 1.421 0.00 -0.01 1 0.00 0.9883 1 -0.01 0.502 0.00 -0.363 0.399 -0.03 -0.011 0.00 0.428 1 -0.02 1 -0.281 1 1 1.414 -0.025 1 1 1.306 0.9892 1 0.9968 CSS-XCSS-Y -3.97 -2.01 0.679 -1.348 0.9948 1 -4.17 -8.87 1.368 3.367 0.9936 0.0288 -0.03 -0.06 1.359 3.143 0.9942 0.0584 CSS-10CSS-11CSS-12CSS-13 -1.50CSS-14 -4.49CSS-15 -4.25CSS-16 -1.70 -2.031CSS-17 -1.52 1.297CSS-18 -3.54 1.108 0.6800CSS-19 -3.64 -1.327 0.9972 -0.26 -2.438 0.9999 -0.99 -0.80 0.252 0.9960 -0.16 0.363 -3.30 0.3522 -3.490 -2.34 -0.159 -1.68 -2.692 1 -2.69 -3.333 0.961 0.0195 1 0.530 0.2005 1 0.141 0.0321 0.684 -0.75 1 -4.20 -2.13 1 -2.38 1 2.04 1 -0.287 0.00 1.283 0.403 0.482 -0.01 0.9977 -1.608 0.00 1 0.00 -0.483 1 0.9470 -0.01 1 0.730 -0.02 -0.248 1 -0.170 0.00 0.02 0.428 1 -0.02 -0.01 1 1.077 1 -0.175 1 -1.129 0.768 1 0.242 0.9998 1 1 1 CSS-Mito nd nd nd nd nd nd nd nd nd C57BL/6J -3.28 nd nd -1.30 nd nd 0.00 nd nd Table II-6. Comparison of replicate HFSC di HFSC of replicate Table II-6. Comparison data set were used to determine the p value for each comparison p value the determine used to data set were the number of the A/J chromosome that is substituted onto the C5 determined. 92

r each strain were strain were r each

) Statistical threshold used is 2

-5.725 <0.0001 0.35 0.03 -4.809 0.0001 -5.370 <0.0001 0.35 0.02 -4.714 0.0002

on the HFSC diet and IW, FW, and BMI fo BMI IW, FW, and HFSC diet and on the

10 -2.831 0.1705 0.37 0.05 -3.207 0.0614 .29 -3.744 0.0099 0.37 0.04 -2.546 0.3266 .49 -4.426 0.0008.92 0.35 -4.448 0.0007 0.04 0.36 -4.410 0.0008 0.04 -3.590 0.0176 dev), and p value dev), for each and p value strain are listed.

8 8.30 -0.062 1 0.41 0.06 0.239 1

#1. CSS males from each strain were raised #1. CSS males from each strain were test. The mean, standard deviation (std deviation (std standard mean, test. The

A/J 30 18.26 1.62 nd nd 31.32 3.15 nd nd 0.31 0.02 nd nd Strain n Mean Std Dev T statistic Value P Mean Std Dev T statistic Value P Mean Std Dev T statistic value P CSS-1CSS-2CSS-3CSS-4 26CSS-5 22 16 25 15.64 21 15.18 18.37 2.53 17.79 3.22 16.89 1.58 -4.218 1.09 -4.700 3.26 -0.115 0.0016 -0.980 0.0002 -2.206 45.83 1 44.11 1 0.5905 38.99 4.86 0.33 36.23 5.71 0.337 5 0.02 -0.886 4.18 1 1 -6.580 -6.663 <0.0001 <0.0001 0.41 0.41 0.33 0.33 0.03 0.05 0.02 0.03 0.136 0.536 -5.763 -6.557 1 <0.0001 1 <0.0001 CSS-6CSS-7CSS-8CSS-9 24 22 17 19 18.25 17.50 17.40 3.25 16.09 1.70 2.70 -0.296 3.63 -1.367 -1.397 -3.245 1 0.9965 0.9951 0.0548 31.78 35.44 37.12 44.15 3.00 4.64 4.19 5.88 -9.556 -7.158 -5.369 <0.0001 -0.824 <0.0001 <0.0001 1 0.30 0.34 0.34 0.03 0.39 0.04 0.03 -9.372 -6.130 -5.360 <0.0001 0.04 <0.0001 <0.0001 -1.052 1 CSS-XCSS-Y 14 18 16.23 17.95 1.95 2.87 -2.774 -0.682 0.1962 1 39.24 36.73 5 4.23 CSS-10CSS-11 25 19 18.49 16.52 1.61 2.09 0.052 -2.657 0.2559 1 38.77 37.92 4 6.42 -5.431 <0.0001 0.35 0.05 -5.195 <0.0001 CSS-12CSS-13CSS-14CSS-15 26CSS-16 20CSS-17 23CSS-18 22CSS-19 16.66 21 16.35 30 17.97 18 1.63 17.73 20 2.73 15.81 2.09 18.48 -2.690 3.02 17.79 -2.931 3.29 19.23 -0.026 2.04 0.2376 -1.044 1.98 0.1321 -3.741 1.90 38.55 0.035 1 34.48 -0.893 1 0.0100 1.089 1 37.83 37.30 3.70 1 0.9999 38.20 4.31 34.73 45.2 -5.019 44.78 2.43 3.52 -7.450 5.78 <0.0001 <0.0001 3.88 -5.599 6.60 0.37 -5.041 <0.0001 0.33 <0.0001 -8.127 -0.390 <0.0001 0.34 0.03 0.35 0.05 1 -3.460 0.32 -6.727 0.03 0.0275 0.04 <0.0001 0.41 -5.357 0.03 -5.110 <0.0001 <0.0001 -7.995 0.05 <0.0001 0.747 1 CSS-Mito 31 18.36 2.08 -0.150 1 41.69 6. C57BL/6J 29 18.45 1.38 nd nd 45.37 5.25 nd nd 0.40 0.04 nd nd Replicate #1Replicate (grams) IW (grams) FW (grams/cm BMI compared to C57BL/6J using an unpaired t- unpaired using an C57BL/6J to compared p<0.05. Table II-7. HFSC diet replicate CSS survey Table II-7. HFSC diet replicate 93

or each strain strain or each

sted. ) 2

HFSC diet and IW, FW, and BMI f BMI and FW, IW, and diet HFSC

eachraised on strain the were

d t-test. The mean, standard deviation (std dev), and p value for each strain are li are strain each for value p and dev), (std deviation standard mean, The d t-test.

A/J 6 16.82 1.09 nd nd 30.94 2.91 nd nd 0.29 0.02 nd nd Strain n Mean Dev Std T statistic Value P Mean Dev Std T statistic Value P Mean Std Dev T statistic value P CSS-1CSS-2CSS-3 7CSS-4 11CSS-5 11CSS-6 19.98 23CSS-7 20.26 24CSS-8 18.26 23CSS-9 18.88 23 1.33 18.40 2.55 22 19.91 1.42 22 19.85 1.88 18.84 -0.900 1.84 -2.370 20.91 1.10 -2.330 2.79 -5.495 0.4084 1.09 -5.536 0.4384 1.85 <0.0001 -6.548 <0.0001 -3.539 46.78 <0.0001 -3.662 46.98 34.92 -5.562 0.0196 37.22 0.0129 3.90 <0.0001 40.35 6.27 5.08 0.9525 36.84 3.48 34.67 0.93 3.68 38.15 0.05 0.18 3.41 48.37 -6.81 4.25 -6.72 <0.0001 1 4.30 1 <0.0001 -4.54 4.87 0.0005 -7.00 -8.54 <0.0001 0.33 <0.0001 -6.00 0.34 <0.0001 0.42 0.42 0.36 0.9992 0.33 0.32 0.04 0.34 0.02 0.04 0.05 0.03 0.42 1.114 0.03 -5.815 -0.495 0.03 0.589 <0.0001 -6.522 0.03 <0.0001 -4.015 0.03 -6.551 0.9998 0.0036 1 -7.752 <0.0001 <0.0001 <0.0001 -5.929 1 CSS-XCSS-Y 16 20 20.19 19.96 1.06 0.82 -4.336 -2.728 0.1932 0.0374 48.12 40.90 3.06 4.15 -2.33 0.93 1 0.0043 0.37 0.43 0.03 0.03 1.891 -1.373 0.0772 0.7891 CSS-10CSS-11CSS-12 25CSS-13 20CSS-14 24 19.99CSS-15 6 21.01CSS-16 20 20.91CSS-17 11CSS-18 1.44 18.06 12 19.49CSS-19 1.60 24 21.26 1.37 24 19.45 19 -1.584 0.89 18.74 1.74 -3.459 18.59 1.58 -1.354 19.40 1.05 0.0255 -1.608 2.44 -4.692 0.9935 1.85 -4.213 0.9444 1.65 -0.748 38.91 0.0002 -3.708 0.0018 42.07 -5.872 40.89 -6.167 1 0.0112 5.96 36.16 <0.0001 5.86 40.52 <0.0001 4.84 1.19 0.0011 39.68 5.25 -5.63 35.48 4.21 42.41 -3.17 46.91 <0.0001 0.0625 3.05 -4.15 43.24 4.44 -4.91 0.0022 6.50 4.89 0.0001 -2.47 0.34 0.0016 5.82 -4.16 -4.23 0.38 -8.04 0.0021 0.37 <0.0001 0.17 0.3367 0.33 0.36 0.04 1 0.4368 0.05 0.35 0.32 0.04 0.953 0.37 0.04 -5.946 0.03 <0.0001 -2.669 0.39 0.43 0.2220 -3.783 0.02 0.03 -4.882 0.0088 0.05 0.0001 -2.852 -4.348 0.0035 0.05 -4.023 0.04 <0.0001 0.0010 0.1438 1.968 -7.889 0.7309 0.9919 CSS-Mito nd nd nd -3.337 nd nd nd -3.97 nd nd nd -3.090 nd C57BL/6J 20 21.73 1.21 nd nd 46.68 3.98 nd nd 0.41 0.03 nd nd Replicate #2 (grams) IW FW (grams) BMI (grams/cm Table II-8. HFSC diet replicate CSS survey #2. CSS males from males CSS #2. CSS survey replicate HFSC diet Table II-8. were compared to C57BL/6J using an unpaire to C57BL/6J compared were 94

Of the 13 chromosomes that consistently conferred decreased body weight and BMI, eight had a significant diet effect in the previous HFSC vs. LFCC CSS analysis. Thus, at least eight genes that confer resistance to high-fat, diet-induced obesity were detected in the CSS surveys.

4. Number and location of QTLs on substituted chromosomes

Although we reproducibly demonstrated which CSSs are resistant to obesity, the

CSS surveys do not provide any clues to the localization of QTLs on individual A/J chromosomes. Thus, to estimate the number and establish the location of QTLs on the

A/J substituted chromosomes and to test whether we could detect QTLs that eluded discovery in the CSS surveys, we performed a genome scan using intercross progeny derived from each CSS (except Y and mito). In addition to FW and BMI, traits measured at the beginning and middle of the time course were assessed to test whether we could determine when the QTLs first produced detectable effects.

Before performing linkage analysis, we tested whether any genetic markers used in the analysis deviated from Mendelian expectations (Appendix I). The genotypes at only one marker near the telomere on chromosome 10 differed significantly from expectations. This marker inflated the genetic map in the region because more than the expected number of recombination events occurred between markers near the end of the chromosome. For the mapping analysis, the distance between the markers near the telomere was set to the expected genetic distance based on their physical location.

We also tested whether the traits analyzed were highly correlated in each of the

F2 crosses. Although the strength of the correlations among most traits measured in the

F2 crosses varied, some general trends were observed. For instance, traits measured at the end of the study, such as FW, WG, and BMI were highly correlated (FW vs. BMI,

>0.8 in all crosses, range = 0.82 – 0.97; FW vs. WG, > 0.7 in all crosses, range = 0.75 –

0.96; BMI vs. WG > 0.7 in all crosses, range=0.73-0.92). FW was also highly positively correlated with MW (>0.8 in all crosses, range = 0.81 – 0.93). In contrast, the correlations between EWG and FWG were variable but usually weak (<0.6 in all crosses, range = 0.09 – 0.60). Similarly, the correlations between IW and other traits were highly variable and generally weak. Therefore, weight gain in the early stage of the study does not appear to be strongly related to weight gain in the second stage of the study, and IW is not necessarily predictive of FW (Table II-9; graphics of trait correlations are in

Appendix II).

We used two different statistical strategies for analyzing the linkage data for the

F2 crosses. First, the F2 crosses were treated as twenty independent crosses, and the p values were adjusted only for the seven traits analyzed in each cross. With this approach, significant QTLs were detected on six chromosomes and suggestive QTLs were detected on 18 chromosomes. The significant QTLs were on chromosomes 1 (FW, MW, BMI,

WG), 6 (FW, MW, BMI, WG, FWG), 10 (FW, MW, IW, WG, EWG), 11 (FW, BMI,

WG, FWG), 13 (FW, MW, BMI, WG, EWG, FWG), and 17 (MW). Suggestive QTLs were detected on all chromosomes except for chromosomes 9 and 14 (Table II-10).

Table II-9. Trait correlations for B6-ChrA CSS F2 crosses. Pearson’s correlation coefficients were calculated for each trait pair within each F2 cross. For graphical presentation, see Appendix II.

Chr 1 FW MW IW BMI WG EWG FWG Chr 11 FW MW IW BMI WG EWG FWG FW * FW * MW 0.89 * MW 0.89 * IW 0.29 0.27 * IW 0.56 0.64 * BMI 0.94 0.83 0.29 * BMI 0.96 0.84 0.52 * WG 0.90 0.79 -0.14 0.84 * WG 0.96 0.80 0.31 0.92 * EWG 0.66 0.79 -0.37 0.60 0.86 * EWG 0.80 0.88 0.22 0.77 0.85 * FWG 0.80 0.44 0.26 0.75 0.73 0.27 * FWG 0.86 0.52 0.30 0.83 0.88 0.49 *

Chr 2 FW MW IW BMI WG EWG FWG Chr 12 FW MW IW BMI WG EWG FWG FW * FW * MW 0.93 * MW 0.90 * IW 0.46 0.54 * IW 0.48 0.49 * BMI 0.93 0.84 0.30 * BMI 0.96 0.82 0.43 * WG 0.94 0.82 0.12 0.92 * WG 0.93 0.82 0.13 0.90 * EWG 0.81 0.84 -0.01 0.80 0.90 * EWG 0.72 0.82 -0.09 0.66 0.85 * FWG 0.87 0.63 0.24 0.85 0.88 0.60 * FWG 0.84 0.53 0.32 0.86 0.82 0.40 *

Chr 3 FW MW IW BMI WG EWG FWG Chr 13 FW MW IW BMI WG EWG FWG FW * FW * MW 0.82 * MW 0.91 * IW 0.11 0.31 * IW 0.29 0.44 * BMI 0.92 0.69 0.08 * BMI 0.92 0.80 0.21 * WG 0.91 0.67 -0.30 0.85 * WG 0.75 0.56 -0.42 0.73 * EWG 0.67 0.69 -0.47 0.58 0.84 * EWG 0.53 0.46 -0.59 0.50 0.92 * FWG 0.89 0.47 -0.07 0.86 0.88 0.48 * FWG 0.84 0.55 0.00 0.83 0.79 0.49 *

Chr 4 FW MW IW BMI WG EWG FWG Chr 14 FW MW IW BMI WG EWG FWG FW * FW * MW 0.87 * MW 0.92 * IW 0.11 0.16 * IW 0.46 0.58 * BMI 0.82 0.61 -0.12 * BMI 0.97 0.86 0.38 * WG 0.76 0.63 -0.55 0.77 * WG 0.95 0.82 0.17 0.95 * EWG 0.50 0.57 -0.71 0.52 0.89 * EWG 0.85 0.89 0.15 0.82 0.90 * FWG 0.79 0.39 0.02 0.78 0.64 0.23 * FWG 0.78 0.47 0.16 0.81 0.82 0.48 *

Chr 5 FW MW IW BMI WG EWG FWG Chr 15 FW MW IW BMI WG EWG FWG FW * FW * MW 0.90 * MW 0.81 * IW 0.33 0.42 * IW 0.29 0.22 * BMI 0.93 0.81 0.36 * BMI 0.90 0.65 0.20 * WG 0.83 0.67 -0.25 0.75 * WG 0.88 0.72 -0.19 0.84 * EWG 0.60 0.62 -0.45 0.50 0.88 * EWG 0.47 0.65 -0.57 0.40 0.75 * FWG 0.81 0.47 0.11 0.79 0.76 0.37 * FWG 0.83 0.36 0.28 0.83 0.72 0.09 *

Chr 6 FW MW IW BMI WG EWG FWG Chr 16 FW MW IW BMI WG EWG FWG FW * FW * MW 0.87 * MW 0.92 * IW 0.47 0.54 * IW 0.54 0.51 * BMI 0.88 0.72 0.38 * BMI 0.96 0.85 0.52 * WG 0.92 0.74 0.09 0.82 * WG 0.92 0.84 0.19 0.89 * EWG 0.66 0.76 -0.13 0.56 0.80 * EWG 0.72 0.82 -0.06 0.65 0.87 * FWG 0.85 0.48 0.26 0.78 0.85 0.36 * FWG 0.82 0.53 0.44 0.83 0.77 0.35 *

Chr 7 FW MW IW BMI WG EWG FWG Chr 17 FW MW IW BMI WG EWG FWG FW * FW * MW 0.84 * MW 0.88 * IW 0.32 0.33 * IW 0.43 0.40 * BMI 0.92 0.70 0.23 * BMI 0.96 0.82 0.42 * WG 0.85 0.68 -0.21 0.82 * WG 0.93 0.81 0.07 0.88 * EWG 0.52 0.67 -0.48 0.47 0.80 * EWG 0.67 0.82 -0.19 0.61 0.82 * FWG 0.81 0.35 0.20 0.81 0.73 0.17 * FWG 0.83 0.47 0.33 0.82 0.78 0.29 *

Chr 8 FW MW IW BMI WG EWG FWG Chr 18 FW MW IW BMI WG EWG FWG FW * FW * MW 0.89 * MW 0.90 * IW 0.31 0.37 * IW 0.37 0.46 * BMI 0.91 0.75 0.21 * BMI 0.96 0.86 0.40 * WG 0.92 0.76 -0.08 0.89 * WG 0.88 0.74 -0.11 0.82 * EWG 0.72 0.78 -0.27 0.68 0.87 * EWG 0.66 0.68 -0.31 0.59 0.87 * FWG 0.85 0.52 0.16 0.85 0.84 0.46 * FWG 0.85 0.56 0.16 0.83 0.83 0.45 *

Chr 9 FW MW IW BMI WG EWG FWG Chr 19 FW MW IW BMI WG EWG FWG FW * FW * MW 0.88 * MW 0.88 * IW 0.48 0.53 * IW 0.55 0.65 * BMI 0.92 0.80 0.45 * BMI 0.97 0.83 0.52 * WG 0.92 0.75 0.09 0.83 * WG 0.96 0.79 0.28 0.94 * EWG 0.72 0.81 -0.07 0.62 0.84 * EWG 0.83 0.91 0.29 0.78 0.85 * FWG 0.78 0.39 0.24 0.73 0.79 0.32 * FWG 0.75 0.36 0.17 0.77 0.81 0.38 *

Chr 10 FW MW IW BMI WG EWG FWG Chr X FW MW IW BMI WG EWG FWG FW * FW * MW 0.82 * MW 0.92 * IW 0.40 0.48 * IW 0.56 0.54 * BMI 0.95 0.75 0.31 * BMI 0.93 0.82 0.51 * WG 0.91 0.70 0.00 0.90 * WG 0.95 0.87 0.28 0.89 * EWG 0.63 0.76 -0.20 0.62 0.80 * EWG 0.76 0.86 0.04 0.67 0.86 * FWG 0.83 0.38 0.18 0.83 0.82 0.31 * FWG 0.86 0.60 0.45 0.85 0.83 0.44 *

4.165 0.005 1.182 0.859 4.337 0.006

to identify QTLs that meet QTLs that identify to

2.589 0.094 2.884 0.060 2.2322.167 0.191 0.321

273 2.277 0.161 0.667 0.980

3.507 0.025 6.803 0.0001

394000991 2.818 0.679 0.935 0.102 0.979 0.951 1.633 1.361 0.504 0.618 0.621 1.000 2.330 0.363 1.217 0.235 1.000 0.824 648 1.381 0.678 1.880 0.367 1.059 0.877 900796161 1.263273 2.280 1.113 0.790 2.003 0.220 0.884 2.039 0.364 1.784 0.728 0.311 1.876 0.467 0.991 1.076 0.434 1.688 1.938 0.892 2.873 0.527 0.371 0.087 996 0.523 0.999 0.308 1.000 1.053 0.876 784 1.008 0.860 1.758 0.374 0.388 1.000 716332 0.632 1.645 0.927 0.475 0.750 2.551 0.864 0.113 0.986 1.938 0.703 0.312 170434 2.320 1.014 0.114 0.488 1.371 0.633 0.514 0.814 1.964 1.004 0.210 0.496

OD Value P LOD Value P LOD Value P LOD Value P 3.918 0.008 4.605 0.002 4.885 0.0012 4.6605.031 0.0018 0.0014 4.2572.638 0.007 0.122 3.264 0.035 3.393 0.028 3.093 0.047 3.242 0.036

re corrected for the sevenyellowused Light traits tested. is used identifysignificant to (p<0.05). were QTLs that

0.710 0.983 2.400 0.224 1.812 0.388 1.945 0.277 1.986 0.260 1.953 0.

F2 genome scan with p values adjusted for traits analyzed. The F2 crosses were considered 20 independent whole whole 20 independent considered were F2 crosses The analyzed. traits for adjusted values p with scan genome F2 A 2.270 0.164 0.721 0.969 3.477 0.019

FW MW IW BMI WG EWG FWG 0.8440.2820.569 0.986 1.000 0.999 0.653 0.484 0.486 0.998 0.998 1.000 1.905 2.148 0.307 0.442 0.200 1.000 1.987 0.311 0.726 0. 1. 0. 1.612 0.524 2.282 0.197 1.695 0.472 1.426 0. 0.6981.4602.308 0.9922.280 0.679 0.209 0.842 0.240 1.036 1.385 0.971 1.749 0.922 0.723 1.206 0.511 1.083 2.467 0.825 0.996 0.902 0.160 1.060 0.947 1.280 2.462 0. 2.197 0. 0. 0. 0.511 0.999 0.405 0.999 0.953 0.922 0.608 0. 1.005 0.862 1.499 0.535 1.897 0.304 1.134 0. 0.9841.995 0.705 0.287 1.041 2.407 0.663 0.146 1.118 0.778 0.607 0.970 0.969 1.898 0. 0. 2.7212.208 0.074 1.084 0.139 0.436 1.815 0.934 0.269 0.554 0.887 0.270 0.850 0.995 2.089 1.087 0. 0.

7 8 9 12 3 4 5 4.4656 0.003 3.568 3.785 0.016 0.018 3.293 0.046 X 12 11 4.025 0.006 10 3.214 0.043 5.827 0.0003 3.166 0.047 14 13 4.139 0.008 4.874 0.0020 15 16 18 17 19 the suggestivecriteria (p<0.63), andyellow bright bold and text is genome scans and the peak p values listed for each chromosome we chromosome each for listed values p peak the and scans genome Table II-10. LOD scoresB6-Chr for Chromosome LOD Value P LOD P Value LOD P Value L 98

We also analyzed the crosses as a single genome scan and corrected the p values

for our genome-wide search (the 20 crosses) and the seven traits analyzed (Table II-11).

Obviously, with the increased multiple testing penalties, fewer QTLs were detected

because the weaker QTLs failed to reach significance. This more conservative method

detected significant QTLs on chromosomes 1 (BMI, WG), 6 (BMI), 10 (MW, EWG), 11

(WG), and 13 (MW) and suggestive QTLs were on 1 (FW, MW), 6 (FW, MW, WG,

FWG), 10 (FW, IW, WG), 11 (FW, BMI, FWG), 13 (FW, BMI, WG, FWG), and 17

(MW) (Tables II-11 and 12; Figure II-5). This CSS whole genome scan is a novel

approach and the appropriate statistical methods have not been established in the

literature. Thus, although the remainder of this chapter will focus on the QTLs detected

with the more conservative approach, the QTLs detected using fewer multiple testing

corrections should not be disregarded as these QTLs may be true QTLs.

To determine whether multiple QTLs may be present on a single chromosome,

the location of the peak LOD scores for various traits on each chromosome were

compared (Table II-12). On chromosomes 6 and 10, for example, the locations of the

peaks for the various traits suggested that multiple QTLs may be present on these

chromosomes. For chromosome 6, the 1.5 LOD intervals (an estimation of the 95%

confidence interval) for several traits did not overlap and thus, provided evidence for two

QTLs on the chromosome. Furthermore, a closer examination of the LOD scores for FW and WG on chromosome 6 indicated that two QTLs with non-overlapping 1.5 LOD support intervals influenced these traits. For example, the slightly stronger FW QTL on chromosome 6 is near 57 cM (LOD=3.78, 1.5 LOD support interval: 38-76 cM), but a second QTL near 28 cM (LOD=3.74, 1.5 LOD support interval: 21-33 cM) was also 99

ven traits tested. ven tested. traits 1.182 1.000

p<0.05). p<0.05).

2.884 0.723 2.232 0.988 6.803 0.001 2.589 0.866 4.165 0.091

4.257 0.109 2.167 0.998 4.337 0.094 4.605 0.041

4.885 0.024 4.660 0.036 5.031 0.023

ificant QTLs are highlightedQTLs are bright yellow in ificant bold and text ( lues for each chromosome were corrected for the 20 crosses and se 20 crosses for the corrected were each chromosome for lues

3.166 0.583 2.638 0.906 3.507 0.369 1.812 1.000 3.264 0.535 3.393 0.456 3.093 0.648 3.242 0.548

F2 genome scan with p values adjusted for number of crosses and number of traits analyzed. The F2 crosses F2 crosses The analyzed. of traits number and of crosses number for adjusted p values with scan genome F2 A 5.827 0.006 4.874 0.042

FW MW IW BMI WG EWG FWG 4.465 0.054 3.568 0.292 0.710 1.000 0.6981.460 1.0002.308 1.0002.280 0.842 0.9913.785 1.036 0.9790.844 1.000 1.385 0.2590.282 1.000 1.749 1.0000.569 1.206 1.000 3.293 1.0003.214 1.083 1.000 0.653 1.000 1.000 2.467 0.539 0.484 0.550 1.000 0.996 1.000 0.486 1.060 0.971 2.400 1.000 1.280 1.000 1.905 1.000 1.000 2.462 0.984 2.148 1.000 2.197 1.000 0.307 1.263 0.971 0.995 2.280 0.988 1.987 1.000 1.000 1.113 0.311 0.990 2.003 1.000 0.726 2.039 1.000 1.000 1.784 0.998 2.818 1.000 0.996 0.728 0.679 1.000 1.876 0.840 0.935 1.076 1.000 1.000 1.688 1.000 1.633 1.000 1.000 1.938 1.361 1.000 2.873 1.000 0.504 0.999 1.000 0.718 2.330 1.000 0.363 0.989 1.217 1.000 1.000 4.0251.612 0.1144.139 1.000 2.270 0.151 2.282 0.974 0.987 0.721 1.695 1.000 1.000 3.918 1.426 0.143 1.000 1.381 1.000 1.880 1.000 1.059 1.000 0.5111.005 1.0000.984 1.0002.721 0.405 1.0001.995 1.499 0.8602.208 1.000 1.041 0.9971.084 1.000 3.477 0.993 0.953 1.000 2.407 1.000 1.897 0.382 1.815 1.000 1.118 0.935 0.934 1.000 1.945 1.000 0.608 1.000 0.778 1.000 1.134 1.000 0.887 1.000 0.969 1.000 0.270 1.000 1.986 1.000 0.523 1.000 1.898 1.000 1.008 1.000 2.089 1.000 0.632 0.999 1.087 1.000 1.953 0.998 0.308 1.000 1.645 1.000 1.758 1.000 2.320 1.000 0.750 1.000 1.014 1.000 2.277 0.982 1.053 1.000 2.551 1.000 0.388 0.989 1.371 1.000 0.986 0.875 0.633 1.000 0.667 1.000 1.000 1.938 1.000 1.000 1.964 0.999 1.004 1.000 1.000

1 2 3 4 5 6 7 8 9 X 10 11 12 13 14 15 16 17 18 19 Chromosome LOD P Value LOD P Value LOD P Value LOD P Value LOD P Value LOD P Value LOD P Value were considered a single whole genome scans, and the peak p va p peak the and a single whole genome scans, were considered sign and (p<0.63), yellow light in highlighted are QTLs Suggestive Table II-11. LOD scores from B6-Chr from scores LOD Table II-11.

100

Table II-12. Support intervals for significant and suggestive QTLs. QTLs with overlapping support intervals were given the same name. For a single chromosome, we only called two peaks independent QTLs if the 1.5 LOD support intervals were completely non-overlapping. A. Significant QTLs from the intercrosses are listed with the 1.5 LOD support interval, an estimate of the 95% confidence interval. B. Suggestive QTLs from the intercrosses are listed with the 1.5 LOD support interval.

Chromosome (Trait) QTL name Location (cM) 1.5 LOD Support Interval (cM) 1 (BMI) Obrq4 67 57 - 69 1 (WG) Obrq4 60 47 - 69 6 (BMI) Obrq1 56 51 - 62 10 (MW) Obrq5 65 57 - 90 10 (EWG) Obrq5 95 79 - 103 11 (WG) Obrq6 49 38 - 55 13 (MW) Obrq7 24 12 - 31

Chromosome (Trait) QTL name Location (cM) 1.5 LOD Support Interval (cM) 1 (FW) Obrq4 62 47 - 69 1 (MW) Obrq4 65 37 - 69 6 (FW) Obrq1 57 37 - 76 Obrq3 28 20 - 33 6 (MW) Obrq1 56 39 - 64 6 (WG) Obrq3 28 19 - 33 Obrq1 43 36 - 73 6 (FWG) Obrq3 28 18 - 33 10 (FW) Obrq5 62 25 - 103 10 (IW) Obrq5 65 56 - 80 10 (WG) Obrq5 90 68 - 103 11 (FW) Obrq6 49 31 - 57 11 (BMI) Obrq6 47 32 - 57 11 (FWG) Obrq6 49 40 - 58 13 (FW) Obrq7 38 8 - 53 13 (BMI) Obrq7 15 0 - 51 13 (WG) Obrq7 23 0 - 47 13 (FWG) Obrq7 35 6 - 51 17 (MW) Obrq8 00 - 11

101

Figure II-5. B6-ChrA CSS whole genome scan. The LOD scores for each cross are plotted. 0.05 (-----) and 0.63 ( ) statistical (p value) thresholds are indicated. These thresholds reflect correction for the seven traits and 20 crosses (“conservative statistical approach”).

102

detected. Likewise, although the QTLs on chromosome 10 had overlapping 1.5 LOD support intervals, the localization suggests that the peaks may also represent distinct

QTLs. Consequently, the actual number of QTLs detected in the crosses is probably

under-estimated because at least two chromosomes may have multiple QTLs (Table II-

12).

5. Obesity resistance and obesity promoting QTLs detected in F2 crosses

We expected that the QTLs detected in the crosses would be A/J-derived obesity

resistance QTLs because no A/J-derived obesity promoting QTLs were detected in the

CSS surveys. To test this hypothesis, the F2 mice were sorted by their genotype at the

marker nearest each significant or suggestive peak, and the mean FW for each genotype

at that marker was calculated. Surprisingly, A/J-derived resistance QTLs (FW) were

detected on only chromosomes 6, 10, and 13, and A/J-derived obesity promoting QTLs

(FW) were detected on chromosomes 1 and 11 (Table II-13). Therefore, using the

intercrosses, we detected two obesity-promoting QTLs that were not detected in the surveys. A graphical summary of the intercross results is presented in figure II-6.

6. Resistance QTLs on 10 A/J chromosomes were not detected in the crosses

In the replicate studies, 13 CSSs reproducibly conferred resistance, but in the

crosses, significant or suggestive QTLs were only detected on three of these 13 A/J

chromosomes. We hypothesized that the genetics of obesity resistance on the

chromosomes on which we expected but did not detect QTLs may be complex involving

either multiple QTLs or parental effects. To test whether the F2 crosses provided 103

Table II-13. QTL inheritance patterns. The marker nearest the peak LOD score for each significant or suggestive QTL was identified. The F2 mice were sorted based on their genotype at the peak marker, and the mean body weight at each genotype was calculated. Abbreviations: B/B = homozygous C57BL/6J alleles, A/A = homozygous A/J alleles, B/A = heterozygous alleles

Chromosome Marker Trait B/B A/A B/A 1 m7 FW 39.27 43.39 44.50 1 m8 MW 31.94 35.73 35.28 1 m8 BMI 0.34 0.39 0.39 1 m7 WG 0.20 0.24 0.26 6 m18 FW 37.55 35.56 39.69 6 m18 MW 30.97 30.71 32.74 6 m18 BMI 0.35 0.33 0.37 6 m9 FW 37.04 35.92 39.77 6 m9 WG 0.19 0.17 0.21 6 m9 FWG 0.14 0.12 0.17 10 m9 FW 40.77 34.52 37.11 10 m10 MW 33.45 28.92 31.29 10 m10 IW 20.46 18.50 20.38 10 m11 WG 0.20 0.16 0.16 10 m11 EWG 0.23 0.19 0.18 11 m7 FW 36.58 44.03 40.78 11 m7 BMI 0.32 0.38 0.36 11 m7 WG 0.17 0.24 0.21 11 m7 FWG 0.13 0.23 0.18 13 m9 FW 38.05 31.82 34.65 13 m6 MW 30.64 27.37 30.59 13 m4 BMI 0.33 0.30 0.33 13 m6 WG 0.18 0.13 0.17 13 m9 FWG 0.15 0.09 0.11 17 m1 MW 31.31 30.50 33.52

104

e . For n Obrq5 19 Genome Scan Significant QTL (p<0.05) Suggestive QTL (p<0.63) 1.5 LOD Support (20 scans) 1.5 LOD Support (1 scan) Obrq Obrq 16 Obrq3 Obrq1 of chromosomes and numberphenotypes of presented tested) are ly ly for number of phenotypes tested) and single whole genome sca X 15 FW QTL or the BMI QTL if no FW QTL was detected. detected. was QTL no FW if QTL or the BMI QTL FW 14 Obrq7 12 QTL detected No QTL detected CSS Surveys Obrq6 11 13 17 18 Obrq4 Figure II-6. Comparison of CSS surveys and whole genome scan analysis. The FW and/or BMI QTLs detected using both the 20 whol 20 the both using detected QTLs BMI and/or FW The analysis. scan genome whole and surveys CSS of Comparison II-6. Figure genomewith scan method multiple (liberal testing corrections on (conservative method with multiple testing corrections for number this graphic, the LOD supportthis presented graphic,for the the is 105

evidence for such complexity, we investigated FW in each F2 population by comparing

the FW from the F2 progeny to that of C57BL/6J. If no QTLs were present, we

hypothesized that the mean FW from the crosses would be similar to C57BL/6J.

Interestingly, the mean FW from the F2 males derived from B6-Chr 3A, 4A, 5A,

6A, 7A, 8A, 10A, 12A, 13A, 15A, 17A, 18A, and XA was significantly lower (all with

p<0.007) than the FW from C57BL/6J indicating that these 13 F2 populations were

obesity resistant relative to C57BL/6J. In contrast, the F2 progeny derived from B6-Chr

14A and B6-Chr 16A were not significantly different from C57BL/6J. Thus, although

QTLs were not detected in the F2 crosses derived from chromosomes 3, 4, 5, 7, 8, 12, and 17, these chromosomes must harbor QTLs because, if not, the FW in the F2 progeny

would not be significantly different from C57BL/6J (Figure II-7).

Similarly, we hypothesized that if QTLs were present on the A/J chromosomes on which QTLs were expected but not detected, the variance in the crosses would also be larger than C57BL/6J. Alternatively, if no QTLs were present on these chromosomes, we expected the variance in the crosses would be similar to C57BL/6J. Surprisingly, the variance in FW from all intercross populations, even intercrosses in which QTLs were detected, did not significantly differ from C57BL/6J. To further investigate this result, we used a larger C57BL/6J sample (all C57BL/6J male mice used in this chapter as a pooled sample) with the hypothesis that a larger sample size may increase our power to detect differences in variance. For this analysis, we only used the 13 strains in which resistance QTLs were expected. Of 13 intercrosses analyzed, only one, the intercross

106

5) after

from each F2 cross is indicated by a bar, and each and by a bar, is indicated F2 cross from each * ** CSS * * * QTL detected not Obese CSSs detected QTL * * * CSS F2 progeny vs. C57BL/6J (B6). The mean FW The (B6). vs. C57BL/6J CSS F2 progeny A * * ** Lean CSSs QTL detected QTL not detected not QTL * * 3 3 410126155188X1711142169191B6 137

60 50 40 30 20 FW (grams) FW individual is indicated by a circle. * indicates that the FW for the F2 cross was significantly different from C57BL/6J (p<0.0 C57BL/6J from different significantly was cross F2 the for FW the that * indicates circle. a by indicated is individual multiple testing correction. Figure II-7. FW for B6-Chr

107 derived from B6-Chr 4A, had a variance that differed significantly from C57BL/6J, and the variance in this intercross was actually lower than C57BL/6J. Even crosses in which

QTLs were detected did not have a larger variance than C57BL/6J. Despite the unexpected variance results, the FW from the F2 progeny together with the replicate CSS surveys provide strong evidence that resistance QTLs are present but undetectable in the

F2 crosses.

7. Duration of diet exposure reveals time of QTL action

Because of the lack of a strong relationship between EWG and FWG in the F2 crosses, we hypothesized that some QTLs may produce effects at early vs. later stages of the time course. To test this hypothesis, we compared the F2 mapping results for traits at various time points on the same chromosome. The results demonstrated that the effects of certain QTLs are stronger at specific time points. For example, the QTL on chromosome 11 influenced weight gain in the second half of the study because significant or suggestive QTLs were detected for BMI, FW, WG and FWG, but not EWG or MW. In contrast, the chromosome 10 QTL influenced weight gain in the first half of the study because LOD scores for EWG and MW were the strongest among the traits and the LOD score for FWG was extremely low. Chromosome 6 was also complex with at least two QTLs. FWG and WG QTLs were detected near 28 cM on chromosome 6.

Because no EWG QTL was detected, this QTL most likely influences weight gain during the second half of the study. In contrast, the second QTL near 56 cM may influence weight gain throughout the study because FW and MW QTLs were detected. 108

Consequently, duration of diet exposure or age influences the genetics of obesity

resistance in our model.

8. Obesity resistance is not confirmed in B6-Chr MitoA

In the initial CSS survey, B6-Chr MitoA exhibited resistance to high-fat, diet-

induced obesity, but in the replicate CSS survey, this strain was not significantly different

from C57BL/6J. An intercross could not be used to investigate this chromosome because

the mitochondrial genome does not recombine. Instead, because the mitochondrial

genome is maternally inherited, we utilized a reciprocal cross strategy as an independent test to investigate whether the A/J-derived mitochondria confer resistance. If the A/J- derived mitochondria conferred resistance, we hypothesized that male progeny derived from a B6-MitoA female x C57BL/6J male cross should be resistant to obesity whereas male progeny derived from a C57BL/6J male x B6-Chr MitoA female cross should be obese. Prior to initiating the reciprocal cross study, we confirmed that the CSS harbored

A/J-derived mitochondria by genotyping a SNP (mitochondrial) in CoIII, which encodes

a subunit of the mitochondrial cytochrome c oxidase enzyme. Analyses of FW in the

reciprocal crosses revealed that the FW was not significantly different (p=0.28) (Figure

II-8) between the progeny derived from the two crosses. Thus, the A/J-derived

mitochondria do not reproducibly confer resistance to diet-induced obesity.

109

Figure II-8. Maternal inheritance was not detected in B6-MitoA. Male offspring from reciprocal crosses derived from B6-MitoA (mito) and C57BL/6J (B6) were fed the high-fat diet and final body weight was analyzed. The FW was compared using an unpaired t-test (p=0.28).

50 (grams) 25 Final weight body

0 B6 x Mito Mito x B6 Cross -

110

D. DISCUSSION

1. C57BL/6J males are susceptible to high-fat, diet-induced obesity

Genetic factors that increase susceptibility to diet-induced obesity are suspected to

be one cause for the rapidly rising prevalence of human obesity. Animal models, such as

C57BL/6J and A/J inbred mouse strains, are useful models for investigating the genetics

of diet-induced obesity. C57BL/6J males develop high-fat, diet-induced obesity and

metabolic syndrome whereas A/J males remain resistant (COLLINS et al. 2004; SURWIT et al. 1988). Investigations of C57BL/6J and A/J males on diets of varied fat and carbohydrate composition demonstrated that regardless of diet composition, C57BL/6J males are heavier than A/J males, but that this difference is enhanced on the high-fat diet.

Thus, C57BL/6J males have a genetic predisposition for both high-fat, diet-induced obesity and elevated body weight independent of the diet consumed. Studies using these strains must dissect the genetics of these two traits because different genes may explain them. For instance, variation in growth or lean body mass may explain the diet- independent weight differences whereas variation in energy metabolism may explain the weight differences that are diet-dependent.

2. Individual chromosome substitution confers obesity resistance in B6-ChrA CSSs

Despite the extensive investigations of the physiologic differences between

C57BL/6J and A/J inbred mouse strains fed the HFSC diet (BLACK et al. 1998; COLLINS et al. 2004; REBUFFE-SCRIVE et al. 1993; SURWIT et al. 1995; SURWIT et al. 1988), few

genetic studies of body weight have been performed using these strains. HFSC diet

surveys using the B6-ChrA CSSs provided the first estimate of the genetic complexity of 111

high-fat, diet-induced obesity in the C57BL/6J and A/J strains (SINGER et al. 2004). The surveys were surprising for several reasons. First, the substitution of individual A/J chromosomes was sufficient to confer obesity resistance on the C57BL/6J background despite the presence of many obesity-promoting QTLs and the high-fat content of the diet. Second, the obesity resistance in A/J is highly complex because 13 A/J-derived chromosomes reproducibly conferred decreased FW and BMI. Thus, at least 13 genes contribute to the trait difference between C57BL/6J and A/J.

We hypothesized that the CSS F2 crosses would enable us to determine the number and location of the A/J-derived obesity resistance QTLs detected in the surveys.

Instead, despite the reproducibility of the surveys, in the crosses using the conservative multiple testing correction strategy, we were able to detect A/J-derived obesity resistance

QTLs on only three chromosomes that reproducibly conferred resistance in the surveys suggesting that additional strategies are needed to detect the QTLs on the remaining 10 chromosomes.

One possible explanation for inability to detect QTLs on many chromosomes on which we expected them is overly conservative multiple testing corrections. Our genome scan is a novel approach and the appropriate statistical methods have not been established for such an analysis. If the genome scan is treated as 20 independent crosses, many more

QTLs are detected. Further analyses of these resistant CSSs using congenic strains or larger intercross sample sizes may reveal whether the QTLs detected using fewer multiple testing penalties are true QTLs or false positive results.

Multiple QTLs and parental effects are two other possible explanations for our inability to detect all of the expected QTLs in the crosses. First, in the F2 crosses, we 112

detected evidence for multiple, strong QTLs on chromosomes 6 and 10, but it is possible

that we do not have the power to detect multiple QTLs on other chromosomes. For

instance, multiple, weak QTLs with additive effects may cumulatively confer obesity

resistance when an intact A/J chromosome is inherited, but the effects of each

independent QTL may be too weak to detect in the crosses. Likewise, interacting QTLs

may not be detectable in the crosses, especially if the QTLs have no independent effects.

Larger sample sizes should increase power to detect interacting QTLs. Obviously, the

sample size needed will depend on the spacing of the QTLs. QTL interactions in obesity

are discussed in a later section.

Alternatively, parental effects are another explanation for the unexpected

intercross results. For instance, the parents of the F2 progeny and CSSs differ, and we

were unable to distinguish the parent of origin of the alleles in the intercrosses. Thus, we

were unable to test for parental effects in our data. The existence of parental effects in

our model would not be surprising because several studies have demonstrated the

influence of parental and grand-parental effects, such as maternal milk content or

imprinting, on obesity-related traits in several different mouse models (DIAMENT and

WARDEN 2004; JARVIS et al. 2005; MANTEY et al. 2005; REIFSNYDER et al. 2000; YORK et al. 1997; YU et al. 2000).

3. Many A/J-derived QTLs are associated with high-fat, diet-induced obesity resistance

Because A/J males have a genetic predisposition for lower body weight relative to

C57BL/6J males in the presence and absence of a high-fat diet, we investigated whether

the QTLs detected in the CSS surveys were influencing high-fat, diet-induced obesity. 113

Several pieces of evidence suggest that many of these QTLs confer high-fat, diet-induced obesity resistance. First, the LFCC survey demonstrated that many of the A/J chromosomes that conferred resistance in the surveys conferred resistance to weight gain specifically on the high-fat diet. In addition, all of the strains that reproducibly conferred a decreased FW in the HFSC surveys also conferred a decreased BMI. Because the BMI is an estimate of adiposity, the decreased FW in these CSSs suggests that decreased adipose tissue explains the weight differences. Measures of body composition or percentage of body fat will definitively test whether these QTLs are associated with

decreased fat deposition.

4. Intercrosses reveal obesity-promoting QTLs in the CSSs

Because C57BL/6J and A/J were not selected for body weight or body size during

inbred strain development, we expected to detect both obesity resistance and obesity

promoting QTLs. In the CSS surveys, whole chromosome substitution did not confer a

significantly increased body weight relative to C57BL/6J in any CSS. In contrast, in the

intercrosses, we discovered two new obesity-promoting QTLs. An obesity promoting

QTL was detected on chromosome 1 and a second was detected on chromosome 11. The

obesity-promoting QTL on chromosome 11 was particularly surprising because the FW

in B6-Chr 11A was less than C57BL/6J in all three HFHS surveys. Thus, the A/J strain

has both obesity-promoting and obesity resistance QTLs, but an intercross analysis rather

than a CSS survey was necessary to uncover the obesity promoting QTLs.

5. Duration of diet exposure adds complexity to the genetics of obesity resistance 114

The CSS surveys and intercross studies demonstrated that the genetic factors that

influence body weight may produce effects after particular durations of diet exposure or

at particular ages. First, of all the traits analyzed in the HFSC CSS survey, IW was the least reproducible and least correlated among the replicate surveys. Despite this variation in IW, the FW in the replicate surveys was highly reproducible and highly correlated.

Consequently, many prenatal and postnatal environmental factors rather than genetic

factors might explain body weight variation prior to diet exposure. In contrast, following

exposure to the HFSC diet, genetic factors contribute to weight variation among the

strains.

Mapping analyses for traits at various time points during the diet studies also

suggested that age or duration of diet exposure influences the genetics of weight gain in

our model. For instance, in the intercrosses, we identified several QTLs that produced

effects after short durations of diet exposure or at early ages whereas others required

longer diet exposures or older age before exhibiting an effect. Several other studies have

described time-dependent or developmental stage-specific QTLs influencing body

weight, obesity, or size-related traits in the mouse (BROCKMANN et al. 2004; CHEVERUD et al. 1996; MORRIS et al. 1999; ROCHA et al. 2004). Perhaps, QTLs at different time points represent different stages of growth or weight gain and may represent distinct

physiologic processes. Alternatively, some QTLs may act sequentially such that one

QTL may require the presence and previous action of a second QTL to exhibit an effect.

Although many studies focus on the end stage of a disease process, such as an obese adult, studies of QTLs that act early in the disease progression may reveal genes that can be targeted to prevent weight gain early in the development of obesity. 115

6. Epistasis characterizes obesity resistance in A/J and in B6-ChrA CSSs

The CSS surveys demonstrate that epistasis or gene interactions must contribute

to the obesity resistance phenotype. For instance, if the QTLs detected in the CSS surveys

were completely additive, A/J males would be much leaner than they are (SINGER et al.

2004). Furthermore, if the 13 A/J chromosomes showed additive effects, the sum of the

weight differences conferred by each A/J resistance chromosome should be equivalent to

the total weight difference between C57BL/6J and A/J (SINGER et al. 2004). Obviously, the individual A/J chromosomes confer a much larger effect than would be predicted using this type of additive model (SINGER et al. 2004). Furthermore, the fact that an

obesity-promoting QTL was detected on chromosome 11 even though the FW of B6-Chr

11A was reproducibly less than C57BL/6J indicates that epistasis may be occurring.

Consequently, the CSS data proves that interactions between QTLs on different

chromosomes must occur. Moreover, the intercrosses introduce the possibility of

interactions even on individual chromosomes.

The existence of epistasis in our model is not unexpected because recent studies

indicate that epistasis plays an important role in the genetics of obesity and related traits,

but the high level of epistasis suggested by the number of A/J chromosomes that individually confer obesity resistance is surprising (BROCKMANN et al. 2000; CHEVERUD et al. 2001; CORVA et al. 2001; STYLIANOU et al. 2006; WARDEN et al. 2004; YI et al.

2004; YI et al. 2006). To investigate epistasis, strains with two or three A/J-derived

chromosomes could be generated using the CSSs. Likewise, crosses in which multiple

chromosomes segregate may also be useful. Furthermore, congenic strains derived from 116

CSSs will be useful tools for testing whether interactions on individual chromosomes contribute to obesity resistance in this model.

7. Monogenic and multigenic models: complementary resources for studies of resistance

The panel of B6-ChrA CSSs is a unique resource for studying obesity because

several CSSs are resistant to diet-induced obesity. The CSSs are complementary to the

many single gene knock-out and transgenic models of obesity resistance because the

CSSs model polygenic rather than monogenic resistance. Unlike the monogenic models

which involve large perturbations in genes encoding a wide variety of proteins ranging

from signaling molecules to proteins involved in lipid metabolism and mitochondrial

function, the discovery of the genes underlying the resistance in the CSSs may reveal a

new repertoire of genes that when subtly perturbed produce resistance. Consequently,

single gene knock-out and transgenic models provide a bottom-up approach whereas the

CSSs provide a unique top-down, systems approach for genetic and physiologic studies

of diet-induced obesity resistance.

8. Obesity resistance in CSSs models human obesity

Overall, these B6-ChrA CSS studies expose a new dimension to the genetics of

obesity because the CSS surveys demonstrate that obesity resistance is a genetically

complex trait. We detected at least 13 QTLs that contribute to obesity resistance in the

CSSs and at least eight of these QTLs are specific for high-fat, diet-induced obesity. We

localized at least three of these QTLs and identified two new obesity promoting QTLs

using F2 crosses. Furthermore, our studies of individual chromosomes demonstrated that 117 several chromosomes may be similar to a mini-genome because each chromosome may have several QTLs with possible interactions among them that produce resistance on the

C57BL/6J background. This high level of complexity in a single pair of inbred strains provides a potential explanation for why the genetics of obesity in human populations has proven so challenging for investigators. In human populations, the genetics of diet- induced obesity may be significantly more complex with larger numbers of QTLs, some of which confer resistance and others that confer susceptibility to obesity, and complex networks of interactions among them.

118

CHAPTER III: GENETIC DISSECTION OF OBESITY RESISTANCE ON A/J-DERIVED CHROMOSOME 6

The work described in this chapter was performed by the candidate. 119

A. INTRODUCTION

Typical forms of human obesity result from a combination of many genetic and

environmental factors, but genetic studies of human obesity generally do not control for

variation in environmental factors, such as diet composition, because they are difficult to

control in human populations. For example, although the recommended intake of dietary

fat is < 30% of total calories (KRAUSS et al. 2000), individual dietary fat intake varies

and may exceed these recommendations. Furthermore, self-reported food diaries may not

provide an accurate representation of the quantity or types of food consumed and,

therefore, are not necessarily reliable for use in research studies (HILL and DAVIES 2001).

Variation in food consumption and types of food consumed may confound genetic studies of obesity in humans and explain why obesity QTLs are challenging to detect in human populations. Consequently, because many environmental factors, such as diet composition, can be controlled in the laboratory, animal models such as the C57BL/6J and A/J inbred mouse strains, which differ in body weight when fed a high-fat diet,

(SURWIT et al. 1988), provide an important resource for investigating the genetics of diet-

induced obesity.

Although many environmental factors can be controlled, obesity QTL discovery

in animal models remains challenging because many genes contribute to the trait even in

inbred strains, such as C57BL/6J and A/J. For instance, analyses of the B6-ChrA CSSs discussed in the previous chapter demonstrated that at least 13 A/J chromosomes confer decreased body weight on a high-fat diet. Furthermore, the CSS F2 crosses provided evidence for genetic complexity even on individual chromosomes. Thus, many genes contribute to the genetics of obesity resistance in A/J males and the identification of 120

individual QTLs is probably difficult to dissect in the A/J strain. In contrast, individual

CSSs simplify the analysis because they enable analyses of individual A/J chromosomes

and their contribution to obesity resistance. Because at least 13 A/J chromosomes

reproducibly confer resistance to high-fat, diet-induced obesity, the CSS panel provides at least 13 independent models for genetic and physiologic studies of the trait.

B6-Chr 6A is an example of a CSS that is resistant to high-fat, diet-induced

obesity. When fed the high-fat diet, B6-Chr 6A males weighed approximately 9-10 grams

less than C57BL/6J in each survey, a value that represents ~20% of the C57BL/6J body

weight. B6-Chr 6A males also consistently exhibited a significantly lower BMI, an

estimate of adiposity, relative to C57BL/6J males. The decreased final body weight and

BMI in B6-Chr 6A demonstrates that at least one obesity resistance QTL is present on the

A/J-derived chromosome 6.

A single gene or multiple genes on chromosome 6 may explain the resistance in

B6-Chr 6A. Results from the B6-Chr A CSS F2 crosses demonstrated that chromosome 6

has at least two resistance QTLs. Instead of detecting a single, well-defined LOD peak as

would be expected if a single QTL was segregating in the F2 population, we discovered a

broad LOD peak for FW that spans much of the chromosome and that consists of

multiple individual peaks. For the two highest peaks of the LOD curve, the 1.5 LOD

support intervals, which estimate the 95% confidence interval for a QTL, did not overlap, suggesting that these two peaks represent distinct QTLs. Thus, chromosome 6 probably has at least two QTLs associated with obesity resistance.

As presented in the previous chapter, a closer examination of the genetic markers nearest to these two potential QTLs on chromosome 6 indicated that mice with a 121

heterozygous genotype at the markers nearest the peaks were obese relative to mice that were homozygous for alleles derived from C57BL/6J or A/J (Table II-10). This phenomenon is called overdominance and suggests that an interaction may be occurring.

For instance, the mice that were homozygous for the C57BL/6J-derived alleles at either

marker were lean relative to C57BL/6J males. Obviously, the C57BL/6J males have

C57BL/6J-derived alleles at these markers but are not lean. Therefore, several QTLs on

chromosome 6 must interact because if they act independently, mice with C57BL/6J-

derived alleles would be expected to be obese like C57BL/6J males. Consequently, we

hypothesized that several genes influence obesity resistance on chromosome 6.

Mouse chromosome 6 is approximately 150 Mb (NCBI mouse genome build 35:

http://www.ncbi.nlm.nih.gov/genome/guide/mouse) with over 1800 genes, and the broad

LOD peak detected in the B6-Chr 6A F2 cross did not provide precise clues for QTL

localization. Therefore, the QTL(s) on chromosome 6 must be further localized before

candidate genes are identified and tested. Congenic strains, in which small segments of

a chromosome from one strain (donor) are substituted onto the genetic background of a

second strain (host), are useful for QTL localization (SILVER 1995; SNELL 1948). George

Snell pioneered the generation and utilization of congenic strains to dissect histocompatibilty genes in the mouse (SNELL 1948). Since then, congenic strains have

been used to study many traits. Congenic strains may provide increased power for QTL

detection relative to F2 crosses because they enable the investigation of genetically

identical individuals. Previous studies demonstrated that congenic strains are a useful

method for investigating QTLs detected in CSSs (YOUNGREN et al. 2003). 122

Historically, the generation of congenic strains derived from two inbred strains

using traditional methods required three to four years and 10+ mouse generations to

construct (SILVER 1995; SNELL 1948; WAKELAND et al. 1997). With the increased

availability of genetic markers, the process of generating congenic strains has been

reduced to approximately 1.5 years using the “speed congenic” strategy, but this method

of congenic strain generation requires extensive genotyping and is, therefore, time-

A consuming and expensive (MARKEL et al. 1997). With the generation of B6-Chr CSSs, the process of making congenic strains has been reduced to only three to four generations or approximately one year, and unlike the “speed congenic” strategy, genotyping is only required for a single chromosome. Furthermore, several panels of congenic strains have already been constructed using CSSs and are available for localizing QTLs

(http://genetics.case.edu/nadeaulab/).

In the present chapter, I describe a panel of congenic strains derived from B6-Chr

6A that was generated and screened to determine the number and location of obesity resistance QTLs on chromosome 6. Mapping studies using congenic strains localized one, independent obesity resistance QTL and provide evidence for two interacting QTLs that influence body weight. Additional marker genotyping refined these QTLs to regions

ranging from approximately 7 Mb to 24 Mb. Consequently, B6-Chr 6A is a polygenic

model for resistance to high-fat, diet-induced obesity, and congenic strains derived from

this CSS are useful tools for dissecting the genetics and physiology of resistance to high-

fat, diet-induced obesity.

123

B. MATERIALS AND METHODS

1. Mice: C57BL/6J and A/J mice were obtained from the Jackson Laboratory (Bar

Harbor, ME) and colonies were established at CWRU. The B6-Chr 6A CSS was generated

at CWRU (NADEAU et al. 2000; SINGER et al. 2004), and a colony was maintained at

CWRU. To generate heterosomic (B6-Chr 6A F1) mice, B6-Chr 6A females were crossed

to C57BL/6J males, and the male offspring were collected for phenotyping studies. All

mice were weaned at 3-4 weeks of age and maintained in microisolator cages with a 12

hour:12 hour light:dark cycle. All mice were fed LabDiet 5010 (LabDiet, Richmond, IN)

ad libitum until diet studies were initiated.

2. Diet studies: For all studies, the mice were introduced to the HFSC diet (D12331,

Research Diets, New Brunswick, NJ) at 35 days of age and fed ad libitum. Body weights were collected every two weeks for approximately 100 days. At the final time point, nasoanal lengths were measured so that BMI could be calculated. Diet composition is provided in Chapter 2 (Table II-1). For the analysis, FW and BMI were used.

3. Congenic strain panel construction: A panel of 17 reciprocal, overlapping, homozygous congenic strains derived from B6-Chr 6A was constructed (Figure III-1). F2 male and female mice (B6-Chr 6A x C57BL/6J) with various segments of A/J

chromosome 6 were selected by genotyping microsatellite markers. These F2 mice were

backcrossed to C57BL/6J, and offspring that were heterozygous for the selected region

124

). A ).

microsatellite markers and only the numberprovided. is

D6Mit and Mb http://www.ncbi.nlm.nih.gov/genome/guide/mouse/

55.2 75.5 85.3 92.7 104.6 112.2 125.4 139.0 146.5

. All markers used are A

AAAAAAAAAAAAA http://www.informatics.jax.org/ (

Genetic location(cM) 0.7 7.25 19 20.8 27.5 32.5 35.3 37.5 46 49 60.55 67 74 Physical location (Mb) 4.5 29.7 45.4 48.9 A 4ABBAAAAAAAAAAA 74-A BBBBAAAAAAAAA 46-A BBBBBBAAAAAAA 51-A BBBBBBBBBBBAA 62-A 2AABBBBBBBBBBBB 92-A AAAAABBBBBBBB 54-A 7BBBBBBBBAAAAAA 57-B 4BAAAAAAAAABBBB 54-B 2- BBBBBAAAAAAAA 120-A BBBBBBBBBBAAA 114-A 0- AAABBBBBBBBBB AAAABBBBBBBBB 108-A 109-A AAAAAABBBBBBB AAAAAAABBBBBB 115-A 105-A 1- AAAAAAAAAAABB 115-B 2B AABBBBBBBBBBB 62-BL Strain Marker 138 159 223 274 384 188 391 284 36 287 254 59 15 29ABBBBBBBBBBBBA 62.9-A BBBBBAAABBBBB 114.3-A B6-Chr 6

Figure III-1. Congenic panel derived from B6-Chr 6 B6-Chr from derived panel Congenic III-1. Figure denoteshomozygous A/J-derived alleles and B denotes homozygous B6-derived alleles. Marker locations are provided in cM in provided are locations Marker 125

(identified by genotyping) were intercrossed to homozygose the segment. The congenic

strains were maintained by brother-sister mating the progeny. Due to breeding

difficulties, 54-A was not used for the body weight screen.

4. Polymorphism selection: Informative microsatellite markers were selected using

NCBI (http://www.ncbi.nlm.nih.gov/genome/guide/mouse/) and the MGI

(http://www.informatics.jax.org/) websites. All marker locations refer to NCBI Build 35

(http://www.ncbi.nlm.nih.gov/genome/guide/mouse/) released in August 2005. To refine

the recombination breakpoints of the QTL critical regions, microsatellite markers and

SNPs were genotyped. SNP data was obtained from four sources: NCBI dbSNP

(http://www.ncbi.nlm.nih.gov/SNP/MouseSNP.cgi), SNPview at GNF

(http://snp.gnf.org/GNF10K/) (WILTSHIRE et al. 2003), The Inbred Laboratory Mouse

Haplotype Map developed at the Broad Institute of MIT and Harvard,

(http://www.broad.mit.edu/personal/claire/MouseHapMap/Inbred.htm), and the Celera

genome database (website no longer available). All markers used for congenic strain construction and refining the critical intervals of QTLs are listed in Table III-1.

5. Genotyping: Tail tissue was collected and digested with proteinase K (Invitrogen,

Carlsbad, CA) in 1X PCR buffer (Invitrogen, Carlsbad, CA) overnight at 55ºC. The enzyme was inactivated at 100ºC for one hour prior to using the DNA for genotyping studies. The reactions were performed in 1X PCR buffer, 2 mM MgCl2, 0.3mM of each

dNTP, 0.03 units of Taq polymerase/µL, and 0.4 - 0.5 µM of forward and reverse primers

in 25 µL total reaction volume. The reaction conditions were: 94°C for 2 minutes, 94°C 126

arker source source arker

A The Inbred LaboratoryMouse Haplotype Map G Map Haplotype Mouse Laboratory The Inbred A/J Allele Source A A G Map Haplotype Mouse Laboratory The Inbred

Allele C57BL/6J C57BL/6J

I α Taq Enzyme Enzyme for SNPs Restriction

the panel.the microsatellite markers B. SNPs and were

AAAGGTATGTG Mbo II G A Celera

s were obtainedwere from the mouses genome informatics website

CCCCCTTCGTACATCTGTGT I Xcm

AAAA CACCCATGGCATATTTCTGA AAATGA GTGGAAATGGACATGCTGAA HpyCH4 V C T The Inbred LaboratoryMouse Haplotype Map AGTGTCCCTAGGGGGTGG GGGGCCTTAGAGGTAGCAAC 139 bp 117 bp NCBI/MGI GCAATGCCAAAATGTTCAAA TCCTTCTCCATTTACACTTACAACA 115 bp 97bp NCBI/MGI ACCATCTGCATGGACTCACA GTTGAAGAGGACGACCAAGTG 196 bp 178 bp NCBI/MGI GGCTGCTGAGAAACAACCTC TGAGTATTGAGCCAAATCCTCC 145 bp 134 bp NCBI/MGI CACTGACCCTAGCACAGCAG TCCTGGCTTCCACAGGTACT 260 bp 195 bp NCBI/MGI

GCCATCCTTTGTAATAACAAACA CGTCTGGGAAAACCTCAAAA 168 bp 178 bp NCBI/MGI TTCTCTCAGTCTTGTCTGTGTACA GTGAGGCTCAAAGAAAGGGC 100 bp 110 bp NCBI/MGI CATATTCAAGACGGAGACTAGTTCC CACATGAAACACATGCACACA 116 bp 140 bp NCBI/MGI CTGTCTCAAAAAATAAAGTGACAAGC ATGCCCATATTGCATATCTGC 88 bp 150 bp NCBI/MGI GCTCTTATTAATGAAGAAGAAGGAGG CAAAGAAAGCATTTCAAGACTGC 111 bp 131 bp NCBI/MGI AATGCTTTATATGCAAACTACTCTCTC GAATATAGCAAGACAAGGGAGACA 126 bp 142 bp NCBI/MGI CTTTAGTCATTATTAGGATTGCCTATG TGGGATAGCATTGGAAACGT 130 bp 159 bp NCBI/MGI ACATTGGTAGTAGACTTGATATTTCCA TTACATGCTTGAGTGCTGGC 126 bp 102 bp NCBI/MGI

AAAGCCCCACTTGTCCAAC TAAACCTTGGGGTTTTGCAG I Bsm G

). A. Microsatellite markers were used Microsatellitewere construct markers to A. ).

4456709-4456823 (bp) 29683368-29683480 (bp) 45355206-45355328 (bp) 48686615-48686727 (bp) 55163236-55163360 (bp) 75461271-75461398 (bp) 85306769-85306869 (bp) 92694579-92694723 (bp) 104599315-104599510 (bp) 112164928-112165015 (bp) 125382949-125383088 (bp) 139021038-139021206 (bp) 146507305-146507557 (bp) 36546124 (bp)36546124 TCCCCTGCTCACTAATTCTTGT TGCAGCTGTCCTGCTTTAAC I Ale T C GNF/NCBI

16609642-16609765 (bp)16609642-16609765 GAGATTTCCCGTTACTATCTGACA AGCAATCCAGACAATGAGTTACA --- bp 124 bp 116 NCBI/MGI ) D6Mit59 D6Mit36 D6Mit287 D6Mit15 D6Mit15 D6Mit223 D6Mit159 D6Mit188 D6Mit138 D6Mit284 D6Mit254 D6Mit274 D6Mit391 D6MIt384

Marker Name Build 35 Location Forward Primer Reverse Primer C57BL/6J Allele Allele A/J Source http://www.informatics.jax.org/ are provided. All primer sequences for microsatellite marker ( Table III-1. Markers used for congenic panel construction. The marker names, locations, primer sequences, allele sizes, and m and sizes, allele sequences, primer locations, names, The marker construction. panel congenic for used Markers III-1. Table used to refine the recombination breakpoints in several strains. strains. several in breakpoints recombination the refine to used SNP #10SNP (bp) 10762148 AGCAGAGGGCTTTGTTGAAA GCCCCATTC SNP #23SNP #32SNP (bp) 23368662 (bp) 32233587 GCCACCGAATTCAACTTCTT CTCTGGCTACTGATGGCACA GGGGTTCTCTGGGTTATGGT GGGTCACACAGGGACTCATT I Fok I Rsa T A C G Celera Celera D6MIt264 available) Location 35 Build Primer Forward Primer Reverse 6-98704278 (bp) 98415121 CAGAGCGCACTAGC 6-95048665 (bp) 94772024 AAGAGCTCAGGCAACC 6-108104805 (bp) 107808913 GCTCATGCAAGGGGGTACT Marker Name Name Marker (ref SNP ID if ID SNP (ref 6-43395279 (43)6-43395279 (bp) 43483155 CTCACTCCTGCCAGTCACCT CATGGCTGCTTGTTACTCCA I Tsp45 C T Map Haplotype Mouse Laboratory The Inbred A. B. rs13478633 (12.2)rs13478633 (bp) 12191562 TGGTAGATTCAAGAGCACAGGA TCCCTATAAAGCACCCCTGAI Rsa C T NCBI 6-38783944 (38.7)6-38783944 (bp) 38785751 TGCATATCAGCCCACAGGTA TCACTTACGTGCATCATCCAT I EcoR A G Map Haplotype Mouse Laboratory The Inbred 6-35919902 (35.9)6-35919902 (bp) 35935144 GCTTCTGGTGTTTGCTCCTC GCACAACCTTGAACCCATTT I Gsu C T Map Haplotype Mouse Laboratory The Inbred 6-32961136 (32.9)6-32961136 (34.5)6-34541420 (bp) 33008024 (bp) 34570964 AGCCTGTCTACTGTGAAGC T TGTAACTCACGGCCTCTGTII Hpa T C Map Haplotype Mouse Laboratory The Inbred rs3024195 (90159 rs3024195 127

for 1 minute, 60°C for 1 minute, 72°C for 1 minute, repeat steps 2-4 34 times, 72°C for 5

minutes, and 4°C for 15 minutes. For microsatellite markers, PCR products were

separated using 6% polyacrylamide gel electrophoresis and visualized under ultraviolet

light with ethidium bromide (Gel Doc 2000, Bio-Rad, Hercules, CA). For SNPs, PCR

products were digested with the appropriate enzyme (Table III-1), and the products were

visualized using the same method.

6. Generation of the 62-BS congenic strain: Genotyping of additional markers revealed

that the A/J-derived segments in 62-B had uneven recombination breakpoints.

Consequently, we generated two sub-strains by intercrossing two 62-B mice and selecting

for mice with identical, homozygous A/J-derived segments. Mice with homozygous

segments were intercrossed, and two lines were developed (62-BS which contains the

short 62-B segment that extends from the centromere to a breakpoint between 32.2 Mb

and 36.6 Mb and 62-BL which contains the long 62-B segment that extends from the

centromere to a breakpoint between 38.7 Mb and 43.4 Mb).

7. 62-BL reciprocal crosses to investigate parental effects: F1 male progeny were

collected from reciprocal crosses (62-BL female x C57BL/6J male and C57BL/6J female x 62-BL male) and analyzed on the high-fat diet as previously described.

8. Statistical methods

For all studies, mice that lost > 10% of their body weight during a two-week period were not used for the analysis because such weight loss is rare and may indicate illness. 128

a. B6-Chr 6A x C57BL/6J F1 analysis: The FW and BMI of A/J, B6-Chr 6A, and B6-Chr

6A x C57BL/6J F1 male mice were compared to that of C57BL/6J using an unpaired t-test

(Graphpad Prism, version 3.0). If the variance differed significantly (based on an F test)

between the two strains, Welch’s t-test was used (Graphpad Prism, version 3.0).

b. Congenic strain analysis: The mean final body weight for each strain was compared

to C57BL/6J and B6-Chr 6A using an unpaired t-test (Graphpad Prism, version 3.0). If

the variance differed significantly (based on an F test), Welch’s t-test was used

(Graphpad Prism, version 3.0). Comparisons were made to both C57BL/6J and B6-Chr

6A because the C57BL/6J (unlike the B6-Chr 6A) sample was collected several months

after the congenic strain survey and because the C57BL/6J sample was characterized by a

large variance. For the comparison to C57BL/6J, we concluded that strains with a significant p value (<0.05) had a QTL in the A/J-substituted segment. For the comparison to B6-Chr 6A, we concluded that strains with a non-significant p value

(greater than 0.05) and those with a significant p value and a mean trait value less than

B6-Chr 6A had at least one QTL influencing the trait in the A/J-derived segment. In addition to the p=0.05 statistical threshold, the data set was analyzed using a Bonferroni corrected threshold of p<0.0028 which takes into account the 19 strains tested (0.05 / 18

= 0.0028). The Bonferroni correction is overly conservative because the congenic strains are obviously not independent strains. In addition, the trait values for all strains with an

A/J-derived allele at D6Mit36 (Obrq1) were pooled to calculate the mean trait values associated with Obrq1. The trait values for the pooled population were compared to 129

C57BL/6J using an unpaired t-test. Likewise, the trait values for all strains with Obrq3

(108-A, 109-A, 105-A, 115-A) were pooled and the mean trait value was compared to

62-BL. Mean trait values +/- one standard deviation are presented in the text.

c. 62-BS replicate analysis: The FW and BMI from 62-BS male mice were compared to the FW and BMI from the 62-BL replicate data using the same statistical tests as described above. The replicate 62-BL sample was selected because the 62-BS sample and replicate 62-BL sample were collected simultaneously.

d. 62-BL reciprocal crosses with C57BL/6J: The FW and BMI for the male offspring from the reciprocal crosses were compared to each other and to C57BL/6J using an unpaired t-test. If the variance differed significantly (based on an F test), Welch’s t-test was used. The reciprocal cross data were also pooled and the pooled trait values compared to the trait values from C57BL/6J and 62-BL using the same statistical tests.

C. RESULTS

1. B6-Chr 6A x C57BL/6J F1 male mice do not exhibit overdominance

In the intercross progeny derived from B6-Chr 6A, overdominance of the obesity

phenotype was observed at markers nearest the peak LOD scores. Thus, we hypothesized

that chromosome 6 may harbor a single or multiple, additive, overdominant QTLs. To

test this hypothesis, we examined FW and BMI in an F1 population derived from B6-Chr

6A and C57BL/6J. We expected that if chromosome 6 harbored a single, overdominant

QTL or multiple, additive, overdominant QTLs, the F1 population would exhibit 130

overdominance because the F1 males are heterozygous for all alleles on chromosome 6.

Contrary to expectation, the BMI in the F1 population was not significantly different

from C57BL/6J males (F1: 0.38 +/- 0.03 grams/cm2 vs. C57BL/6J: 0.39 +/- 0.04

grams/cm2; Welch corrected t=0.6757, df=48, p = 0.50), and the body weight in the F1

population was slightly less than C57BL/6J (F1: 43.89 +/- 4.53 grams vs. C57BL/6J:

46.56 +/- 5.85 grams; t=2.391, df=92, p = 0.02) (Figure III-2). Thus, the F1 males do not

exhibit overdominance, and the intercross results may reflect multiple, interacting QTLs.

2. Congenic analyses reveal at least three QTLs on chromosome 6

The complicated peak structure and overdominance in the intercross provided

evidence for the existence of multiple QTLs on chromosome 6. To fine map these QTLs,

a panel of overlapping congenic strains spanning chromosome 6 was generated, and the

congenic strains were surveyed for FW and BMI on the HFSC diet. Then, the regions of

chromosome 6 that were shared among the obesity resistant strains were identified as

likely locations for a resistance QTL.

Overall, eight congenic strains had a FW that was significantly less than

C57BL/6J (p<0.0028, Figure III-3). Of these eight strains, six shared a common A/J-

derived segment including D6Mit284 and D6Mit36. The only strain (46-A) with this A/J-

derived segment that did not significantly differ from C57BL/6J also weighed less than

C57BL/6J (p=0.01). Consequently, we concluded that an A/J-derived resistance QTL

must reside in this region. In contrast, the 114.3-A congenic strain, which has A/J- derived alleles at D6Mit284 but not D6Mit36 (Figure III-1), was obese, and hence, we could exclude D6Mimt284 from the candidate region. Thus, this QTL (named Obrq1)

131

Figure III-2. FW and BMI in the F1 [B6-Chr 6A (CSS-6) x C57BL/6J (B6)] male mice. Neither BMI nor FW in the F1 was significantly higher than C57BL/6J. Thus, unlike in the F2 cross overdominance was not observed. A. The mean FW and BMI for each strain are indicated with a bar. The values are derived from an unpaired t-test (or Welch’s t-test if variances differed) with comparison to C57BL/6J. B. The weight gain curve for C57BL/6J, A/J, B6-Chr 6A (CSS-6), and the F1 male mice was plotted using the mean body weights, which were measured at two-week intervals.

A. P=0.02

P<0.0001 75 P<0.0001

(grams) 25

Final weight body

0 B6 A/J CSS-6 F1 Strain

P<0.0001

P<0.0001 0.5

) 2 0.4

0.3

0.2

BMI (grams/cm 0.1

0.0 B6 A/J CSS-6 F1

Strain

132

Figure III-2 continued.

50 B6 F1

CSS-6 A/J 30

Body weight (grams) weight Body

B6 10 AJ CSS-6 F1 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 Age (days)

133

must reside in the A/J-derived segment extending between but not including D6Mit284

and D6Mit287, a region of approximately 20 Mb. The FW of all strains with Obrq1 (A/J-

derived alleles at D6Mit36) is 37.55 grams which is similar to the B6-Chr 6A mean of

37.66 grams and which differs significantly from C57BL/6J (B6 mean FW = 43.89 g, t=4.052, Welch’s corrected df=30, p=0.0003). BMI analyses produced similar results

(Figures III-1 and III-3, Table III-2 and III-3).

Evidence for additional QTLs was provided by two other congenic strains, 62.9-A and 62-BL, because both strains weighed significantly less than C57BL/6J. No other strains provided additional evidence for QTLs in these two A/J-derived segments, so replicate samples were collected and analyzed. For 62.9-A, a replicate sample was analyzed and neither the FW nor BMI in the replicate sample differed significantly from

C57BL/6J (Table III-4). Thus, the initial analysis of 62.9-A may have produced a false positive result.

In contrast, a replicate sample of 62-BL males provided confirmatory evidence for a second obesity resistance QTL (Obrq2) on chromosome 6 because the FW and BMI of the replicate were also significantly less (p<0.0001 for both traits) than C57BL/6J

(Table III-4, Figure III-4). Similarly, a third independent 62-BL sample also confirmed

Obrq2 and is discussed in the appendix. Because 92-A, a congenic strain that shares the centromeric region of 62-BL, is obese, the candidate interval for Obrq2 was localized to between but not including D6Mit138 and D6Mit223.

134

Figure III-3. FW and BMI in the HFSC diet congenic strain survey. The FW and BMI for C57BL/6J (B6), A/J, B6-Chr 6A (CSS-6), and each congenic strain are presented. The mean value for each strain is indicated with a bar. * indicates p<0.0028, the conservative Bonferroni threshold using an unpaired t-test or Welch’s t-test (if variance differed significantly using an F test).

Final bodyweight (grams)

20 ******** B6 CSS-6 62-BL 120-A 57-B 74-A 62.9-A 115-B 51-A 54-B 109-A 46-A 105-A 114.3-A 62-A 115-A 92-A 114-A 108-A Strain

0.500

0.475

0.450

0.425

0.400 2) 0.375

0.350

BMI (grams/cm 0.325

0.300 0.275

0.250

0.225

0.200 ****** B6 CSS-6 62-BL 120-A 57-B 74-A 54-B 115-B 62.9-A 51-A 109-A 46-A 105-A 114.3-A 115-A 62-A 92-A 114-A 108-A Strain

135

Obrq1

Obrq2

A) Unpaired t-tests (or 0.6940 0.7958 0.1141 0.0301 0.5960 0.2326 0.0001 0.5565

value P to B6-Chr 6 to B6-Chr (Comparison (Comparison

and the p values are presented. presented. p values are and the A df ) A 0.3950.2611.597 85 2.207 35* 0.532 82 1.202 84 83 85 3.978 89 0.590 84 esented for each strain. esented T statistic (Comparison (Comparison to B6-Chr 6 to B6-Chr

value P standard deviation is pr standard

B6) to (Comparison

T statistic

are bold. * indicates Welch’s t-test used. used. t-test Welch’s * indicates bold. are (Comparison to B6) (Comparison

Obrq2 and

(grams) Obrq1 Mean FW Mean

et congenic strain survey. The mean FW +/- one The mean FW strain survey. et congenic

FW FW n=60 4.05 37.66 +/- 3.858 34* 0.0005 ** ** ** sample size sample

62-A n=24 4.87 40.77 +/- 1.710 46* 0.0941 2.998 82 0.0036 92-A n=21 4.09 40.99 +/- 1.636 43* 0.1091 3.236 79 0.0018 74-A46-A51-A n=26 n=24 6.02 37.32 +/- n=25 4.49 39.27 +/- 4.03 38.17 +/- 3.344 2.592 3.309 53 44* 41* 0.0015 0.0129 0.0020 57-B n=27 3.99 36.54 +/- 4.301 41* 0.0001 54-B n=26 3.55 38.20 +/- 3.387 38* 0.0017 114-A n=25 4.95 42.31 +/- 0.865 46* 0.3917 4.511 83 <0.0001 108-A109-A115-A105-A n=32 n-26 n=27 5.47 42.54 +/- n=25 2.83 38.73 +/- 4.53 40.78 +/- 5.12 40.07 +/- 0.747 3.175 1.766 2.077 47* 35* 44* 47* 0.4590 0.0031 0.0843 0.0433 4.435 1.402 3.201 49* 2.305 66* <0.0001 85 83 0.1656 0.0019 0.0237 120-A n=26 4.40 35.51 +/- 4.780 43* <0.0001 115-B n=27 4.22 38.04 +/- 3.384 42* 0.0016 62-BL n=31 4.92 33.82 +/- 5.708 45* <0.0001 Strain 62.9-A n=25 5.48 37.57 +/- 3.362 49* 0.0015 0.075 35* 0.9406 114.3-A n=22 4.36 40.73 +/- 1.764 44* 0.0846 2.980 80 0.0038 C57BL/6J n=29 8.22 43.89 +/- ** ** ** ** ** ** B6-Chr 6 The strains that provide evidence for provide evidence that The strains Table III-2. FW in the HFSC di in the HFSC Table III-2. FW Welch’s t-test if variance was significantly different) were performed with comparisons to C57BL/6J (B6) and B6-Chr 6 B6-Chr and (B6) C57BL/6J to comparisons with performed were different) significantly was variance if t-test Welch’s 136

Obrq2 Obrq1 )

A strain. T-tests (or T-tests strain.

P value value P B6-Chr 6 B6-Chr

to (comparison df

) A

T statistic and the p values are presented. Strains expected to Strains expected are presented. and the p values A (Comparison (Comparison to B6-Chr 6 to B6-Chr

e standard deviation is presented for each for presented is deviation e standard

P value value P

(comparison to B6) (comparison

T statistic ngenic strain survey. The mean BMI +/- on to B6) (Comparison ) 2

Mean BMI (grams/cm

are bold. *indicates Welch’s t-test used. used. t-test Welch’s *indicates bold. are BMI BMI size n=39 0.35 +/- 0.03 3.631 38* 0.0008n=22 ** 0.37 +/- 0.04 ** 1.880 ** 47* 0.0663 2.129 59 0.0375 sample sample

Obrq2 A or A

92-A n=21 0.37 +/- 0.03 1.380 43* 0.174862-A n=24 0.37 +/- 0.03 3.168 1.896 58 0.0024 44* 0.0646 2.400 61 0.0195 74-A46-A n=2551-A n=23 +/- 0.04 0.34 +/- 0.04 0.36 n=25 +/- 0.03 0.35 3.564 2.551 2.998 49* 47* 41* 0.0008 0.0141 0.0046 0.526 1.144 62 0.840 60 0.6007 62 0.2571 0.4040 54-B n=26 +/- 0.03 0.35 57-B 3.461 n=27 +/- 0.03 0.33 41* 4.500 0.0013 39* 0.092 P<0.0001 63 0.9273 1.490 64 0.1412 108-A109-A115-A n=29105-A n=26 +/- 0.03 0.39 n=27 +/- 0.02 0.35 n=25 +/- 0.03 0.37 +/- 0.04 0.36 0.463 3.044 2.023 2.213 44*114-A 34* 42* 47* n=25 0.6458 0.0045 +/- 0.04 0.38 0.0495 0.0318 0.650 4.874 1.351 2.435 66 48* 1.738 62* 64 <0.0001 62 0.1815 0.5185 0.0177 0.0871 4.039 62 0.0002 120-A n=26 +/- 0.03 0.33 4.583 41* P<0.0001 1.723 63 0.0899 115-B n=25 +/- 0.03 0.35 3.340 41* 0.0018 0.311 62 0.7570 62-BL n=15 +/- 0.03 0.30 6.580 41* <0.0001 4.665 52 <0.0001 Strain 62.9-A n=25 0.35 +/- 0.04 2.836 49* 0.0066 0.545 62 0.5874 Obrq1 114.3- C57BL/6J n=29 0.39 +/- 0.06 ** ** ** ** ** ** B6-Chr 6 Welch’s t-test if variances differed) were performed with comparisons to C57BL/6J (B6) and B6-Chr 6 and B6-Chr (B6) C57BL/6J to comparisons with performed were differed) if variances t-test Welch’s Table III-3. Final BMI for the HFSC diet co BMI for the HFSC Table III-3. Final have have 137

Table III-4. 62.9-A and 62-BL congenic strain replicate analyses. The FW and BMI from the original and replicate samples for the two strains were compared using an unpaired t-test or Welch’s t-test if variances differed. The mean +/- the standard deviation is presented. * indicates Welch’s t-test used. Abbreviations: B6 = C57BL/6J.

Mean FW T statistic P value Strain Sample size df (grams) (Comparison to B6) comparison to B6 62.9-A original n=25 37.57 +/- 5.48 3.362 49* 0.0015 62.9-A replicate n=12 42.03 +/- 4.49 0.926 35* 0.3611

62-BL original n=31 33.82 +/- 4.92 5.708 45* <0.0001 62-BL replicate n=57 35.99 +/- 4.77 4.779 37* <0.0001

Mean BMI T statistic P value Strain Sample size df (grams/cm2) (Comparison to B6) comparison to B6 62.9-A original n=25 0.35 +/- 0.04 2.836 49* 0.0066 62.9-A replicate n=12 0.38 +/- 0.04 0.341 39 0.7352

62-BL original n=15 0.30 +/- 0.03 6.580 41* <0.0001 62-BL replicate n=34 0.33 +/- 0.04 4.952 44* <0.0001

138

Figure III-4. 62-BL congenic strain HFSC diet replicate analysis. A second, independent 62-BL sample was analyzed and confirmed the presence of an obesity resistance QTL in the 62-BL congenic region of A/J-derived chromosome 6. The mean trait value for each sample is indicated with a bar. The samples were compared using an unpaired t-test or Welch’s t-test if variances differed. Abbreviations: B6=C57BL/6J and CSS-6=B6-Chr 6A

20 P=0.0005 P<0.0001

Final body weight (grams) 10 P<0.0001 0 B6 CSS-6 62-BL original 62-BL replicate Strain

0.5

0.4 ) 2

0.3

0.2 P=0.0008 BMI (grams/cm P<0.0001 0.1 P<0.0001

0.0 B6 CSS-6 62-BL original 62-BL replicate Strain

139

Surprisingly, several strains with Obrq2 were obese. For instance, the FW and

BMI in three strains (108-A, 105-A, and 115-A) did not significantly differ from

C57BL/6J (using the Bonferroni corrected threshold) even though these strains have an

A/J-derived allele at Obrq2. Therefore, an interaction must be occurring between Obrq2

and a nearby QTL to produce obesity in these three strains (referred to as Obrq3). For

FW, only 109-A did not fit this interaction model because the FW was not significantly lower than C57BL/6J, but when the data from these four strains (108-A, 109-A, 115-A,

105-A) are pooled, the mean FW is 40.64 grams, a value that is significantly greater than

62-BL (t=6.957, df=139, p<0.0001). Similar results are obtained with comparisons to

B6-Chr 6A. Overall, the congenic panel provided evidence for one independent and two interacting QTLs associated with FW and BMI on the HFSC diet.

3. 62-BS confirms Obrq2 and refines the candidate interval

During congenic panel construction, we discovered that one of the homozygous

A/J-derived segments in 62-B was longer than the other as a result of uneven recombination breakpoints. Consequently, two substrains, 62-BL and 62-BS were generated, and 62-BL was used for the congenic survey. To test whether 62-BS also has

Obrq2, 62-BS males were analyzed on the HFSC diet. The FW and BMI of 62-BS males did not significantly differ from that of the replicate 62-BL sample (FW: t=0.6501, df=85, p=0.52; BMI: t=0.7176, df=59, p=0.72). Consequently, 62-BS provided additional evidence for Obrq2. Moreover, because the A/J-derived segment in 62-BS is shorter than 62-BL, the Obrq2 candidate region was narrowed, and the telomeric end of 140

the critical region was changed to 36.6 Mb. Overall, the critical region for Obrq2 extends from 4 Mb to 36.6 Mb and is approximately 32.6 Mb.

4. Additional markers refine critical intervals of resistance QTLs on chromosome 6

To refine the critical intervals for the three QTLs detected using the congenic strains, additional markers were genotyped in informative strains. The critical region of

Obrq1 is defined by the 54-B congenic strain, which contains the QTL and delineates the telomeric end of the critical region, and 114.3-A, which does not contain the QTL and defines the centromeric end of the critical region. To refine the interval, a SNP (6-

108104805) near 108 Mb was genotyped in 54-B, and a SNP near 94.8 Mb (6-95048665) was genotyped in 114.3-A. Although the telomeric end of the interval could not be further localized because 54-B had A/J-derived alleles at 6-108104805, the centromeric end was refined because 114.3-A had A/J-derived alleles at 6-95048665. Thus, the critical interval extends from 94.8 to 112.2 Mb, a region of approximately 17.5 Mb

(Figure III-5).

Likewise, the critical region of Obrq2 was refined using 92-A and 62-BS. 92-A, an obese congenic strain, defines the centromeric end of the critical region, and 62-BS defines the telomeric end of the region. To refine the critical interval, D6Mit264 (16.5

Mb) and two SNPS (SNP#10 at 10.7 Mb and rs13478633 at 12.2 Mb) were genotyped in

92-A. 92-A had homozygous A/J-derived alleles at both proximal markers but homozygous C57BL/6J-derived alleles at D6Mit264, the more distal marker. Likewise, two SNPs (6-34541420 at 34.5 Mb and 6-35919902 at 35.9 Mb) were genotyped in 62-

BS. 62-BS had homozygous A/J-derived alleles at 34.5 Mb and homozygous C57BL/6J- 141

D6Mit287 D6Mit287 (112.2 Mb) (112.2

Unknown (107.8 Mb) (107.8

6-108104805

(94.8 to 112.2 Mb)

D6Mit36 D6Mit36 Mb) (104.6 Obrq1

alleles B6-derived

is presented.

Obrq1 Obrq1 (98.4 Mb) (98.4 6-98704278

for interval Critical

(94.8 Mb) (94.8 A/J-derived alleles A/J-derived 6-95048665

D6Mit284 D6Mit284 Mb) (92.7

critical interval. The critical intervalfor

Obrq1

54-B (lean) 54-B

Figure III-5. III-5. Figure 114.3-A (obese) 114.3-A Chromosome 6 Chromosome 142

-derived alleles at 35.9 Mb. Consequently, the critical region extends from but does not include the SNPs at 12.2 Mb to 35.9 Mb (Figure III-6).

Lastly, the critical region for the suppressor of Obrq3 was refined by genotyping two additional SNPs (6-38783944 at 38.8 Mb and 6-43395279 at 43.5 Mb). 62-BL had homozygous A/J-derived alleles at the more proximal marker, but homozygous

C57BL/6J-derived alleles at the more distal marker (43.5 Mb). Thus, the critical interval extends from 38.8 Mb to 48.9 Mb, a region of only 10.1 Mb (Figure III-7).

The QTLs discovered on chromosome 6 are summarized in figure III-8.

5. Absence of parental effects for Obrq2

Because the critical interval of Obrq2 includes at least one known imprinted region (BEECHEY 2000), F1 males derived from reciprocal crosses were generated and analyzed to test whether a parental effect could be observed. If a parental effect such as imprinting explains the trait difference, we hypothesized that the FW from the progeny derived from the two reciprocal crosses would differ significantly from each other because the 62-BL segment would be inherited from a different parent in the progeny of each cross. For example, in the C57BL/6J x 62-BL cross, the 62-BL segment is inherited from the father but in the reciprocal cross, the same segment is inherited from the mother. Although males derived from both reciprocal crosses weighed less than

C57BL/6J, the FW of the reciprocal F1 male mice did not differ significantly from each other (Table III-5). Furthermore, males derived from both crosses were significantly heavier than 62-BL for both traits. Consequently, a parental effect does not explain the trait difference. 143

Unknown (43.5 Mb) 6-43395279

(38.8 Mb) 6-38783944 6-38783944

Mb) (35.9 6-35919902 6-35919902

(34.6 Mb) 6-34541420 6-34541420

alleles B6-derived

Obrq2 is presented. presented. is

Obrq2

(12.2 Mb to 35.9 Mb) Critical Interval for A/J-derived alleles D6Mit264 D6Mit264 (16.6 Mb)

critical interval. The critical intervalfor (12.2 Mb) (12.2 rs13478633 Obrq2 Obrq2

Chromosome 6 Chromosome Figure III-6. III-6. Figure 92-A (obese) 62-BS (lean) 62-BL (lean) 144

Unknown

D6Mit274 Mb) (48.9

D6Mit223 D6Mit223 (45.5 Mb) -derived alleles

B6 (38.8 Mb to 48.9 Mb)

43395279 43395279 (43.5 Mb) 6-

is presented. Obrq3

Obrq3

A/J-derived alleles Critical Intervalfor

38783944 (38.8 Mb) 6-

critical interval. The critical interval for

Obrq3 BL (lean) BL -

(obese) -A 62 108 Figure III-7. III-7. Figure Chromosome 6 145

Lean Obese

Obrq1 = = =

Obrq3

Obrq2 Obrq1 Obrq1 Obrq1

or or or +

Obrq3 + + Obrq3

Obrq2 Obrq3 Obrq2 Obrq2 Obrq1 Obrq2

Interacting QTL: Interacting Obesity Resistance QTLs: Resistance Obesity 1. 2. 3.

Figure III-8. Summary of QTLs discovered on chromosome 6. chromosome on discovered QTLs of Summary III-8. Figure 146

P value P value

(Comparison to B6) (Comparison ch reciprocal ch cross

(Welch corrected)

T statistic (Comparison to B6) (Comparison

62-BL x C57BL/6J x 62-BL n=26 0.03 +/- 0.36 0.768 44 0.4465

) 2

FW 62-BL x C57BL/6J n=28 6.33 +/- 40.73 ** ** ** BMI 62-BL x C57BL/6J n=28 0.04 +/- 0.36 ** ** ** Trait Strain size Sample value trait Mean (grams) C57BL/6J x 62-BL n=26 3.57 +/- 39.72 0.732 43 0.468 (grams/cm Table III-5. 62-BL reciprocal cross analysis. The mean +/- the standard deviation for FW and BMI from progeny derived from ea from derived progeny from BMI and FW for deviation standard the +/- mean The analysis. cross reciprocal 62-BL III-5. Table are presented. are presented. 147

Because the mean values were not significantly different, the FW and BMI from

the two crosses were pooled and compared to both C57BL/6J and 62-BL. For both traits,

the pooled F1 sample had a significantly higher FW than 62-BL (FW: t=4.505, df=109,

p<0.0001; BMI: t=4.182, df=86, p<0.0001). Similarly, for both traits, the FW was

slightly less than C57BL/6J (FW: t=02.170, df=40, p=0.04; BMI: t=2.584, df=37,

p=0.01). Because the body weight of the pooled F1 population is intermediate between

C57BL/6J and 62-BL, an additive effect may characterize the trait, but because the trait

values are closer to those of C57BL/6J, further investigations are needed to confirm this

observation.

D. DISCUSSION

1. Congenic strains successfully localize QTLs

We had previously tested whether QTLs detected in CSSs could be localized

using CSS F2 crosses. For some chromosomes, the F2 crosses successfully localized at

least one QTL, but on chromosome 6, the F2 cross produced a broad LOD peak that

extended over most of the chromosome and QTLs could not be localized precisely using this method. Analyses of the inheritance pattern at the markers near the highest LOD scores revealed evidence for overdominance, but overdominance was not observed in an

F1 population derived from C57BL/6J and B6-Chr 6A. Because the LOD peak was broad

and because overdominance was observed in the F2 but not F1 males, we concluded that

multiple genes contribute to obesity resistance on chromosome 6.

Because the F2 cross provided evidence for QTLs but did not precisely localize

them on chromosome 6, we hypothesized that congenic strains may be useful for fine- 148

mapping the QTL(s) on chromosome 6. Unlike an F2 cross in which each individual has

a unique genetic composition, congenic strains enable analysis of large samples of

genetically identical individuals and therefore, typically have more power than an F2 cross. The congenic panel derived from B6-Chr 6A successfully localized at least three

QTLs, and the complexity revealed in the congenic panel survey may explain why

precise QTL localization was not possible in the F2 cross.

2. Obrq1 is localized to a 17.4 Mb candidate interval

The critical interval of Obrq1 (Figure III-5) consists of approximately 17.4 Mb and contains at least 39 known and predicted gene sequences. This QTL probably accounts for the peak LOD near D6Mit284 (near 57 cM) in the F2 cross discussed in the previous chapter. The two most obvious candidate genes for the obesity resistance

phenotype are Slc25a26 (solute carrier family 25, mitochondrial carrier, phosphate carrier, member 26) and Suclg2 (succinate-Coenzyme A ligase, GDP-forming, beta subunit). Slc25a26 is known to encode a mitochondrial phosphate carrier protein and may be involved in energy metabolism. Likewise, Suclg2 is involved in the tri-

carboxylic acid cycle, which converts acetyl-CoA to CO2 and releases reducing

equivalents that are used to generate ATP in the electron transport chain. The candidate

interval for Obrq1 is in conserved synteny with various regions of human chromosome 3

(3p13-14, 3p21, 3p25-26, 3q21). Several obesity-related QTLs have been mapped to this

region in mouse (http://www.informatics.jax.org) and to the region in conserved synteny

in humans (PERUSSE et al. 2005).

149

3. The Obrq2 critical interval contains many genes associated with energy metabolism

The Obrq2 candidate interval, which contains at least 125 genes, extends from

12.2 Mb to 35.9 Mb. This region contains many strong candidate genes that are associated with various aspects of energy metabolism including Lep (leptin), Bpgm (2,3- bisphosphoglycerate mutase), Cav-1 (caveolin-1), Mest (mesodermal expressed transcript), Nrf1 (nuclear respiratory factor-1), Chchd3 (coiled-coil-helix-coiled-coil- helix domain containing 3), Ndufa5 (NADH dehydrogenase (ubiquinone) 1 alpha subcomplex), St7 (suppressor of tumorigenicity which encodes low density lipoprotein- related protein 12), and Ppp1r3a (protein phosphatase 1, regulatory (inhibitor) subunit

3A). The candidate interval is in conserved synteny with human chromosome 7 (7q21,

7q31-36) and as with Obrq1, several obesity-related QTLs have been mapped to this region in mouse (http://www.informatics.jax.org) and to the region in conserved synteny in humans (PERUSSE et al. 2005).

At least one imprinted region is present in the Obrq2 candidate interval. Mest, a

strong candidate gene for the obesity resistance phenotype in the imprinted region,

(BEECHEY 2000), is expressed in adipose tissue (TAKAHASHI et al. 2005). Although the

exact function of Mest is not known, Mest gene expression increases upon exposure of mice to a high-fat diet (TAKAHASHI et al. 2005). Because the reciprocal crosses derived from C57BL/6J and B6-Chr 6A produced male offspring with similar phenotypes when fed the high-fat diet, we concluded that the imprinting of genes, such as Mest, is probably not involved in the obesity resistance. The reciprocal crosses also demonstrated that other types of parental effects, such as maternal milk content, are not contributing to the phenotype. 150

4. Obrq2 and Obrq3 interact

Obrq3 was discovered as a suppressor of Obrq2 because many strains with A/J-

derived chromosomal segments including and extending beyond Obrq2 were obese. For

instance, 108-A, 115-A, and 105-A were obese even though these strains have the A/J- derived region associated with Obrq2. Consequently, a third QTL (Obrq3) must suppress the effects of Obrq2 in this region. Obrq3 extends from 38.7 Mb to 48.9 Mb, a region near the telomeric boundary of the critical interval for Obrq2 and may account for the

LOD peak near D6Mit274 (28 cM) in the F2 cross. Consequently, these interacting

QTLs are closely spaced. The critical interval for Obrq3 contains at least 114 genes and many interesting candidate genes including Ndufb2 (NADH dehydrogenase (ubiquinone)

1 beta subcomplex 2), a component of the electron transport chain. Interestingly, several genes, including Chchd3, Ndufa5, and Nrf1, within the Obrq2 candidate interval are also involved in oxidative phosphorylation, electron transport, or mitochondria biology.

Although any hypothesis about the nature or mechanism of the interaction between

Obrq2 and Obrq3 is speculative, mitochondrial biology could be central to this interaction. In addition to Ndfub2, the critical interval also includes several olfactory receptors and taste receptors, which could serve as excellent candidate genes because variation in the ability to taste or smell food may contribute to food consumption and thus, leanness. The critical interval for Obrq3 is in conserved synteny with the same regions of human chromosome 7 as Obrq2. Obrq3 appears to suppress only Obrq2 because all strains with Obrq1 and Obrq3 were resistant to obesity.

151

5. Further studies will refine QTL critical intervals.

The critical intervals for all three QTLs are large and contain many genes.

Furthermore, many of the genes within the regions are poorly characterized. Although

many genes associated with energy metabolism are identified as possible candidate

genes, no genes within the interval can be formally eliminated at this stage. Thus, further

mapping studies, including the construction of subcongenic strains, are needed to narrow

the critical interval and hence, the list of candidates. Furthermore, additional

phenotyping may enable characterization of the mechanism of resistance associated with these QTLs. Lastly, an understanding of the mechanism of resistance will assist in the

selection of important tissues for gene expression studies which may result in the identification of differentially expressed genes in the candidate interval or differential

expression of key pathways that may provide clues to the identity of the candidate gene.

6. Gene interactions contribute to the complexity of obesity genetics

The congenic strain survey revealed that chromosome 6, a single A/J-derived

chromosome, has at least three QTLs, one which acts independently and two which

interact. Thus, if other individual chromosomes exhibit similar complexity, then the

genetics of obesity resistance in the A/J inbred strain is more complex than predicted by

the CSS surveys. Evidence from other rodent models indicates that large numbers of

genes and gene interactions contribute to the genetics of obesity (BROCKMANN and

BEVOVA 2002; WARDEN et al. 2004). Gene interactions influencing obesity and obesity-

related traits have been observed in a DU6i X DBA/2 intercross (BROCKMANN et al.

2000), in a LG/J and SM/J intercross (CHEVERUD et al. 2001), in the BSB model for 152

obesity (YI et al. 2004), and in other crosses or mapping studies (WARDEN et al. 2004; YI et al. 2006). Most of these examples involve interactions between genes or QTLs on different chromosomes, but at least one study used subcongenic strains to demonstrate the existence of at least three interacting genes in a 24 cM region on chromosome 7

(DIAMENT and WARDEN 2004). It is unclear whether obesity genes are clustered on

certain chromosomes or chromosomal regions, such as the chromosome 6 region

containing Obrq2 and Obrq3, or whether this high density of obesity genes and

interactions among them characterizes other regions of the genome (DIAMENT and

WARDEN 2004). Thus, studies using the CSSs and congenic strains derived from them

will test whether the clustering of obesity-related genes on individual chromosomes is the

exception or the rule. Consequently, analyses with CSSs and congenic strains will begin

to narrow the regions and the list of candidate obesity resistance genes and provide an

understanding of the genetic architecture of obesity resistance and related traits.

153

CHAPTER IV: PHENOTYPIC DISSECTION OF RESISTANCE TO DIET-INDUCED OBESITY IN B6-CHR 6A CHROMOSOME SUBSTITUTION STRAIN MICE

The work described in this chapter was performed by the candidate under the supervision of Dr. Colleen Croniger and in collaboration with Dr. William O’Brien, who performed the sterol, carnitine, and amino acid analysis. The method used to measure liver triglyceride levels was developed by Dr. David Sinasac. 154

A. INTRODUCTION

As the prevalence of obesity continues to rise, the health conditions associated

with obesity, such as metabolic syndrome, also become major public health concerns.

Metabolic syndrome, which is defined as obesity, hypertension, insulin resistance, and

dyslipidemia, is a major risk factor for the development of type II diabetes mellitus and

cardiovascular disease (ALBERTI and ZIMMET 1998; BALKAU and CHARLES 1999;

EXPERT PANEL ON DETECTION 2001). Furthermore, although nonalcoholic fatty liver

disease is not a formal component of the clinical criteria for metabolic syndrome, it has

also been strongly associated and may be a hepatic component of the condition

(CHITTURI et al. 2002; MARCHESINI et al. 2001; MARCHESINI et al. 2003). Currently,

over 20% of U.S. adults have metabolic syndrome (FORD et al. 2002), and this prevalence

is likely to increase as obesity rates continue to rise. Despite the association of obesity

with these other phenotypes of metabolic syndrome, a single, shared pathogenesis has not

been discovered (KAHN et al. 2005), and the exact etiology remains unclear.

Like obesity, metabolic syndrome is obviously influenced by both genetic and

environmental factors. Unfortunately, studies of the genetics of metabolic syndrome have

proven even more challenging than obesity. The individual components of metabolic

syndrome are heritable (ARYA et al. 2002; KISSEBAH et al. 2000; MCQUEEN et al. 2003;

SHMULEWITZ et al. 2006), and studies of the genetics of obesity and other component traits of metabolic syndrome have identified unique QTLs for one or two individual component traits (ARYA et al. 2002; KISSEBAH et al. 2000; SHMULEWITZ et al. 2006). In

contrast, few studies have identified QTLs that strongly influence several component traits simultaneously (MCQUEEN et al. 2003; TANG et al. 2003). Furthermore, no genes 155

have been unambiguously associated with all components of metabolic syndrome in the

human population. Consequently, it is unclear whether the genetic control of metabolic

syndrome is due to a collection of genes that increase susceptibility to one or two

component phenotypes, to a collection of genes that increase susceptibility to many or all

component phenotypes, or to a combination of both.

Mouse models of obesity provide important tools for studying the genetics of

metabolic syndrome. In particular, inbred mouse strains are useful because of their wide

range of phenotypic variation and because of the availability of strains that differ in

susceptibility to obesity and features of metabolic syndrome. For example, in addition to

obesity, C57BL/6J male mice also develop many features similar to metabolic syndrome

phenotypes (COLLINS et al. 2004), including hypertension (MILLS et al. 1993),

hypercholesterolemia (REBUFFE-SCRIVE et al. 1993), hyperglycemia (REBUFFE-SCRIVE et al. 1993; SURWIT et al. 1995; SURWIT et al. 1988), hyperinsulinemia (REBUFFE-SCRIVE et al. 1993; SURWIT et al. 1995; SURWIT et al. 1988), and fatty liver (D. Sinasac, personal

communication), but no rigorous genetic investigations of the metabolic syndrome have

been performed in these strains.

A Because of the variation in body weight among the B6-Chr CSSs (SINGER et al.

2004), these strains are useful for dissecting the genetics and phenotypes associated with obesity. In the previous chapters, we demonstrated that B6-Chr 6A is resistant to high-fat,

diet-induced obesity and our genetic analyses using an intercross and a panel of congenic

strains revealed that chromosome 6 harbors at least three QTLs associated with body

weight. Whether the obesity resistance QTLs on chromosome 6 also confer resistance to

other features of the metabolic syndrome has not been investigated. 156

In the present study, we characterized the obesity resistance in B6-Chr 6A males and tested whether A/J-derived chromosome 6 confers resistance to other features of the metabolic syndrome. First, we confirmed that the decreased body weight and BMI are surrogate measures of adiposity by demonstrating that the fat pads are smaller in the

resistant strain. We also demonstrated that food intake does not explain the resistance.

Then, we demonstrated that B6-Chr 6A males are resistant to several aspects of the

metabolic syndrome, including hyperglycemia, hypercholesterolemia, and elevated liver

triglyceride levels in addition to obesity. Similarly, we tested an obesity resistant

congenic strain (62-B) with Obrq2, which was described in the Chapter 3, and demonstrated that this strain is also resistant to hyperglycemia, elevated liver triglyceride levels, and possibly hypercholesterolemia. Consequently, B6-Chr 6A is a polygenic

model for resistance to high-fat, diet-induced obesity, and congenic strains derived from

this CSS, such as 62-B, are useful tools for investigating the underlying mechanisms of

obesity resistance and the association of other obesity-related traits on chromosome 6.

B. MATERIALS AND METHODS

1. Mice: For phenotyping studies, one set of C57BL/6J and A/J male mice was obtained

from the Jackson Laboratory at four weeks of age (Bar Harbor, ME). For the congenic phenotyping study, a second set of C57BL/6J and A/J male mice was raised at CWRU.

All B6-Chr 6A and 62-B male mice were raised at CWRU. Mice were weaned at 3-4

weeks of age and maintained in microisolator cages with a 12 hour:12 hour light:dark

cycle. The 62-B mice used in this study were obtained prior to the generation of the 62-

BL and 62-BS congenic strains discussed in the previous chapter. 157

2. Diet studies: For all studies, the mating colonies and weaned mice were fed LabDiet

5010 (LabDiet, Richmond, IN) ad libitum until diet studies were initiated. The mice

were introduced to the HFSC diet (D12331, Research Diets, New Brunswick, NJ) at 35

days of age and fed ad libitum. Body weights were collected every two weeks for the

duration of the study (approximately 50 or 100 days). Diet composition is provided in

chapter 2 (Table II-1).

3. Fat pad mass: To assess adiposity, gonadal fat pads were collected from C57BL/6J

(n=17), A/J (n=10), and B6-Chr 6A (n=13) males and weighed. To calculate BMI,

nasoanal length was measured, and the FW (grams) was divided by length (in

centimeters) squared.

4. Histology: Sections (n=4 individuals) of liver were collected, stored in 10% formalin,

and submitted to the CWRU Histology Core for hematoxylin and eosin (H&E) staining after both 50 and 100 days of high-fat diet exposure.

5. Food consumption studies: Six to eight males per strain (C57BL/6J, A/J, and B6-Chr

6A) were collected, singly housed, and introduced to the HFSC diet at 35 days of age.

Twenty-four hour food consumption was measured for four consecutive days at 35, 85,

and 135 days of age. During the four-day periods, a thin layer of bedding was provided

and a known quantity of food was placed in the cage. Twenty-four hours later, the remaining food (including any large pieces that fell into the bedding) was weighed, the 158

quantity was recorded, and the procedure was repeated for three additional days. The

mean food consumption for each mouse at each time point was calculated. Mean body

weight, absolute food consumption, and food consumption per gram body weight were

analyzed at each time point.

6. Blood chemistry: Blood chemistry was performed in approximately 10 C57BL/6J,

A/J, and B6-Chr 6A male mice after 50 days (85 days of age) and 100 days (135 days of

age) of HFSC diet exposure. Prior to blood collection, the mice were fasted for 24 hours.

Then, the mice were anesthetized with an intraperitoneal injection of tribromoethanol

(stock solution: 1.6 g/mL in tert-amyl alcohol; working solution: 20.5 mg/mL in

potassium-free phosphate buffered saline; dosage: 0.20 mL/10 grams of body weight).

The tail was clipped, and tail blood glucose was measured (One Touch Ultra blood

glucose monitoring system, LifeScan, Milpitas, CA). Blood was collected by cardiac

puncture into a 1 cc syringe (Allergy Syringe, 26G1/2, Becton Dickinson, Franklin

Lakes, NJ). The blood was placed in a heparinized tube (Microtainer plasma separator

tube with heparin, Becton Dickinson, Franklin Lanes, NJ), centrifuged at 2000g for 5

minutes. Plasma was collected, stored at 4°C, and submitted to Marshfield Laboratories

(Marshfield, WI) within 12 hours for the following measurements: glucose, triglycerides,

total cholesterol, β-hydroxybutyrate, and blood-urea nitrogen (BUN). Insulin was

measured using the Mercodia ultrasensitive insulin ELISA kit (Mercodia, Uppsala,

Sweden). Sections of liver were collected, frozen, and stored at -20ºC for additional

studies (see below). Similar studies were performed in 62-B and a second set of

C57BL/6J and A/J control male mice. 159

7. Liver triglyceride measurements: The level of triglyceride in liver sections was

measured based on the Salmon and Flatt’s method (SALMON and FLATT 1985) of liver

saponification (Dr. David Sinasac, personal communication). 100-200 mg of liver tissue was saponified using 3M potassium hydroxide (in 65% ethanol) for one hour at 70ºC and then overnight at room temperature. Each sample was diluted to 100 mg/500 ml in 50mM

Tris-HCl. The triglyceride (GPO) reagent set (Pointe Scientific, Canton, MI) was used to

determine the concentration of glycerol in each sample. The concentration of glycerol

was then converted to the concentration of triglyceride in the sample using the average

molecular weight of triglyceride (885 grams/mole).

8. Plasma carnitine, sterol, and amino acid studies: All plasma carnitine, sterol, and

amino acid measurements were performed at Baylor College of Medicine (Houston, TX).

Plasma neutral sterols were measured with gas chromatography/mass spectrometry

(GC/MS; Hewlett Packard 6890 GC with a 5973 mass spectrometer) using selected ion monitoring mode. Cation exchange chromatography on a high performance liquid chromatography (HPLC) system (Dionix) using post-column derivitization with ninhydrin reagent (Pickering) was used to quantify plasma amino acids. Acylcarnitine profiles were measured with tandem mass spectroscopy (Micromass Quatro) (VREKEN et al. 1999).

160

9. Statistical methods

a. Correlation coefficients: Pearson’s correlation coefficients were calculated for gonadal

fat pad weights and both FW and BMI.

b. Metabolic phenotyping: An unpaired t-test was used to compare the traits for all

strains to C57BL/6J, to compare C57BL/6J and A/J males obtained from the Jackson

Laboratory and those raised at CWRU, and to compare the 62-B blood chemistry to the

pooled C57BL/6J (JAX and CWRU) (GraphPad Prism version 3.0). If the variance of

the samples differed significantly (based on an F test), Welch’s t-test was used (GraphPad

Prism, version 3.0).

C. RESULTS

1. Decreased adiposity in B6-Chr 6A males

To test whether the decreased body weight in B6-Chr 6A male mice is associated

with decreased adiposity, the BMI and size of gonadal fat pads from C57BL/6J, A/J, and

B6-Chr 6A male mice fed the HFSC diet was investigated. Both A/J and B6-Chr 6A had significantly lower BMI and smaller gonadal fat pads relative to C57BL/6J. The fat pads from A/J were 35% smaller (A/J: 1.70 +/- 0.66 grams vs. C57BL/6J: 2.60 +/- 0.56 grams; t=3.737, df=25, p =0.001) than C57BL/6J, and the fat pads from B6-Chr 6A were 27%

smaller (B6-Chr 6A/J: 1.89 +/- 0.85 grams; t=2.755, df=28, p=0.01) than C57BL/6J

(Figure IV-1). Although fat pad mass was highly variable in B6-Chr 6A males, gonadal

fat pad mass was highly correlated with both FW (r=0.92) and BMI (r=0.92) (Figure IV-

2).

161

Figure IV-1. Gonadal fat pad weights and BMI in C57BL/6J (B6) relative to A/J and B6-Chr 6A (CSS-6) fed the HFSC diet for 100 days. The mean value for each strain is indicated with a bar. The p values derived from an unpaired t-test with comparison to C57BL/6J are indicated.

P=0.01 4 P=0.001

(grams)

Gonadal fat weightpad 0 B6 A/J CSS-6 Strain

P<0.0001 0.5 P<0.0001

) 2 0.4

0.3

0.2

BMI (grams/cm 0.1

0.0 B6 A/J CSS-6 Strain

162

Figure IV-2. Pearson’s correlation coefficients for fat pad weight vs. final body weight and fat pad weight vs. BMI in B6-Chr 6A males fed the HFSC diet. The dashed lines indicate the 95% confidence intervals for the best-fit line.

36 Final bodyFinal weight (grams)

30 0.835 1.256 1.575 1.881 2.305 2.601 2.961 3.396 Fat pad weight (grams)

0.42

0.40

0.38 ) 2 0.36

0.34 BMI (grams/cm

0.32

0.30

0.28 0.835 1.256 1.575 1.881 2.305 2.601 2.961 3.396 Fat pad weights (grams)

163

Together with the decreased body weight and BMI, the smaller gonadal fat pads provide

further evidence for reduced adiposity in B6-Chr 6A males.

2. Hypophagia does not explain obesity resistance in B6-Chr 6A

We hypothesized that the obesity resistance in B6-Chr 6A males may be due to

hypophagia relative to C57BL/6J. To test this hypothesis, we measured HFSC diet

consumption in C57BL/6J, A/J, and B6-Chr 6A males. A/J males did not eat significantly

less food than C57BL/6J males at any time point. In contrast, although B6-Chr 6A males consumed more food than C57BL/6J upon initial exposure to the diet (p=0.04), this strain consumed less food after 50 days (p=0.02) and 100 days (p=0.005) of diet exposure

(Figure IV-3, Table IV-1). The differences in food consumption at the later two time points were between 0.2 and 0.4 grams per day. When body weight is taken into account,

A/J consumed significantly more food per gram body weight (35 days: p=0.002; 85 days: p=0.002; 135 days: p=0.002), but the food intake per gram body weight did not significantly differ between C57BL/6J and B6-Chr 6A at any time point. Consequently,

relative to C57BL/6J, hypophagia does not appear to explain the resistance in B6-Chr 6A.

3. Hypertriglyceridemia but decreased plasma cholesterol and glucose in B6-Chr 6A

To test whether B6-Chr 6A males are resistant to various aspects of metabolic

syndrome and to investigate possible differences in fuel utilization in the fasted state,

blood chemistry was analyzed in C57BL/6J, A/J, and B6-Chr 6A male mice. After

consuming the high-fat diet for 50 days, the only significant difference in blood

164

Figure IV-3A. HFSC diet consumption in C57BL/6J (B6), A/J, and B6-Chr 6A (CSS-6) males: 35 days of age. Body weight, mean 24-hour food intake, and mean 24-hour food intake per gram body weight is presented for each strain. The mean for each strain is indicated with a bar. The p values were derived from unpaired t-tests with comparison to C57BL/6J.

22.5 P=0.0006

20.0

17.5

(grams)

15.0

days 35 at weight Body 12.5 B6 A/J CSS-6 Strain

3.25 P=0.04

3.00

2.75

(grams) 2.50

2.25 Food intake atdays 35 2.00 B6 A/J CSS-6 Strain

P=0.002 0.200

0.175

0.150

35 days

0.125 Food intake (grams) / (grams) intake Food

body weight (grams at

0.100 B6 A/J CSS-6 Strain 165

Figure IV-3B. HFSC diet consumption in C57BL/6J (B6), A/J, and B6-Chr 6A (CSS-6) males: 85 days of age. Body weight, mean 24-hour food intake, and mean 24-hour food intake per gram body weight at 85 days of age (50 days of high-fat diet consumption) is presented for each strain. The mean for each strain is indicated with a bar. P values were derived from unpaired t-tests with comparison to C57BL/6J or Welch’s t-test if variance significantly differed.

P=0.0008 35

(grams) 25

days 85 at weight Body 20 B6 A/J CSS-6

Strain

3.5 P=0.02

3.0

(grams) 2.5

intakeFood atdays 85 2.0 B6 A/J CSS-6 Strain

0.150 P=0.002

0.125

0.100 0.075

85 days 0.050 0.025 Food intake (grams) /

body weight (grams) at 0.000 B6 A/J CSS-6 Strain 166

Figure IV-3C. Food consumption in C57BL/6J (B6), A/J, and B6-Chr 6A (CSS-6) males: 100 days of diet consumption. Body weight, mean 24-hour food intake, and mean 24-hour food intake per gram body weight at 135 days of age (100 days of high-fat diet exposure) is presented. The mean value for each strain is indicated with a bar. The p values were derived from an unpaired t-test (or Welch’s t-test if the variance significantly differed using an F test).

50 P=0.0004

(grams) 20

Body weight atdays 135 0 B6 A/J CSS-6

Strain

P=0.005 3.0

2.5

(grams) 2.0

Food intake at days 135 1.5 B6 A/J CSS-6 Strain

0.15 P=0.002

0.10

135 days 0.05

Food intake (grams) / (grams) intake Food

at (grams) weight body 0.00 B6 A/J CSS-6 Strain 167

Table IV-1. HFSC diet consumption in C57BL/6J, A/J, and B6-Chr 6A males. The mean trait value for each sample +/- the standard deviation is presented for each trait. An unpaired t-test was used to compare the mean from each sample to that of C57BL/6J. *indicates Welch’s t-test was used. Graphical representation of the data is in Figure IV-3.

A. Food consumption

Food intake Strain at 35 days (grams) n T statistic df P value C57BL/6J 2.46 +/- 0.16 7 ** ** ** A/J 2.54 +/- 0.11 8 1.046 13 0.31 A B6-Chr 6 2.71 +/- 0.25 7 2.263 12 0.043

Food intake Strain at 85 days (grams) n T statistic df P value C57BL/6J 2.65 +/- 0.24 7 ** ** ** A/J 2.92 +/- 0.29 6 1.855 11 0.0906 A B6-Chr 6 2.37 +/- 0.14 7 2.697 12 0.0194

Food intake Strain at 135 days (grams) n T statistic df P value C57BL/6J 2.60 +/- 0.24 8 ** ** ** A/J 2.59 +/- 0.27 7 0.1182 13 0.9077 A B6-Chr 6 2.21 +/- 0.19 7 3.418 13 0.0046

B. Food consumption per gram body weight

35 Day Food intake / Strain gram body weight n T statistic df P value C57BL/6J 0.14 +/- 0.012 7 ** ** ** A/J 0.16 +/- 0.010 8 3.981 13 0.0016 A B6-Chr 6 0.15 +/- 0.019 7 1.034 12 0.3215

85 Day Food intake / Strain gram body weight n T statistic df P value C57BL/6J 0.084 +/- 0.011 7 ** ** ** A/J 0.12 +/- 0.016 6 4.208 11 0.0015 A B6-Chr 6 0.087 +/- 0.0049 7 0.6124 12 0.5517

135 Day Food intake / Strain gram body weight n T statistic df P value C57BL/6J 0.069 +/- 0.015 8 ** ** ** A/J 0.096 +/- 0.013 7 3.788 13 0.0023 B6-Chr 6A 0.066 +/- 0.0053 7 0.5484 9* 0.5968

168

chemistry was slightly elevated plasma insulin (p=0.04) in B6-Chr 6A males relative to

C57BL/6J. In contrast, after consuming the high-fat diet for 100 days, significant

differences in plasma triglyceride, glucose, cholesterol, BUN, and β-hydroxybutyrate

were detected. While plasma cholesterol (p=0.005) and glucose (p=0.02) in B6-Chr 6A

were significantly lower than C57BL/6J, the plasma triglyceride levels were significantly

higher (p=0.01) in B6-Chr 6A males. Likewise, β-hydroxybutyrate was also elevated in

B6-Chr 6A males (p=0.03) indicating that triglyceride catabolism may be increased in this

resistant strain. In addition, BUN was significantly elevated in B6-Chr 6A relative to

C57BL/6J (p<0.0001) indicating that protein catabolism or kidney function may differ in

these strains. In contrast, plasma insulin did not differ significantly among the strains,

but insulin values were highly variable in the study. Perhaps, analyses of plasma insulin

after a shorter fast may be more informative. Overall, when fed the high-fat diet for 100

days, B6-Chr 6A males are resistant to several components of metabolic syndrome

including hypercholesterolemia and hyperglycemia. Blood chemistry is listed in Table

IV-2 and IV-3.

4. B6-Chr 6A males are resistant to fatty liver

To test whether C57BL/6J, A/J, and B6-Chr 6A males differ in susceptibility to fatty liver, we examined H&E stained histological sections of liver from each strain

(Figure IV-4 and IV-5). After 50 days of high-fat diet exposure, the liver from C57BL/6J males was characterized by extensive microvesicular and macrovesicular fat deposition. 169

st. **One outlier value of outlierst. **One value

using an unpaired t-test (or t-test unpaired an using

T statistic df P value

(n~10)

1 +/- 22 3 +/- 14 1.495 9* 1.364 0.1692 17 0.1903 152 +/-30 1.440 17 0.1680 0.23 +/- 0.08 2.288 15 0.0371

28.26 +/-2.04 1.909 18 0.0723

B6-Chr 6

at 85 days of age (50 days of high-fat diet consumption). The The consumption). diet high-fat of days (50 age of 85 days at A

st), and significant p values are bold. *indicates Welch’s t-te esented for each trait. Each sample was compared to C57BL/6J to C57BL/6J was compared trait. Each sample for each esented

2 +/- 17 4.240 5 +/- 18 11* 3.833 0.0014 13* 0.0021 77 +/- 12 77 +/- 12 4.058 9* 0.0028 +/- 11 64 5.037 9* 0.0007 214 +/- 20 21420 +/- 3.274 18 0.0042

25.832.04 +/- 4.006 18 0.0008

6 +/- 25 2 +/- 12 173 +/- 34 23 +/- 113 9 +/- 103 118 +/- 22 1.367 13026 +/- 133 +/- 40 11* 0.1989 1.098 18 8 +/- 107 0.2868 12222 +/- 0.8561 11* 0.4102 0.3469 17 0.7329 **97 +/- 74 +/- **97 +/-16 69 1.139 9* 0.2840 25+/- 68 1.157 11* 0.2718 0.15 +/- 0.08 +/- 0.15 0.07 +/- 0.19 1.2090 15 0.2455 30.46 +/- 3.03

g/dl)

Trait (n~10) B6 (n~10) A/J Tstatistic df P value (mg/dl) (mg/dl) (mg/dl)

( Insulin -Hydroxybutyrate -Hydroxybutyrate Glucose (mg/dl) Glucose Total Cholesterol Cholesterol Total Liver Triglycerides Β Blood Urea Nitrogen Nitrogen Urea Blood Triglycerides (mg/dl) Triglycerides (mg per gramtissue) Tail Glucose (mg/dl) Glucose Tail Body Weight (grams) Weight Body mean trait value for each sample +/- the standard deviation is pr is deviation standard the +/- sample each for value mean trait Welch’s t-test if variances differed from C57BL/6J using an F te an using C57BL/6J from differed variances if t-test Welch’s 275mg/dl,without andoutlier the is 76.67value mean41.90. +/- Table IV-2. Blood chemistry and liver triglycerides in C57BL/6J (B6), A/J, and B6-Chr 6 B6-Chr and A/J, (B6), C57BL/6J in triglycerides and liver chemistry Blood IV-2. Table 170

2.694 18 0.0148 /6J using an unpaired t- unpaired an using /6J T statistic df P value

Welch’s t-test used.Welch’s t-test (n~10) A

103 +/- 31 31 103 +/- 150 +/- 15 +/- 150 1.992 18 0.0678 204 +/- 29 29 204 +/- 2.447 18 0.0249

B6-Chr 6

sample was compared to C57BL was compared sample at 135 daysof (100 agedays high-fat of diet consumption).

A (<00.5) are bold. * indicates * indicates bold. are (<00.5)

presented for each trait. Each for each presented

an F test). Significant p values Significant p values an F test). 17 +/- 2 2 +/- 17 4.674 3 +/- 18 18 3.079 0.0002 18 0.0065 4 +/- 29 4.9942 +/- 16 18 <0.0001 2.430 18 0.0258 80 +/- 24 24 +/- 80 5.506 13* 0.0001 33 +/- 109 3.457 18 0.0028 101 +/- 15 15 +/- 101 6.054 13* 26 +/- 138 <0.0001 2.656 18 16 +/- 129 3.334 0.0161 13* 0.0054 30.77 +/-3.55 +/-3.55 30.77 7.073 18 <0.0001 5.92 +/- 36.27 2.956 18 0.0084

21 +/- 3 21 +/- 3 13 +/- 66 +/- 29 +/- 66 26 +/- 76 0.7733 18 0.4494 245 +/- 45 +/- 245 30 +/- 166 17 +/- 224 1.378 29 +/- 171 11* 0.1957 45 +/- 169

0.55 +/- 0.87 0.28 +/- 0.49 1.957 13* 0.0722 1.04+/- 1.04 0.4536 13* 0.6576 43.05 +/- 4.18 +/- 43.05

g/dl) µ

Trait (n~10) B6 (n~10) A/J T statistic df P value (mg/dl)

( Insulin Glucose (mg/dl) Glucose Liver Triglycerides Triglycerides Liver Blood Urea Nitrogen Nitrogen Urea Blood Triglycerides (mg/dl) Triglycerides (mg per gram tissue) Tail Glucose (mg/dl) Glucose Tail Body Weight (grams) Weight Body -Hydroxybutyrate (mg/dl) -Hydroxybutyrate Total Cholesterol (mg/dl) Total Cholesterol Β The mean trait value for each sample +/- the standard deviation is deviation standard the +/- sample each for value trait The mean test (or Welch’s t-test if variancesdifferedWelch’s t-test if by significantlytest (or Table IV-3. Blood chemistry and liver triglycerides in C57BL/6J (B6), A/J, and B6-Chr 6 B6-Chr and A/J, (B6), C57BL/6J in triglycerides liver and chemistry Blood IV-3. Table 171

The extent of fatty deposition appeared to be lower in A/J and B6-Chr 6A males. These

differences were consistent after 100 days of diet exposure.

To confirm that hepatic triglyceride content was elevated in C57BL/6J vs. A/J and

B6-Chr 6A males, we measured the level of triglyceride in liver sections from all three

strains after 50 and 100 days of high-fat diet exposure (Tables IV-2 and IV-3; Figures IV-

4 and IV-5). After 50 days of high-fat diet consumption, the liver triglycerides were 42% lower in A/J (77 +/- 12 mg/gram liver; t=4.058, Welch’s df=9, p=0.003) and 52% lower in B6-Chr 6A (64+/- 11 mg/gram liver; t=5.037, Welch’s df=9, p=0.0007) as compared to

C57BL/6J (133+/-40 mg/gram liver). Similar trends were observed following 100 days

of diet consumption with significantly lower levels of liver triglycerides detected in A/J

(80 +/- 24 mg/gram liver; t=5.506, Welch’s df=13, p=0.0001) and B6-Chr 6A males

(109+/-33 mg/gram liver; t=3.457, df=18, p=0.003) compared to C57BL/6J (169 +/- 45 mg/gram liver). Consequently, B6-Chr 6A males are also resistant to the development of fatty liver in response to the high-fat diet (Figure IV-4 and IV-5).

5. 62-B is also resistant to metabolic syndrome phenotypes

To test whether Obrq2, an obesity resistance QTL described in the previous

chapter, also confers resistance to various aspects of the metabolic syndrome, blood

chemistry and liver triglyceride measurements were obtained in C57BL/6J, A/J, and 62-B

male mice raised at CWRU and fed the HFSC diet. As described previously, 62-B is a

172

(n=4)

in TableIV-2.

The p values were derived from from derived were p values The CSS-6

P=0.0007 50 days of HFSC diet consumption. A. H&E sections (400x (400x H&E sections A. consumption. diet HFSC of days 50 r each strain is indicated with a bar. bar. indicated with a is strain r each Strain mparison to C57BL/6J. Data also presentedmparison to C57BL/6J. P=0.003 A/J (n=4) A/J (n=4) B6 A/J CSS-6 at 85 days of age (50 days of high-fat diet consumption) are presented. B. Liver B. presented. are consumption) diet high-fat of days (50 of age days 85 at

300 200 100 days of age age of of days days

(mg/gram liver) at 85 at liver) (mg/gram

85 at liver) (mg/gram

Liver triglyceride triglyceride Liver Liver

ers of all threeers The strains.value mean fo

variances differed using an F test) with co using differed variances

. (CSS-6) is resistant to(CSS-6) is resistant the development of fatty liver after

C57BL/6J (n=4) (n=4) C57BL/6J

B. Figure IV-4.B6-Chr 6 magnification) of liver from C57BL/6J (B6), A/J, and B6-Chr 6 B6-Chr and A/J, (B6), C57BL/6J from of liver magnification) triglyceride levels were measured the in liv an unpaired t-test (or Welch’s t-test if unpairedan (or t-test A.

173

so presented in Table IV- 3. IV- Table in presented so (n=4)

The p values were derived from from were derived p values The

CSS-6

P=0.003 st) with comparisonData al st) to C57BL/6J. Strain A/J (n=4) A/J (n=4) P=0.0001 at 135 days of age (100 days of high-fat diet consumption) are presented. B.The liver liver B.The presented. are consumption) diet high-fat of days (100 age of days 135 at A B6 A/J CSS-6 ins. The mean value for each strain is indicated with a bar. bar. with a strain is indicated for each value mean ins. The

300 200 100

days of age of days

age of days

(mg/gram liver) at 135 at liver) (mg/gram

135 at liver) (mg/gram Liver triglyceride triglyceride Liver Liver variance differed significantly using an F te

(CSS-6)resistant is todevelopment the of fatty liver after100days of HFSC diet consumption.H&E A. sections(400x A

C57BL/6J (n=4) (n=4) C57BL/6J

B. triglyceride levels were measured livers in of the all three stra an unpaired t-test (or Welch’s t-test if unpairedan (or t-test Figure IV-5.B6-Chr 6 A. magnification) of liver from C57BL/6J (B6), A/J, and B6-Chr 6 B6-Chr and A/J, (B6), C57BL/6J from of liver magnification) 174

congenic strain derived from B6-Chr 6A that harbors Obrq2. Before comparisons were

made among the strains, the blood chemistry and the body weights from the C57BL/6J

and A/J male mice obtained from Jackson Laboratory were compared to those raised at

CWRU, and no significant differences (p<0.05) were observed. Like B6-Chr 6A, 62-B males exhibited elevated BUN (p<0.0001) and plasma triglyceride (p<0.0001) levels with decreased plasma glucose (p=0.003) and liver triglyceride content (p=0.001) relative to

C57BL/6J. Unlike B6-Chr 6A, no differences in plasma total cholesterol or β-

hydroxybutyrate were observed (Table IV-4).

To further investigate the blood chemistry results, especially plasma cholesterol,

62-B was also compared to the pooled C57BL/6J (Jackson Laboratory and CWRU)

sample. In this comparison, all results, except total cholesterol, were similar to the un-

pooled comparison. For cholesterol, 62-B had a significantly lower value than C57BL/6J

(p=0.02). A closer examination of the total cholesterol samples identified one outlier

(value = 65 mg/dl, all other samples ranged from 118-180 mg/dl) in the CWRU

C57BL/6J sample. When this outlier is removed, the cholesterol in 62-B is also

significantly different from C57BL/6J in the original comparison (p=0.04). This result

suggests that 62-B may also be resistant to hypercholesterolemia relative to C57BL/6J

but that the level of resistance is not as extreme as B6-Chr 6A.

6. Sterol biosynthesis intermediates are reduced in B6-Chr 6A and 62-B

To further investigate the plasma cholesterol difference observed in B6 vs. B6-

Chr 6A and 62-B, several sterol biosynthesis intermediates were measured in plasma.

175

f the variance differed differed f the variance at 135 days of age (after of age days at 135

relativeother analyses. to the

112 +/- 0.4855 11* 0.6368 159 +/- 18159 +/- 10 133 +/- 3.801 0.8438 12* 12* 0.0025 0.4153

57BL/6J (B6) using an unpaired t-test (Welch’s t-test was used i was used t-test (Welch’s t-test unpaired an using (B6) 57BL/6J e blood chemistry and liver triglyceride analyses were performed were analyses triglyceride liver and chemistry e blood

17 +/- 2 17 +/- 5 19 +/- 5.608 2.485 18 18 <0.0001 0.023 2 30 +/- 6.889 19 <0.0001 94 +/- 20 94 +/- 2.812 10* 0.0184 17 109 +/- 6.037 12* <0.0001 80 +/- 24 80 +/- 5.506 13* 0.0001 28 101 +/- 3.921 17 0.0011 105 +/- 16 105 +/- 3.415 15* 18 121 +/- 0.0038 2.766 20 0.0119 8 103 +/- 5.711 13* <0.0001

5.30 34.11 +/- 4.883 20 <0.0001 3.13 32.95 +/- 7.153 20 <0.0001

23 +/- 3 23 +/- 6 13 +/- 8 74 +/- 220 +/- 49 +/- 220 31 +/- 142 31 205 +/- 0.7961 26 +/- 148 18 0.4363 45 +/- 169 0.87 +/- 0.55 +/- 0.87 0.28 0.49 +/- 1.957 13* 0.0722 28 0.45 +/- 2.122 13* 0.0536 43.46 +/- 3.66 +/- 43.46

g/dl)**

µ Trait B6 (n~10) (n~10) A/J T statistic df P value 62-B (n~10) T statistic df P value (mg/dl) (mg/dl) (mg/dl) Insulin ( Insulin -Hydroxybutyrate -Hydroxybutyrate Glucose (mg/dl)Glucose Total Cholesterol Β

Liver Triglycerides** Blood Urea Nitrogen Triglycerides (mg/dl) Triglycerides Tail Glucose (mg/dl) Glucose Tail (mg per gram tissue) gram per (mg Body Weight (grams) Weight Body 100 days of HFSC diet consumption). Comparisons were made to C made to were Comparisons consumption). diet HFSC of days 100 TableMetabolic IV-4. phenotypingcongenic62-B in strain. Th significantlyusing Indicates*F test). an Welch’s used t-test **Indicates a differentpopulation used was for this analysis 176

As compared to humans, mice synthesize cholesterol at much higher rates (DIETSCHY and

TURLEY 2002). Thus, we hypothesized that variation in the rate or level of cholesterol

biosynthesis may explain the reduced levels of total plasma cholesterol in B6-Chr 6A

males. In humans, desmosterol and lathosterol levels, two cholesterol biosynthesis

intermediates, are markers for cholesterol biosynthesis (MIETTINEN et al. 1990). We,

therefore, measured desmosterol, lathosterol and plasma concentrations of other

cholesterol biosynthetic intermediates. Although for unclear reasons total cholesterol did not significantly differ using this assay, desmosterol, lathosterol and 7- dehydrocholesterol were significantly lower in B6-Chr 6A and 62-B relative to C57BL/6J

suggesting that cholesterol biosynthesis differs among these strains (Table IV-5).

In addition, recent evidence suggests that plant sterols, which are consumed in the diet, also contribute to the plasma cholesterol pool Patients with defects in plant sterol metabolism have high levels of these sterols in the blood and are at risk for coronary heart disease (BERGE et al. 2000; LEE et al. 2001). Furthermore, because plant sterols must be absorbed in the gut (rather than synthesized), the levels of plant sterols, in particular campesterol, may reflect the level of cholesterol absorbed from the diet

(MIETTINEN et al. 1990). Consequently, we tested whether the plasma concentration of

two plant sterols, campesterol and sitosterol, differ in the strains. Both campesterol and

sitosterol were significantly decreased in both B6-Chr 6A (campesterol: p=0.01 and

sitosterol: p=0.02) and 62-B (campesterol: p=0.03 and sitosterol: p=0.004) relative to

C57BL/6J (Table IV-5). Consequently, either differences in plant sterol metabolism or

differences in cholesterol absorption may explain the decreased levels of plant sterols.

Further investigations of cholesterol metabolism are needed to explore these differences. 177

Table IV-5. Sterol measurements in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and 62-B congenic strain. An unpaired t-test was used to compare each trait in each strain to C57BL/6J. Welch’s t-test was used if variance differed using an F test). Significant p values (<0.05) are bold. Abbreviations: std dev = standard deviation.

Sterol (µg/L) C57BL/6J A/J CSS-6 62-B 1255.66 +/- 231.25 1011.96 +/- 131.10 1213.11 +/- 67.33 1138.86 +/- 88.92 Total Cholesterol Mean +/- std dev (n=10) (n=9) (n=8) (n=8)

P value ** 0.0128 0.5922 0.1680

7.67 +/- 2.98 5.11 +/- 0.82 6.47 +/- 0.93 5.61 +/- 0.59 Cholestanol Mean +/- std dev (n=9) (n=9) (n=8) (n=8)

P value ** 0.0348 0.2814 0.0771

3.27 +/- 0.52 2.68 +/- 0.36 2.35 +/- 0.51 1.89 +/- 0.45 7-Dehydrocholesterol Mean +/- std dev (n=10) (n=9) (n=8) (n=8)

P value ** 0.011 0.0017 <0.0001

4.01 +/- 0.76 2.43 +/- 0.29 3.08 +/- 0.26 2.64 +/- 0.78 Lathosterol Mean +/- std dev (n=9) (n=3) (n=7) (n=7)

P value ** 0.0064 0.0065 0.0033

0.77 +/- 0.14 0.56 +/- 0.07 0.59 +/- 0.05 0.56 +/- 0.05 Desmosterol Mean +/- std dev (n=10) (n=7) (n=8) (n=8)

P value ** 0.0020 0.0026 0.0008

1.74 +/- 0.55 2.42 +/- 0.6 1.02 +/- 0.46 0.96 +/- 0.70 Campesterol Mean +/- std dev (n=9) (n=9) (n=7) (n=7)

P value ** 0.0224 0.0144 0.0253

2.73 +/- 0.56 2.85 +/- 0.28 1.74 +/- 0.49 1.34 +/- 0.39 Sitosterol Mean +/- std dev (n=5) (n=7) (n=5) (n=4)

P value ** 0.6385 0.0181 0.0041

178

7. Obrq2 does not modulate variation in plasma amino acids

We hypothesized that the elevated BUN in B6-Chr 6A and 62-B may be due to

either increased ammonia generation from protein catabolism during the 24-hour fast, to

differences in flux through the urea cycle, or to variations in the clearance of BUN by the

kidney. To test this hypothesis, we measured plasma amino acids, including the branched

chain amino acids, and several urea cycle intermediates in the strains following 100 days

of HFSC diet exposure (Table IV-6). We hypothesized that differences in protein

catabolism would be reflected by increased levels of the branched chain intermediates,

which rise during starvation as a result of protein catabolism (VAZQUEZ et al. 1985).

Likewise, we hypothesized that if variation in the urea cycle explained the elevated BUN,

differences in amino acids involved in the urea cycle would be observed. Overall, no

differences in any of the branched chain amino acids or urea cycle amino acids were

observed. Thus, the elevated BUN may be due to variation in renal clearance, but

markers of renal function (e.g. plasma BUN/creatinine ratio) must be measured to

investigate this hypothesis.

8. Minimal differences in plasma acylcarnitine profiles

We hypothesized that the elevated plasma concentration of ketone bodies (ß-

hydroxybutyrate) in B6-Chr 6A males may reflect increased triglyceride catabolism in the

fasted state. To investigate possible elevated triglyceride catabolism in B6-Chr 6A, we analyzed the acylcarnitine profile in all three strains (Table IV-7). Carnitine is required for the transport of long chain fatty acid across the mitochondrial membrane for the

179

Table IV-6. Amino acid measurements in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and 62-B congenic strain. An unpaired t-test was used to compare each amino acid in each strain to C57BL/6J (Welch’s t-test was used if the variance differed significantly from C57BL/6J using an F test). Only the significant p values (<0.05) are indicated: * <0.0001, ** 0.0003, ***0.03, ****0.05. Abbreviations: std dev = standard deviation.

C57BL/6J (n=10) AJ (n=9) CSS-6 (n=8) 62-B (n=8)

Amino Acid (µM) Mean +/- std dev Mean +/- std dev Mean +/- std dev Mean +/- std dev

Proline 98.59 +/- 38.05 97.47 +/- 11.02 106.28 +/- 16.91 101.55 +/- 13.93

Taurine 719.17 +/- 87.24 489.61 +/- 87.27* 509.06 +/- 47.81* 663.54 +/- 177.09

Threonine 137.34 +/- 35.83 174.07 +/- 29.20*** 159.55 +/- 16.46 167.69 +/- 34.28

Serine 139.73 +/- 44.27 146.98 +/- 20.21 128.94 +/- 10.98 127.99 +/- 26.25

Asparagine 64.64 +/- 23.06 57.43 +/- 9.66 67.03 +/- 5.52 65.13 +/- 6.97

Glutamic Acid 26.77 +/- 8.98 26.04 +/- 6.06 33.01 +/- 2.85**** 35.38 +/- 10.44****

Glutamine 843.71 +/- 110.41 464.92 +/- 55.12* 803.40 +/- 74.15 770.08 +/- 154.98

Glycine 251.12 +/- 86.99 229.82 +/- 37.77 248.70 +/- 22.64 244.99 +/- 44.43

Alanine 410.15 +/- 110.77 372.36 +/- 43.09 405.93 +/- 49.83 387.35 +/- 70.44

Citrulline 65.88 +/- 13.86 41.93 +/- 4.65** 65.96 +/- 2.97 74.84 +/- 17.57

Valine 149.09 +/- 46.06 149.48 +/- 19.57 136.46 +/- 15.31 146.43 +/- 30.59

Methionine 47.41 +/- 16.54 42.76 +/- 4.70 46.19 +/- 5.61 51.63 +/- 13.05

Isoleucine 67.59 +/- 22.53 62.27 +/- 11.01 57.23 +/- 7.37 62.29 +/- 11.56

Leucine 138.08 +/- 33.38 128.14 +/- 23.18 131.56 +/- 13.13 131.46 +/- 22.82

Tyrosine 64.32 +/- 24.18 62.46 +/- 9.21 62.00 +/- 6.73 56.89 +/- 11.22

Phenylalanine 72.44 +/- 16.83 70.73 +/- 7.91 70.83 +/- 3.11 74.19 +/- 15.37

Ornithine 86.65 +/- 56.62 50.43 +/- 20.50 55.24 +/- 6.23 45.48 +/- 13.71

Lysine 196.70 +/- 95.20 215.66 +/- 33.49 184.46 +/- 17.69 161.81 +/- 36.56

Histidine 64.85 +/- 17.30 59.00 +/- 11.21 57.33 +/- 5.61 61.60 +/- 12.85

Arginine 70.76 +/- 56.49 66.34 +/- 35.66 49.40 +/- 15.55 66.30 +/- 42.63

180

Table IV-7. Acylcarnitine profile in C57BL/6J, A/J, B6-Chr 6A (CSS-6), and 62-B congenic strain. An unpaired t-test was used to compare the traits from each strain to C57BL/6J (Welch’s t-test was used if the variance differed significantly using an F test). Significant p values (<0.05) are indicated.

C57BL/6J (n=10) AJ (n=9) CSS-6 (n=8) 62-B (n=8) Acylcarnitine (chain length) nM Mean +/- std dev Mean +/- std dev Mean +/- std dev Mean +/- std dev

C3 186.6+/- 96.24 155.78 +/- 53.77 164.13 +/- 31.94 139.378 +/- 35.47

C5 198.1 +/- 92.60 161.44 +/- 39.98 170.00 +/- 40.04 169.38 +/- 33.08

C5-1 18.7 +/-6.00 23.44 +/- 6.41 23.63 +/- 5.68 21.38 +/- 9.66

88.38 +/- 15.75 C5-OH 67.2 +/- 15.22 79.33 +/- 17.36 71.75 +/- 14.82 (P=0.01) 251.56 +/- 61.02 C4 387.5 +/- 173.08 342.75 +/- 61.03 340.25 +/- 39.69 (P=0.04)

C6 75.2 +/- 27.44 57.00 +/- 11.74 95.50 +/- 16.85 91.50 +/- 17.55

C8 14.3 +/- 6.48 13.33 +/- 4.12 15.5 +/- 4.69 18.38 +/ -3.29

20.78 +/- 4.97 19.625 +/- 3.85 C10:1 27.6 +/- 3.66 28.38 +/- 5.78 (P=0.003) (P=0.0004) 23.22 +/- 3.11 C10 31 +/- 8.47 31.63 +/- 3.16 31.75 +/- 6.45 (P=0.02)

C12:1 16.7 +/- 3.02 14.33 +/- 2.74 17.38 +/- 3.20 19.13 +/- 2.59

C12 94.2 +/- 23.08 85.22 +/- 6.65 108.88 +/- 11.67 102.38 +/- 10.20

C14:1 68.4 +/- 11.73 73.22 +/- 10.03 72.63 +/- 8.16 78.13 +/- 7.94

C14 171.7 +/- 47.45 199.33 +/- 37.31 177.75 +/- 18.64 198.75 +/- 14.95

26.33 +/- 4.24 25.63 +/- 5.48 Malonyl 31.8 +/- 6.36 31.63 +/- 2.97 (P=0.04) (P=0.05) 10.56 +/- 1.88 Glutaryl 14.2 +/- 3.39 14.00 +/- 3.30 14.63 +/- 4.96 (P=0.01) 147.13 +/- 13.40 C16:1 128.5 +/- 18.93 117.67 +/- 23.45 126.25 +/- 19.91 (P=0.03)

C16 218.4 +/- 32.43 219.22 +/- 25.19 228.88 +/- 19.45 236.13 +/- 17.03

C16-OH 19.4 +/- 2.88 24.00 +/- 7.79 19.25 +/- 5.47 22.25 +/- 3.58

244.25 +/- 23.01 C18:1 210.1 +/- 18.02 213.67 +/- 30.13 222.38 +/- 30.10 (P=0.0027) 77.75 +/- 8.88 77.13 +/- 5.69 C18 66.5 +/- 8.41 75.56 +/- 12.90 (P=0.01) (P=0.008)

C18:1-OH 25.8 +/- 6.03 30.22 +/- 7.58 28.00 +/ 4.17 30.13 +/- 5.44

20.11 +/- 6.49 18.75 +/- 3.77 C18-OH 14.5 +/- 3.60 18.00 +/- 6.14 (P=0.03) (P=0.03)

181

process of beta-oxidation. Acylcarnitine is the fatty acid-bound form of carnitine. Thus,

we hypothesized that elevated levels of acylcarnitines may reflect increased triglyceride

catabolism. Although elevations in several long chain acylcarnitines were observed in

62-B, no other patterns were observed. The metabolism of acylcarnitines is highly

complex, and the significance of this finding is unclear. Further studies of the

concentrations of plasma free fatty acids, glycerol, and total and free carnitine may

provide more clues to the ketone body elevation in B6-Chr 6A.

IV. DISCUSSION

1. B6-Chr 6A males are resistant to diet-induced adiposity

Our previous studies of B6-Chr 6A demonstrated that, in male mice, the substitution of A/J chromosome 6 onto the C57BL/6J background confers decreased body weight and BMI relative to C57BL/6J. In the present work, we confirmed that the decreased body weight and BMI in B6-Chr 6A males is due to a decrease in adiposity by

demonstrating that the gonadal fat pads were significantly smaller in B6-Chr 6A males

relative to C57BL/6J males. Interestingly, the variation in fat pad mass and BMI in the

B6-Chr 6A males suggests that the obesity resistant phenotype is not completely penetrant in this strain. This incomplete penetrance may be due to the complex genetic nature of the phenotype.

A possible explanation for the decreased adiposity in B6-Chr 6A males is

decreased food consumption relative to C57BL/6J. Subtle differences in caloric

consumption were detected in the strains when raw food intake was analyzed, but the

differences in food consumption were not significant when body weight was taken into 182

account. Body weight is taken into account when evaluating food consumption because

ordinarily more calories are required to maintain the basal metabolic rate in heavier

animals. For instance, C57BL/6J male mice require more calories simply to maintain a

higher body weight than B6-Chr 6A males, so the additional calories consumed by

C57BL/6J must contribute to body weight maintenance, weight gain, and basal

metabolism. Consequently, we concluded that food intake is not the likely explanation for the obesity resistance in B6-Chr 6A.

Other possible explanations for the resistance to diet-induced obesity observed in

the B6-Chr 6A male mice include decreased fat absorption, increased energy expenditure

or increased body activity. Further studies to address these possibilities must be pursued

to understand the etiology of the resistance.

2. B6-Chr 6A males are resistant to various aspects of metabolic syndrome

Blood chemistry and liver triglyceride measurements indicated that B6-Chr 6A

males are resistant to several aspects of the metabolic syndrome in addition to obesity.

For instance, B6-Chr 6A males are resistant to hypercholesterolemia, hyperglycemia, and the development of fatty liver relative to C57BL/6J males (summarized in Figure IV-6).

The only surprising feature of the B6-Chr 6A phenotype is hypertriglyceridemia because

hypertriglyceridemia is usually associated with obesity in human populations.

Regardless, the metabolic studies demonstrate that B6-Chr 6A is a model for resistance to

obesity and features of the metabolic syndrome. It is unclear at this stage whether or not

the same gene or genes confers resistance to all of these traits.

183

male mice A = ↓ ↓ ↓ ↑ ↑ ↓

↓↓

B6-Chr 6

fast after analyzed chemistry

Fat pad weight -hydroxybutyrate Plasma glucose Liver triglyceride Body massBody index β Relative to B6 Plasma triglyceride Plasma cholesterol

Food intake/body Food intake/body weight

Blood p<0.001; indicates or

↑↑ or

A ↓↓

. B6 … … … …

6 B6-Chr

p<0.05; indicates ↑

or fed the HFSC diet for 100 days Figure IV-6. Summary of adiposity, food intake, blood chemistry, and liver triglyceride analyses in C57BL/6J (B6) and B6-Chr 6 B6-Chr and (B6) C57BL/6J in analyses triglyceride liver and chemistry, blood intake, food of adiposity, Summary IV-6. Figure ↓ 184

3. Congenic strain is resistant to metabolic syndrome without elevated ketone bodies

Phenotyping studies of the 62-B congenic strain, which harbors at least one QTL that confers resistance to diet-induced obesity, demonstrated that the 62-B males were also resistant to fatty liver, hyperglycemia, and possibly elevated plasma cholesterol

(summarized in Figure IV-7). Furthermore the 62-B males were also hypertriglyceridemic relative to C57BL/6J males. Phenotyping studies in 92-A, a neighboring strain that does not harbor Obrq2, will determine if these traits map to the same region as the 62-B obesity resistance phenotype. Overall, this congenic strain dissected at least one component of the B6-Chr 6A phenotype, the elevated ketone bodies, from the obesity resistance phenotype, which suggests that these traits are associated with different genes. Phenotyping studies in the complete congenic panel are necessary for a more thorough dissection of the components of metabolic syndrome and related traits on chromosome 6.

4. Differences in triglyceride handling may contribute to resistance

Assuming that the phenotypes are due to the same QTL, the hypertriglyceridemia together with the resistance to obesity and the development of fatty liver suggests that differences in triglyceride handling may contribute to the obesity resistance in the 62-B congenic strain. At least two monogenic mouse models with diet-induced obesity resistance and hypertriglyceridemia have been described. In these examples, extreme perturbations in single genes produce severe phenotypes. For instance, the caveolin-1 knock-out mouse model exhibits severe hypertriglyceridemia and obesity resistance because this mouse has a defect in the clearance of dietary triglyceride and in adipocyte 185

e mice fed the the mice fed e = = ↓ ↓ ↓↓ ↑↑ ↓↓

62-B

fast after analyzed chemistry

Body weight Body -hydroxybutyrate Plasma glucose Liver triglyceride Body mass index Body β Relative to B6 Plasma triglyceride Plasma cholesterol

Blood p<0.001; indicates or ↑↑

↓↓

B6 . … … … … 62-B 62-B

p<0.05; indicates ↑ Figure IV-6. Summary of adiposity, food intake, blood chemistry, and liver triglyceride analyses in C57BL/6J (B6) and 62-B mal 62-B and (B6) C57BL/6J in analyses triglyceride liver and chemistry, blood intake, food of adiposity, Summary IV-6. Figure HFSC diet for 100 days 100 for diet HFSC or ↓ 186

droplet formation (RAZANI et al. 2002). Interestingly, the gene encoding caveolin-1 is in

the Obrq2 candidate region. In contrast, the obesity resistance and hypertriglyceridemia

in the acetyl-Coenzyme A carboxylase 2 knock-out mouse results from increased fatty

acid oxidation (ABU-ELHEIGA et al. 2003). Unlike the acetyl-CoA carboxylase 2 knock-

out mouse, the 62-B males do not have elevated ketone bodies relative to C57BL/6J. Of

course, in our model, the hypertriglyceridemia and obesity resistance phenotypes may

each be associated with entirely different genes. Further studies including VLDL

(derived from hepatic triglyceride synthesis) vs. chylomicron (derived from dietary

triglyceride) concentrations, energy expenditure, lipolysis rates, and fatty acid oxidation rates may reveal clues to the source of the elevated triglycerides in the blood and to the mechanism underlying the hypertriglyceridemia.

5. Subtle differences in cholesterol biosynthesis intermediates and plant sterols

Investigations of cholesterol biosynthesis intermediates and plant sterols in the

62-B strain revealed significantly lower levels of many of these metabolites in 62-B relative to C57BL/6J. In general, large differences in the levels of either cholesterol biosynthetic intermediates (KELLEY 2000) or plant sterols (BERGE et al. 2000; LEE et al.

2001) are due to major, single gene perturbations, and it is unclear whether small differences in these metabolites, particularly the plant sterols, have any pathophysiologic consequence (WILUND et al. 2004). In contrast, lathosterol is often used as a marker of cholesterol biosynthesis and the plant sterols can also be used as a marker for cholesterol absorption in the gut (MIETTINEN et al. 1990). The decreased concentrations of the

various cholesterol biosynthetic intermediates may suggest decreased cholesterol 187

biosynthesis in the liver of 62-B and B6-Chr 6A males (or possibly to increased

clearance). In addition, the decreased plant sterols may result from either variation in the

clearance or metabolism of these sterols or to differences in cholesterol absorption.

Further studies are needed to investigate these hypotheses and to test whether differences

in the biosynthesis or absorption of cholesterol contribute to the metabolic syndrome

resistance phenotype.

6. Congenic strains: tools for studying resistance to obesity and metabolic syndrome

Interestingly, like B6-Chr 6A, the 62-B congenic strain was resistant to both diet-

induced obesity and to various features of the metabolic syndrome. Additional studies

are needed to test whether other regions of chromosome 6 that have been associated with obesity resistance also confer resistance to the metabolic syndrome phenotypes. If so, this result would suggest that genes that confer resistance to diet-induced obesity also confer resistance to many metabolic syndrome phenotypes and that these traits share a common etiology. Moreover, similar studies using other CSSs and congenic strains derived from them will indicate whether the results from chromosome 6 reflect obesity resistance conferred by other chromosomes. Thus, similar studies using other CSSs and congenic strains derived from them will lead to a greater understanding of the etiology of metabolic syndrome and obesity.

188

CHAPTER V: DISCUSSION AND FUTURE DIRECTIONS

189

A. DISCUSSION

1. Multiple genes and gene interactions influence obesity resistance

Although many studies have identified QTLs that increase susceptibility to obesity,

few, if any, studies focus on genes that confer resistance to obesity. In recent years,

alleles that confer protection against disease have been discovered to play an important

role in the genetics of human disease. The most well-recognized resistance alleles are in

the gene encoding CCR5, a chemokine receptor. At least two mutations in the CCR5

gene have been discovered to confer resistance to human immunodeficiency virus

infection (LIU et al. 1996; QUILLENT et al. 1998; SAMSON et al. 1996). Resistance alleles

have also been discovered for non-infectious diseases. For example, mutations or

variants in the PCKS9 have been shown to reduce LDL cholesterol and confer protection

from coronary heart disease in both humans and mice (COHEN et al. 2005; COHEN et al.

2006; KOTOWSKI et al. 2006; RASHID et al. 2005). Lastly, modifier genes, which

minimize the severity of Mendelian disease phenotypes, have also been discovered in

humans and mice (NADEAU 2001; NADEAU 2003). Consequently, susceptibility and

resistance alleles probably contribute to risk for many diseases, including obesity. Thus,

genetic studies of resistance to obesity are necessary to identify genes that protect from

obesity, to identify potential therapeutic targets for obesity, and to minimize risk for

obesity-related diseases, such as metabolic syndrome.

Typical human obesity is the consequence of a combination of genetic and

environmental factors. The environmental factors, particularly diet composition and quantity of food consumed, are difficult to control in human genetic studies.

Consequently, animal models, such as inbred mouse strains, provide unique resources for 190

dissecting the genetics of environmentally-induced forms of obesity. In particular, the

high-fat, diet-induced obesity characteristic of C57BL/6J male mice is analogous to typical human obesity, and the obesity resistance in A/J male mice provides an important contrast for genetic studies of this trait. Because of the similarities between human obesity and the obesity in C57BL/6J males, genetic studies involving this strain may reveal genes and pathways that are associated with diet-induced forms of human obesity.

The first estimate of the genetic complexity of obesity in the C57BL/6J and A/J inbred strains was derived from our HFSC diet body weight surveys using B6-ChrA

CSSs. Thirteen B6-ChrA CSSs were reproducibly resistant to obesity when fed the HFSC

diet. Thus, with the detection of at least 13 QTLs, the CSSs detected the three to four

times more QTLs than traditional mapping strategies (ISHIMORI et al. 2004; REED et al.

2003; SINGER et al. 2004; YORK et al. 1996). Moreover, the large weight reduction

conferred by each A/J chromosome indicates that interactions between these genes must

occur in the A/J males. Consequently, the CSS surveys demonstrated that high-fat, diet-

induced obesity resistance involves many genes and that the B6-ChrA CSSs are unique

and important resources for dissecting the genetics of this trait.

Interestingly, the CSS surveys under-estimated the number of genes involved in

resistance because intercrosses derived from the CSSs and a congenic panel derived from

a single CSS, B6-Chr 6A, demonstrated that even single chromosomes may have multiple

genes that influence this trait. Furthermore, the chromosome 6 congenic panel survey

demonstrated that interactions between QTLs on a single chromosome also influence the

genetics of obesity resistance. In addition, we discovered a lack of strong correlations

between weight gain in the early vs. late half of the diet time course and discovered QTLs 191 that appear to influence weight gain at particular time points in the intercrosses. Our

CSS surveys and intercross analyses, the first comprehensive genetic analysis of obesity resistance in mice, revealed that multiple genes, gene interactions, and age or duration of diet exposure contribute to the genetics of obesity resistance in our model. Such a high level of complexity in the CSSs, which are derived from two inbred strains of mice, provokes the question of whether the human genetics of obesity resistance is as complex as the genetics in our model.

Even if the genetic architecture of obesity in human populations differs from mice, the CSSs are a unique resource in which to develop a catalog of genes, pathways, and mechanisms that confer high-fat, diet-induced obesity resistance in mice. These genes, pathways, and mechanisms can then be explored for application in humans. Once the genes underlying the resistance QTLs are discovered, these genes can be analyzed in human populations to test whether genetic variants with similar effects are associated with resistance in humans or whether genetic variants with opposing effects are associated with obesity susceptibility. Furthermore, the identification of genes and pathways associated with resistance will be useful for the development of dietary and therapeutic strategies for treating obesity. Possibly, a greater understanding of the genetics of obesity will lead to individualized dietary or therapeutic strategies for the condition. If individualized approaches to obesity therapy are possible, compliance and efficacy may improve.

192

2. CSSs provide models for studying obesity resistance and related traits

In addition to studying the genetics of obesity resistance, the CSSs provide a unique

resource for studying the mechanisms underlying obesity resistance and for dissecting the

traits that are associated with obesity. For instance, obesity resistance can result from

decreased food intake or increased energy expenditure through metabolic inefficiency or

increased body activity. Known metabolic mechanisms that produce obesity resistance in

single gene knock-out or transgenic models include pathways involving mitochondrial biology (LI et al. 2000; ZHOU et al. 2003), lipid droplet formation (RAZANI et al. 2002),

steroid metabolism (KERSHAW et al. 2005), thyroid hormone (MACDONALD et al. 2005),

and fatty acid oxidation (ABU-ELHEIGA et al. 2003). Studies of the CSSs, which are

polygenic models of resistance, may reveal additional mechanisms and pathways that

contribute to obesity resistance.

Similarly, many traits or pathologies are associated with the obesity phenotype in

human populations, but it is unclear whether these traits share a common cause. Using

chromosome 6 as a model, we demonstrated that these strains provide useful tools for

dissecting obesity and metabolic syndrome traits and for testing whether single QTLs

confer resistance to the entire spectrum of metabolic syndrome or to only a particular

subset of metabolic syndrome phenotypes. Dr. David Sinasac has initiated an important

study that begins to address these questions in the CSS panel. He is investigating

metabolic syndrome trait relationships in the CSSs by analyzing various aspects of

metabolic syndrome (such as plasma glucose, plasma cholesterol, plasma insulin, plasma

triglycerides, and liver triglycerides) in the CSSs fed the high-fat diet. These studies will

test whether all obesity resistant CSSs are also resistant to metabolic syndrome traits, a 193 discovery that would suggest that these conditions may be due to the same QTLs. In contrast, the discovery of CSSs which are resistant to some, but not all, aspects of metabolic syndrome will reveal that this constellation of phenotypes may be the result of multiple QTLs. In the future, the traits analyzed can be extended beyond metabolic syndrome phenotypes to include forms of cancer, inflammation, and other conditions that are associated with obesity in humans.

3. Evaluation of CSS intercrosses as a method for QTL localization

The CSS surveys and intercross analyses of resistance to high-fat, diet-induced obesity are the first, comprehensive genetic analysis of a complex trait using a panel of

CSSs and highlights many advantages of the CSSs. For instance, the ability to reproduce the high-fat diet survey is an advantage of the CSS approach. The replicate surveys utilized mice that were genetically identical to those used in the initial survey. No additional genotyping or crosses were needed to generate mice for these replicate studies.

Likewise, our ability to refine the phenotype of our trait using the low-fat diet demonstrates another advantage of using the CSS strategy. Obviously, in an F2 study, the same mouse population cannot consume two diets simultaneously, but in the CSSs, simultaneous studies of genetically identical populations using various environmental perturbations (such as diet, pharmacologic, etc.) are possible.

Once CSSs with QTLs influencing a trait of interest are identified, the QTLs must be localized. The CSS whole genome scan used in this study was a novel strategy and the first thorough evaluation of the CSS intercross method as the first step towards QTL localization. The intercrosses successfully localized QTLs on five chromosomes and 194

enabled the detection of two QTLs that were undetectable in the CSS survey. If single,

strong QTLs produced the resistance in the CSS surveys, we hypothesized that we should

detect single, strong QTLs in each of the intercrosses derived from CSSs that were

resistant in the surveys. Surprisingly, for many QTLs reproducibly detected in the CSS

surveys, we were unable to detect significant or even suggestive QTLs using the crosses

even though the mean trait values for the F2 populations were less than C57BL/6J. Our

inability to detect QTLs on these chromosomes suggests that multiple QTLs or parental

effects may influence resistance conferred by these chromosomes. Thus, even single

chromosomes may have multiple QTLs influencing body weight.

Our inability to detect QTLs on some chromosomes that conferred resistance may be

the result of our conservative multiple testing corrections. Our analyses demonstrated

that many more QTLs were detected in the crosses when we considered the genome scan

as multiple, independent crosses with less stringent multiple testing penalties. Because

our mapping strategy is novel, it is unclear if the conservative statistical thresholds were

appropriate for our study. Further investigations of these chromosomes as described in

the next section may reveal which strategy is most appropriate. For instance, if the QTLs

detected using fewer multiple testing penalties are true QTLs, they should be detected

using congenic strains. In contrast, if fewer multiple testing corrections produced false

positive results, we would expect that the additional QTLs discovered using this more liberal strategy should be undetectable with other methods. In addition, an increase in

sample size in the intercrosses (particularly if this sample is limited to mice with

recombinant chromosomes) should increase power for detecting multiple, additive or 195

interacting QTLs and may serve as another method for evaluating the two multiple testing approaches for comprehensive CSS intercrosses.

Analyses of congenic strains derived from B6-Chr 6A demonstrated the utility of the

congenic strain strategy for QTL localization and for dissecting a QTL interaction. Using

the congenic strains, we resolved and localized at least three QTLs, including one QTL

interaction. These QTLs, which probably contributed to the LOD curve from the

intercross, could not be precisely resolved in the F2 population. Consequently, the

congenic approach may be better suited for QTL localization and for the dissection of

multiple and interacting QTLs. Currently, congenic panels derived from chromosome 6,

10, and 17 exist and are available to investigators for studying obesity and other traits.

Furthermore, relative to traditional strategies, the use of CSSs reduces the time required

for congenic strain generation from three or four years to approximately one year, so

even new congenic panels are not very time-consuming to generate.

The discovery for the need for larger intercross sample sizes or more powerful

approaches than intercrosses to detect and localize QTLs discovered in CSS surveys is

profound because it suggests that even CSSs are polygenic models of disease.

Consequently, polygenic diseases may be complex at both a whole genome and single

chromosome level, but as discussed in the following section, the CSSs provide the tools

for dissecting these QTLs and eventually identifying the underlying genes.

196

B. FUTURE DIRECTIONS

1. How do we identify obesity resistance genes detected in CSSs?

An ultimate goal of this work is obesity resistance QTL discovery and eventually gene identification. For chromosome 6, we localized three QTLs with congenic strains,

and similar strategies will enable QTL localization on other chromosomes. The

candidate regions associated with these QTLs still contain many genes, and it is unclear

whether single or multiple genes within each region explain the resistance. To refine

these intervals, various methods, including the generation of subcongenic strains and additional phenotyping, may prove useful.

a. Subcongenic strain analysis

The generation and analysis of subcongenic strains is the first step towards gene identification. Subcongenic strains which span the candidate intervals of QTLs will further localize the QTLs and definitively narrow the list of candidate genes.

Subcongenic strains are generated by backcrossing a congenic strain to C57BL/6J and then backcrossing or intercrossing the offspring to C57BL/6J. Then, mice with recombination events in the candidate interval can be selected by genotyping and crossed to C57BL/6J. Mice that inherit identical segments intact are selected by genotyping and intercrossed to homozygose the A/J-derived segment. Once a series of overlapping, homozygous subcongenic strains are constructed, they can be screened on the high-fat diet to test which strains are resistant to obesity. Then, as with the congenic strain analysis described in Chapter 3, the A/J-derived segments shared by resistant strains will define the new candidate interval. 197

b. Phenotyping studies

Additional phenotyping studies may also enable gene identification. For instance,

phenotyping studies should reveal the mechanism underlying the obesity resistance. An

understanding of the mechanism may enable the identification of key tissues involved in

resistance and may assist prioritization of candidate genes. An outline of important

phenotyping analyses is provided in a later section.

Whole genome expression analysis may also assist in candidate gene identification. Microarray studies in key tissues of obesity resistant vs. obese strains may reveal differential expression of genes in the candidate region or differential expression of pathways that may explain the mechanism of resistance. Results from gene expression analyses in these strains can then be compared to data sets in the literature.

For example, gene expression has been investigated in several tissues derived from

C57BL/6J mice fed a high-fat diet (DE FOURMESTRAUX et al. 2004; KIM et al. 2004;

MORAES et al. 2003; SPARKS et al. 2005; VAN SCHOTHORST et al. 2005). Likewise, gene

expression and proteomic investigations of adipocyte differentiation (SOUKAS et al.

2001), mitochondrial biology (MOOTHA et al. 2003), and various organelles (FOSTER et

al. 2006; KISLINGER et al. 2006) have been published.

Phenotyping studies will be most informative if they are performed after the

region has been narrowed (e.g. subcongenic strains or congenic strains). For instance, if

phenotyping differences are detected in a congenic or subcongenic strain relative to

C57BL/6J or an obese control strain, these differences may or may not be due to the same

QTL that confers obesity resistance. Thus, a small candidate interval will increase the 198

likelihood that the additional trait differences discovered (including expression

differences involving genes outside the region) are the result of the same QTL as the

obesity resistance.

c. Sequence analysis

Genes within the candidate region must be analyzed for the presence of

polymorphisms in C57BL/6J vs. A/J genomic sequence. Various sources of sequence

polymorphisms were used in this work, but a comprehensive SNP resource from Perlegen is currently being released as SNPs are discovered. This resource will be invaluable

because genes within the candidate region with missense or nonsense SNPs may be

prioritized for further investigation and may prevent investigators from having to

sequence large candidate regions.

2. How do we identify obesity resistance genes which did not produce peaks in the CSS

intercrosses?

As described in the previous section, our inability to detect QTLs on several

chromosomes on which we expected them may be the result of the statistical methods

used. Other possible explanations include parental effects and multiple, additive or interacting QTLs.

a. Test for parental effects

The inability to detect QTLs in intercrosses involving chromosomes that conferred

resistance in CSS surveys may be the result of parental effects. For instance, in the CSS 199

surveys, the parents of the CSS males are obviously both CSSs, but in the intercrosses, the parents are from the F1 generation. Consequently, the difference in the genetics of

the parents may contribute to the differences observed between the CSS surveys and

intercross analysis. Possible mechanisms by which parental effects may contribute

include imprinting, uterine environment, growth factors, hormonal environment, and

maternal milk composition. To investigate parental effects, F1 males derived from

reciprocal crosses involving each CSS and C57BL/6J will reveal whether parental effects

play a role. For instance, if parental effects produce the phenotype, the offspring from

the reciprocal crosses should exhibit significantly different body weights. In contrast, if

the body weights are similar, parental effects are probably not involved.

If parental effects are observed, the next question is whether these effects are the result of imprinting, intra-uterine environment, or maternal milk composition. To answer this question, the offspring from the reciprocal F1 crosses can be swapped in the perinatal

period. If the offspring exhibit the same phenotype as they did in the original reciprocal

cross study regardless of whether their milk is derived from a CSS or C57BL/6J, the

intra-uterine environment or imprinting rather than milk composition probably explains

the trait difference. Alternatively, if the offspring with foster mothers do not exhibit the

same phenotype as offspring with the biologic mothers, differences in maternal milk

composition may explain the phenotype. If this is the case, the composition of the milk

can be investigated.

To test whether intra-uterine environment or imprinting plays a role, embryos can

be swapped during development. If the phenotype changes depending on the uterine

environment, some component of the prenatal environment contributes to the phenotype. 200

If not, the phenotype may be due to a uterine effect prior to embryo swap or to

imprinting. To investigate imprinting, allele-specific gene expression in known imprinted regions on the chromosome can be tested in reciprocal F1 mice. QTL localization using congenic strains should be performed first to narrow the candidate

region so that imprinted regions only within the candidate region are investigated.

Imprinting regions can be identified through literature searches and from a recently

published analysis that predicts genome-wide imprinting regions in mice (LUEDI et al.

2005). Metabolic phenotyping studies can be performed to refine the phenotype and

prioritize tissues for such studies.

Likewise, similar studies could be performed by swapping embryos or newborn

pups between C57BL/6J and obesity resistant CSSs to test for possible intra-uterine

effects or the influence of varied milk composition. This type of analysis may be

important if parental effects interact with a recessive offspring genotype.

b. Test for multiple genes or gene interactions

To test whether multiple genes or gene interactions may explain why some QTLs

detected in the CSS surveys were not detected in the crosses, increased sample sizes in

the F2 populations or congenic strain approaches can be used. An increased sample size

in the F2 may increase power to detect multiple, small effect QTLs or interacting QTLs.

The number of mice required for the study will depend on the effect size of the QTL (e.g. smaller effect sizes will require more mice) or the spacing of interacting QTLs. Thus, it is impossible to predict the sample size needed prior to initiating the study. As an initial step, the sample size could be doubled to test whether stronger QTLs are detected. 201

Furthermore, increased sample size and thus, increased recombination, will increase

power to detect interacting QTLs using methods described in Appendix III. To increase

yield from the larger intercross sample, only mice with recombination events could be

used.

Although increased F2 samples sizes may prove useful, a better approach may be

congenic strain generation and analysis as described in Chapter 3. The congenic strain

approach used for chromosome 6 was successful at dissecting three QTLs including one

QTL interaction. If a congenic strain strategy still does not detect QTLs on these other

chromosomes, double congenic strains or congenic crosses can be used. First, double

congenic strains can be generated using crosses with two non-overlapping congenic

strains. In addition, congenic crosses may prove useful for identifying regions of the

chromosome that are involved in an interaction. For instance, a region of the

chromosome with a known QTL can be fixed (for instance, by crossing the congenic

strain to the CSS and then intercrossing the offspring) while the remainder of the

chromosome is allowed to segregate. This method is currently being used to investigate

interactions on chromosome 6. F2 progeny derived from 62-BL and B6-Chr 6A have been collected and are being studied to search for suppressors of the obesity resistance conferred by Obqr2. Similar studies could be performed to investigate Obqr1 and chromosomes on which QTLs were expected but not detected.

202

3. What mechanism explains the resistance to diet-induced obesity in CSSs and congenic

strains?

To develop therapeutic strategies for treating obesity, mechanisms that promote

obesity resistance must be well-understood. The CSSs obviously provide several models of obesity resistance which likely represent a variety of mechanisms. Consequently, physiologic mechanisms of obesity resistance can be evaluated in all resistant strains. In addition, phenotyping studies in congenic or subcongenic strains may prove useful for prioritizing the list of candidate genes.

Although the BMI, the body weight, and the high-fat diet effects provide evidence indicating that decreased adiposity explains the resistance in most CSSs, studies of adipose deposition in each resistant CSS will definitively prove whether decreased adiposity explains the body weight differences. Adiposity can be evaluated by measuring fat pad size, using adipose imaging modalities (DEXA, MRI, etc), or by using stable isotopes to measure body composition. If decreased adiposity rather than decreased lean mass is conclusively identified as the cause of the decreased body weight in each CSS, further studies to investigate the mechanism of obesity resistance can be pursued.

Differences in energy intake or energy expenditure probably explain the resistance to high-fat, diet-induced obesity (Figure V-I). To investigate energy intake, food consumption analyses can be pursued at several time points during the diet survey. Even if strains consume similar quantities of food, the absorption of food may vary.

Consequently, fecal fat quantification or oral fat tolerance tests are also important.

If no food consumption or absorption differences are detected, variation in energy expenditure may explain the resistance. Energy expenditure can be measured using the 203

doubly labeled water method (BRUNENGRABER et al. 2002; KLEIN et al. 1984;

SCHOELLER et al. 1986; WESTERTERP et al. 1986) or with calorimetry. Energy

expenditure differences may result from variations in body activity, which can be

assessed with activity meters or telemetry devices, or in the metabolic handling of fuel.

If differences in the metabolic handling of fuel are suspected, various studies can be

pursued. For example, body temperature analyses using the cold challenge may test whether resistant mice dissipate more energy in the form of heat production as a result of metabolic inefficiency (e.g. uncoupling). Likewise, gene expression or protein studies of

UCP-1, UCP-2, and UCP-3 may provide evidence indicating that uncoupling explains the resistance. Lastly, differences in rates of lipolysis, in concentrations of free fatty acids or ketone bodies during a fast, or in gene expression or protein levels of fatty acid oxidation may provide clues suggesting that variation in fat utilization, triglyceride break-down, or fatty acid oxidation explains the resistance.

204

Figure V-1. Methods for investigating the mechanism of obesity resistance.

Why are the mice obesity resistant?

Decreased energy intake (both studies needed): Increased energy expenditure: 1. Food consumption (food intake studies) 1. Calorimetry or

2. Fat absorption (fecal fat analysis) 2. Doubly labelled water method

Differences would suggest that energy expenditure Differences in either test would suggest that energy intake may explain the resistance may explain the resistance

Energy expenditure differences could be the result of variation in body activity or metabolic rate

Body activity can be measured using: 1. Telemetry devices or 2. Activity meters

If no body activity differences are detected, variation in fuel utilization and metabolic efficiency can be investigated: 1. Cold challenge tests 2. Gene expression or protein levels of UCPs and fatty acid oxidation proteins in brown and white adipose, liver, and skeletal muscle 3. Lipolysis rates 4. Concentrations of free fatty acids, ketone bodies, etc. 5. Other possibilities

205

APPENDIX I: GENETIC MARKERS USED IN CSS INTERCROSS MAPPING STUDIES

206

Genetic markers for F2 crosses. The genetic markers used in the F2 genome scan analysis are listed. In addition, for each marker, the genotype frequencies, the physical location of each marker, and the p values derived from chi square tests are provided. All locations are from the build 33 of the mouse genome (http://www.ncbi.nlm.nih.gov/genome/guide/mouse). Abbreviations: NA = failed genotype, BB = homozygous C57BL/6J alleles, AA = homozygous A/J alleles, BA = heterozygous.

Chr1 Location in bp (2004) NA BB BA AA P value m1 35094165 9 16 44 23 0.48 m2 50597220 7 21 41 23 0.90 m3 69718370 7 18 39 28 0.23 m4 91069033 13 25 28 26 0.035 m5 114007779 2 24 39 27 0.41 m6 135002221 4 23 48 17 0.46 m7 159898156 13 25 35 19 0.38 m8 177278233 1 26 50 15 0.17

Chr2 Location in bp (2004) NA BB BA AA P value m1 3128813 6 20 44 19 0.85 m2 19841432 4 22 44 19 0.85 m3 37617104 8 19 44 18 0.73 m4 51087174 9 20 40 20 1.0 m5 69044186 8 21 40 20 0.98 m6 94517447 17 20 30 22 0.35 m8 123007297 0 22 47 20 0.83 m9 127734668 0 22 46 21 0.94 m10 143676514 0 25 46 18 0.55 m11 153615847 7 24 42 16 0.45 m12 169377735 10 24 42 13 0.18 m13 173671254 3 29 39 18 0.17

Chr3 Location in bp (2004) NA BB BA AA P value m1 8225928 6 28 44 12 0.043 m2 16834103 10 24 45 11 0.065 m3 37780712 6 29 45 10 0.011 m4 44019288 7 30 41 12 0.020 m5 52983683 6 27 44 13 0.088 m6 65931210 6 27 42 15 0.18 m7 76964718 9 27 39 15 0.16 m8 88004314 11 25 39 15 0.28 m9 98954943 6 28 41 15 0.13 m10 116104613 6 26 40 18 0.42 m11 127999894 8 26 35 21 0.31 m12 139974556 9 24 36 21 0.54 m13 151076160 6 23 39 22 0.80 m14 154924980 6 23 38 23 0.68

207

Chr4 Location in bp (2004) NA BB BA AA P value m1 10707511 1 22 47 21 0.90 m2 20935977 3 19 51 18 0.32 m3 34378374 2 23 44 22 0.98 m4 44951705 3 24 43 21 0.88 m5 61404453 1 22 50 18 0.48 m7 89680838 1 22 47 21 0.90 m8 99014671 4 19 46 22 0.78 m9 112735224 1 26 46 18 0.48 m11 152746734 2 28 44 17 0.26 m12 153413672 4 28 41 18 0.27

Chr5 Location in bp (2004) NA BB BA AA P value m2 3960568 4 15 42 31 0.050 m1 6717056 6 16 42 28 0.18 m3 26871837 3 20 40 29 0.26 m4 42981971 3 17 50 22 0.38 m5 52515299 4 17 49 22 0.43 m6 65104536 3 17 52 20 0.26 m7 75748496 5 19 49 19 0.50 m8 76181989 6 20 48 18 0.53 m9 97381748 3 24 50 15 0.20 m10 103248599 4 23 50 15 0.21 m11 121516895 3 27 47 15 0.17 m12 127118903 6 27 47 21 0.050 m13 144894798 11 18 42 21 0.85 m14 147388532 6 24 44 18 0.64

Chr6 Location in bp (2004) NA BB BA AA P value m1 3790155 2 23 47 21 0.91 m2 (D6Mit138 ) 4206900 1 23 47 22 0.97 m3 23971862 2 21 49 21 0.76 m4 (D6Mit159 ) 29621200 1 21 50 21 0.71 m5 36299590 3 22 48 20 0.78 m6 (D6Mit223 ) 45234635 3 20 50 20 0.57 m8 47205158 3 20 49 21 0.69 m9 (D6Mit274 ) 48637500 1 22 50 20 0.68 m10 51785906 3 21 47 22 0.9 m11 54131745 3 20 49 21 0.69 m12 54488587 3 20 48 22 0.78 m13 (D6Mit384 ) 55161700 1 20 50 22 0.68 m14 73555817 5 18 53 17 0.16 m15 (D6Mit188 ) 75683000 1 18 56 18 0.11 m16 (D6Mit391 ) 85526100 1 19 52 19 0.44 m17 87127482 1 20 51 20 0.57 m18 (D6Mit284 ) 92962700 1 20 51 20 0.57 m19 (D6Mit36 ) 104884300 1 19 47 19 0.57 m20 107020358 2 19 46 19 0.58 m21 (D6Mit287 ) 112512600 1 18 49 18 0.48 m22 (D6Mit254 ) 125974300 1 17 51 17 0.34 m23 128402933 3 15 51 15 0.18 m24 137657411 2 16 48 16 0.23 m26 (D6Mit15 ) 146549700 1 17 42 17 0.044 m27 146788578 1 20 39 20 0.055 m28 147857079 2 18 41 18 0.074 208

Chr7 Location in bp (2004) NA BB BA AA P value m1 4916555 1 21 40 26 0.57 m2 9356195 0 22 41 25 0.74 m3 (D7Mit117 ) 18904700 0 21 45 22 0.97 m4 19393930 2 21 44 21 0.98 m5 33671286 0 23 45 20 0.88 m6 (D7Mit145 ) 38861000 0 24 42 22 0.87 m7 51665183 2 24 41 21 0.82 m8 58692545 7 25 37 19 0.47 m9 (D7Mit318 ) 60993300 0 26 42 20 0.61 m10 67366561 2 24 43 19 0.75 m11 74405248 1 28 39 20 0.30 m12 (D7Mit301 ) 78817100 2 28 41 17 0.22 m13 81292882 0 27 42 19 0.44 m14 88809220 0 26 44 18 0.48 m15 (D6Mit238 ) 106390800 0 26 43 19 0.56 m16 (D7Mit245 ) 112387700 2 25 42 19 0.64 m17 122140285 2 23 39 24 0.68 m18 127227445 2 17 46 23 0.53

Chr8 Location in bp (2004) NA BB BA AA P value m1 7689226 8 24 47 13 0.13 m2 14730798 3 27 46 16 0.24 m3 19022011 2 29 44 17 0.20 m4 30061096 3 30 39 20 0.16 m5 42356024 3 30 39 20 0.16 m6 54584799 4 30 40 18 0.14 m7 67280757 4 32 38 18 0.048 m8 68092380 6 31 38 17 0.057 m9 80047620 3 30 41 18 0.15 m10 105950853 3 26 49 14 0.13

Chr9 Location in bp (2004) NA BB BA AA P value m1 6171173 17 20 38 16 0.78 m2 13295479 7 17 36 15 0.84 m3 21389209 4 17 39 15 0.67 m4 30738777 0 18 41 16 0.68 m5 43262862 6 16 38 15 0.69 m6 45902696 6 14 41 14 0.29 m7 56001518 8 18 34 15 0.87 m8 67670008 3 20 36 16 0.80 m9 79330819 1 20 38 16 0.78 m10 91150672 18 16 25 16 0.65 m12 102830828 2 18 37 18 0.99 m13 113069561 4 20 30 21 0.42 m14 117119728 5 17 33 20 0.78 m15 123255864 1 19 35 20 0.89 209

Chr10 Location in bp (2004) NA BB BA AA P value m1 3052687 1 23 46 19 0.76 m2 15437618 0 26 44 19 0.57 m3 27219779 0 26 41 22 0.63 m4 45781665 0 26 42 21 0.66 m5 58149652 0 27 40 22 0.48 m6 67964071 0 23 44 22 0.98 m7 82650425 0 22 44 23 0.98 m8 94107537 0 22 43 24 0.91 m9 108436452 3 19 42 25 0.64 m10 114001997 2 20 41 26 0.57 m11 128973001 2 30 49 8 0.0019 m12 129459263 1 32 48 8 0.0010

Chr11 Location in bp (2004) NA BB BA AA P value m1 19457791 5 18 41 22 0.82 m2 36341041 3 20 40 23 0.85 m3 47650415 4 17 41 24 0.55 m4 60566202 4 21 39 22 0.90 m5 74187295 3 17 46 20 0.55 m6 87585530 4 17 48 17 0.30 m7 102994859 3 20 45 18 0.71 m8 118014838 1 21 44 20 0.94

Chr12 Location in bp (2004) NA BB BA AA P value m1 9550011 3 15 39 32 0.024 m2 20022437 4 17 34 34 0.0061 m3 24036240 2 21 32 34 0.0069 m4 28405425 4 21 31 33 0.0081 m5 38149164 2 24 34 29 0.094 m6 52054989 7 22 34 26 0.25 m7 62010273 2 18 40 29 0.19 m8 72080415 1 21 42 25 0.76 m9 83292957 0 22 42 25 0.79 m10 94576281 4 28 37 20 0.23 m11 102161478 3 30 40 16 0.083 m12 111241851 5 32 39 13 0.011

Chr13 Location in bp (2004) NA BB BA AA P value m1 16612389 5 23 32 20 0.40 m2 30200200 1 23 36 20 0.65 m3 30647841 0 23 36 21 0.64 m4 44995693 3 21 32 24 0.30 m5 46041913 2 20 35 23 0.59 m6 59560600 4 20 36 20 0.90 m7 65710544 1 20 38 21 0.93 m8 73056119 4 17 38 21 0.81 m9 86587619 2 15 43 20 0.48 m10 113675568 0 14 49 17 0.11 m11 116181561 2 14 46 18 0.23 210

Chr14 Location in bp (2004) NA BB BA AA P value m1 11381643 1 15 45 29 0.11 m2 21913498 6 14 49 21 0.17 m3 26695690 1 14 46 29 0.076 m4 44931499 2 12 51 25 0.048 m5 70763611 1 15 57 17 0.029 m6 79098395 3 15 58 14 0.0079 m7 91780378 1 18 54 17 0.13 m8 103985865 3 18 50 19 0.37 m9 111354233 1 21 47 21 0.87 m10 114380983 3 19 52 16 0.17

Chr15 Location in bp (2004) NA BB BA AA P value m1 10378181 0 21 40 22 0.94 m2 19173040 5 21 36 21 0.79 m3 25082949 3 23 41 16 0.53 m4 42601022 0 23 39 21 0.82 m5 54473665 1 22 39 21 0.90 m6 67996429 1 23 37 22 0.67 m7 80709371 0 22 40 21 0.94 m8 91709857 0 21 41 21 0.99

Chr16 Location in bp (2004) NA BB BA AA P value m1 7018645 0 19 44 28 0.39 m2 35660153 1 20 43 27 0.53 m3 44373006 0 21 43 27 0.59 m4 56618639 1 18 46 26 0.48 m5 74387994 0 19 45 27 0.49

Chr17 Location in bp (2004) NA BB BA AA P value m1 16218065 6 17 40 27 0.28 m2 18750525 13 15 36 26 0.18 m3 34118550 5 16 41 28 0.17 m4 34143218 18 15 33 24 0.25 m5 49624943 13 12 41 24 0.13 m6 64248781 5 13 50 22 0.10 m7 65220567 21 12 37 20 0.33 m8 78006312 14 13 44 19 0.24 m9 86210376 5 20 42 23 0.89 m10 91847343 20 20 30 20 0.49

211

Chr18 Location in bp (2004) NA BB BA AA P value m1 3683904 2 26 50 14 0.12 m2 15613577 9 19 48 16 0.32 m3 33014830 0 23 49 20 0.75 m4 44899534 13 20 41 18 0.90 m5 62154186 1 26 46 19 0.58 m6 75426251 9 20 46 17 0.55 m7 88763180 2 23 49 18 0.53

Chr19 Location in bp (2004) NA BB BA AA P value m1 8994305 2 22 45 19 0.82 m2 27167099 1 21 56 10 0.0069 m3 37145091 1 24 53 10 0.013 m4 44868391 1 25 50 12 0.054 m5 54991668 1 24 52 11 0.027 m6 57861953 1 24 48 15 0.25

ChrX Location in bp (2004) NA BB BA AA P value m1 3990232 0 42 - 47 0.59 m2 40760823 0 41 - 48 0.46 m3 44405254 0 43 - 46 0.75 m4 62443623 0 49 - 40 0.34 m5 76720550 0 50 - 39 0.24 m6 97504418 1 50 - 38 0.20 m7 120300412 0 52 - 37 0.11 m8 133229757 0 50 - 39 0.24 m9 145632366 1 51 - 37 0.14 m11 156142219 1 50 - 38 0.20 m12 156864921 1 50 - 38 0.20

212

APPENDIX II: CORRELATIONS AMONG TRAITS IN CSS INTERCROSS PROGENY

213

Correlation matrices for intercrosses. This is the graphical version of table II-7. The Pearson’s correlation coefficients for all pairs of traits in the CSS intercrosses were calculated and plotted. The chromosome segregating is indicated for each graphic. Analyses of trait correlations in CSSs and in segregating populations derived from them are a future direction of this work.

Chromosome 1

214

Chromosome 2

215

Chromosome 3

216

Chromosome 4

217

Chromosome 5

218

Chromosome 6

219

Chromosome 7

220

Chromosome 8

221

Chromosome 9

222

Chromosome 10

223

Chromosome 11

224

Chromosome 12

225

Chromosome 13

226

Chromosome 14

227

Chromosome 15

228

Chromosome 16

229

Chromosome 17

230

Chromosome 18

231

Chromosome 19

232

Chromosome X

233

APPENDIX III: MULTIPLE QTL ANALYSIS IN CSS INTERCROSS PROGENY

The F2 analysis was performed as described in Chapter II. Dr. Karl Broman is responsible for the analysis described in this section and wrote the methods section.

234

A. INTRODUCTION

Although 13 obesity resistance QTLs were detected in the CSS HFSC surveys,

obesity resistance QTLs were detected on only three of these chromosomes in the CSS

intercross analysis. Because so few of the expected QTLs were detected using traditional

QTL mapping methods, we used a two-dimensional QTL mapping strategy to test

whether multiple QTLs, in particular, multiple interacting QTLs, may explain the

resistance. The results of this analysis, which is still a work in progress, are presented.

B. METHODS

Mouse husbandry, genotyping, etc.: See Chapter II Methods

F2 crosses two-QTL analysis: For each chromosome, we performed a two-dimensional, two-QTL scan of the chromosome. For each pair of putative QTLs, we fit an additive model (two QTLs, acting additively) and a full model (two QTLs, allowed to interact).

From these results, we calculate four LOD scores: Lf, comparing the full, two-QTL

model to the null model (of no QTLs); Li, comparing the full model to the additive two-

QTL model (concerning the interaction term); Lci, comparing the full model to the best

single-QTL model for that chromosome; and Lca, comparing the additive model to the best single-QTL model. Lca and Lci indicate evidence for a second QTL on that

chromosome; Li indicates evidence for an interaction between the QTLs. Two QTLs on a

chromosome are indicated when Lf is large and either Lca or Lci are large. The QTLs are

indicated to interact when Li is also large. The analysis was performed using R/QTL

software (BROMAN et al. 2003). The methods for determining statistical thresholds and 235 significant p values are still being explored. Thus, the top LOD scores are presented for each chromosome with potentially interesting findings highlighted.

C. RESULTS

Because we were unable to detect many of the A/J-derived resistance QTLs that we expected in the intercrosses, we suspected that gene interactions may contribute to obesity resistance in this model. To test this hypothesis, we investigated interactions on single chromosomes in each of the F2 crosses. To this end, we performed two- dimensional, two-QTL scans, which tested whether interactions between markers along single chromosomes influence any of the seven traits. The results of the two-dimensional analysis provided preliminary evidence for interactions among QTLs on at least two chromosomes because these chromosomes had LOD scores that were higher relative to the other chromosomes (Figure AIII-1; Table AIII-1). On chromosome 4, there is a potential interaction between two QTLs (near markers m3 and m11) for FW because FW for mice with a heterozygous genotype at m3 appears to vary with genotype at m11. The possibility of epistasis on chromosome 4 may explain why resistance QTLs were undetectable in the F2 cross. Likewise, on chromosome 13, there is evidence for two, additive QTLs, near m6 and m9. Evidence for effects involving many markers on chromosome 6 were detected, and the congenic strain survey previously discussed demonstrated that at least one gene interaction exists on A/J-derived chromosome 6.

Lastly, for several chromosomes, the spacing of the markers that were detected to show strong effects is very small. With closely spaced markers, these effects are likely artifacts because very few mice will have recombination between closely spaced markers. 236

Table AIII-1. Two dimensional genome scan in B6-ChrA CSS F2 crosses. The LOD scores from the two dimensional scan are presented and should be evaluated as described in the methods section. The markers involved are provided for the chromosomes with potentially interesting findings. Marker locations are provided in Appendix I. On chromosomes 4 (WG only), 8, 10, and 11, the markers involved are closely spaced, and since so few recombination events occur between closely spaced markers, the findings are explained by one or a few outliers and thus, are likely artifacts.

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 1.41 3.2 3.9 1.8 2 1.3 2.96 4.17 1.66 3 3.08 4.19 5.28 1.11 4 2.11 3.03 5.45 0.91 5 1.09 2.39 3.38 1.3 6 0.74 3.36 5.71 2.62 7 1.21 2.29 4.16 1.08 8 1.2 3.01 5.15 1.81 9 0.86 1.83 2.14 0.97 10 2.06 3.76 6.87 1.7 11 2.36 2.86 3.58 0.51 12 1.03 1.96 3.64 0.93 13 1.35 3.23 5.04 1.88 14 0.54 1.09 2.04 0.55 15 0.78 2.24 4.14 1.46 16 0.92 1.26 2.37 0.34 17 2.65 3.96 5.9 1.31 18 0.99 2.97 3.75 1.98 19 0.79 1.48 2.33 0.68 X 1.06 1.44 1.71 0.38 MW

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 0.56 1.69 5.26 1.14 2 1.07 2.67 3.51 1.6 3 0.51 1.2 2.23 0.68 4 1.54 2.96 4.33 1.42 5 0.76 2.28 4.02 1.52 6 1.14 3.06 6.35 1.92 7 0.98 3.02 3.68 2.04 8 4.94 7.67 8.15 2.73 m8 m9 9 0.59 2.34 2.81 1.75 10 0.73 2.01 7.8 1.28 11 1.83 3.17 5.44 1.34 12 0.76 2.55 4.84 1.79 13 3.38 3.51 8.36 0.13 m6 m9 14 1.13 1.66 2.05 0.53 15 1.63 2.05 3.54 0.43 16 0.55 1.04 2.08 0.48 17 0.67 2.35 5.82 1.67 18 0.84 1.76 4.17 0.93 19 1.07 1.3 3.06 0.24 X 0.44 2.16 3.09 1.72 237

Table AIII-1 continued.

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 1.01 2.85 7.32 1.85 2 0.91 3.36 4.06 2.45 3 0.8 1.39 2.85 0.59 4 1.55 3.96 6.27 2.41 m3 m11 5 0.94 1.9 4.16 0.96 6 2.47 4.07 7.85 1.6 many? 7 0.91 1.92 2.76 1.01 8 3.84 7.41 7.69 3.56 m8 m9 9 0.54 2.23 2.79 1.7 10 1.41 3.33 6.55 1.92 11 1.56 1.86 5.88 0.3 12 0.57 2.66 4.28 2.1 13 2.99 3.28 7.42 0.29 m6 m9 14 0.98 1.5 1.98 0.51 15 0.85 1.75 2.74 0.9 16 0.82 1.3 2.29 0.48 17 0.62 1.97 4.68 1.35 18 1.22 2.29 4.29 1.07 19 1.12 1.76 3.97 0.65 X 0.94 2.27 3.33 1.33

BMI

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 0.69 2.27 7.15 1.58 2 0.63 3.27 4.33 2.64 3 1.76 3.27 4.54 1.51 4 1.04 3.24 5.68 2.21 5 0.73 1.54 3.72 0.82 6 1.45 2.53 7.56 1.09 7 2.43 3.67 5.65 1.24 8 2.25 5.38 5.69 3.13 9 0.7 2.85 3.58 2.15 10 2.45 3.9 6.54 1.45 11 1.67 2.13 6.05 0.46 12 0.69 2.59 4.02 1.9 13 2.25 3.88 7.14 1.63 14 1.04 1.51 2.12 0.47 15 0.92 2.09 3.21 1.17 16 1.07 1.21 2.18 0.14 17 0.62 1.81 3.79 1.19 18 1.09 2.5 4.39 1.41 19 1.51 2.17 4.26 0.66 X 0.69 1.49 2.55 0.8

238

Table AIII-I continued.

EWG

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 0.65 2.77 5.66 2.12 2 1.52 2.8 4.84 1.28 3 3.01 3.93 5.71 0.92 4 1.33 4.89 5.62 3.56 5 0.77 2.25 4.12 1.47 6 1.59 2.93 5.1 1.35 7 1.57 2.3 3.94 0.74 8 2.18 6.54 7.9 4.36 m8 m9 9 0.93 2.01 2.5 1.08 10 1.26 3.32 10.12 2.06 m10 m11 11 2.32 3.17 5.76 0.85 12 1.04 2.3 4.18 1.26 13 2.71 5.73 8.82 3.01 14 1.13 1.38 1.68 0.25 15 0.83 1.41 3.15 0.58 16 0.19 1.12 1.87 0.93 17 1.09 2.63 4.91 1.54 18 1.49 2.88 5.43 1.39 19 0.84 1.49 2.86 0.66 X 0.3 1.81 2.43 1.5

FWG

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 1.71 2.6 4.83 0.89 2 1.01 3.15 4.23 2.15 3 0.88 2.68 4.36 1.8 4 1.28 2.39 4.32 1.11 5 1.71 3.12 5.99 1.41 6 3.87 4.63 8.94 0.77 many? 7 1.18 2.16 4.45 0.98 8 1.51 3.52 3.88 2.01 9 1.04 2.01 3.21 0.97 10 1.11 1.94 3.11 0.83 11 1.05 1.99 6.12 0.95 12 0.89 2.98 4.04 2.09 13 0.96 1.5 4.75 0.55 14 1.37 2.45 3.5 1.08 15 1.47 2.7 3.07 1.22 16 1.77 1.95 2.93 0.17 17 0.72 1.5 2.17 0.78 18 1.31 1.97 3.91 0.66 19 0.55 2.95 4.91 2.39 X 2.01 3.09 4.07 1.08

239

Table AIII-1 continued

Chromosome Lca Lci Lf Li Marker 1 Marker2 1 0.56 3.13 7.79 2.58 2 1.12 3.81 5.07 2.69 3 0.35 1.41 3.68 1.06 4 2.11 6.56 7.68 4.45 m7 m9 5 0.66 2.33 4.33 1.67 6 2.39 3.85 8 1.46 many? 7 1.15 2.36 5.18 1.22 8 2.65 6.44 7.12 3.79 m8 m9 9 0.6 1.87 2.79 1.27 10 0.61 1.97 5.48 1.36 11 2.24 2.42 7 0.18 m7 m8 12 1.03 3.17 4.55 2.13 13 1.99 3.01 6.4 1.03 14 1.24 1.96 2.48 0.72 15 0.54 1.87 2.86 1.33 16 0.85 1.52 2.16 0.68 17 0.3 1.32 3.25 1.02 18 1.8 2.35 3.99 0.55 19 0.67 1.39 3.71 0.72 X 0.56 2.07 3.04 1.51

240

Figure AIII-1. Two dimensional genome scan in the B6-ChrA F2 crosses. The trait values for mice are plotted based on their genotypes at two markers along the chromosome. These data provide evidence for multipl eQTLs, but further studies are needed for confirmation. Abbreviations: B/B = homozygous C57BL/6J genotype, A/A = homozygous A/J genotype, and B/A = heterozygous genotype. Marker locations are provided in Appendix I.

Final weight (grams) 25

15 BB-BB BB-BA BB-AA BA-BB BA-BA BA-AA AA-BB AA-BA AA-AA Chromosome 4: Genotype at m3 and m11

Mid-point body weight (grams) 20

BB-BB BB-BA BB-AA BA-BB BA-BA BA-AA AA-BB AA-BA AA-AA Chromosome 13: Genotypes for m6 and m9 50

Final weight body (grams) 20

BB-BB BB-BA BB-AA BA-BB BA-BA BA-AA AA-BB AA-BA AA-AA Chromosome 13: Genotypes for m6 and m9

241

D. DISCUSSION AND FUTURE DIRECTIONS

Overall, the possibility of multiple QTLs on individual chromosomes must be investigated further because the results are limited by the small number of F2 progeny and the limited recombination in F2 crosses. Furthermore, the significance thresholds have not yet been determined. Regardless, these results provide evidence for multiple

QTLs on single chromosomes, in particular 4, 6, and 13.

To further evaluate these results, our collaborator, Dr. Karl Broman, is exploring methods to establish significance thresholds. Furthermore, an increase in sample size may provide more power for discovering interacting QTLs or QTLs with additive effects.

Lastly, the congenic strains derived from chromosome 6 confirmed the existence of multiple QTLs on the chromosome and suggest that HFSC diet surveys of congenic strains derived from B6-Chr 4A and B6-Chr 13A may prove useful for testing whether multiple QTLs exist on these chromosomes.

242

APPENDIX IV: GENETIC AND PHENOTYPIC ANALYSES OF OBRQ2

The work described in this section was performed by the candidate. Dr. Gary Churchill and Dr. Keith Shockley assisted in experimental design and planned analysis of expression arrays.

243

A. INTRODUCTION

To further characterize Obrq2, an obesity resistant QTL on A/J-derived

chromosome 6, a series of phenotypic, genetic, and gene expression studies was initiated

using 62-BL, a congenic strain that has Obrq2, and 92-A, an obese control strain (Figure

III-6). This series of work, which begins to address several of the questions posed in

Chapter 5, has only just begun and is presented as a work in progress in combination with

future directions.

In the previous chapters, we demonstrated that 62-BL, a congenic strain with

Obrq2, is resistant to diet-induced obesity and hyperglycemia and possibly to elevated

plasma cholesterol. In the present study, we demonstrate that the obesity resistance

associated with Obrq2 is specific to the HFSC diet and that decreased HFSC diet

consumption does not explain the resistance. Then, we demonstrate that the resistance to

high-fat, diet-induced hypoglycemia also maps to the Obrq2 region because 62-BL is

hypoglycemic relative to 92-A. To further localize Obrq2 and possibly identify candidate genes and pathways, gene expression and subcongenic strain analyses were initiated. Lastly, the chromosome 6 HFSC diet congenic strain survey indicated that at least one suppressor of Obrq2 exists on the chromosome. To test whether additional enhancers or suppressors of Obrq2 are present on chromosome 6, a mapping study involving intercross progeny derived from 62-BL and B6-Chr 6A was initiated.

B. MATERIALS AND METHODS

1. Mouse husbandry: 92-A and 62-BL colonies are maintained at CWRU. For the 62-

BL interaction cross, B6-Chr 6A female mice were crossed with 62-BL male mice to 244

generate F1 mice. The F1 mice were intercrossed and the F2 progeny (n~100) were

collected. All mice were weaned at ~3-4 weeks of age, housed in groups of 2-4, and fed

LabDiet 5010 autoclavable rodent diet (LabDiets, Richmond, IN) ad libitum until diet studies were initiated. All mice were housed in microisolator cages in a 12:12 light:dark

cycle.

2. Diet studies: For the diet studies, the HFSC diet (Research Diets, New Brunswick,

NJ, composition in Table II-1) or LabDiet 5010 autoclavable rodent diet was used. All

mice were introduced to the diet at 35 days of age and weighed every two weeks for the

duration of the study. The composition of the 5010 diet is in Table AIV-1.

3. Food consumption studies: Food consumption studies were initially performed as

described in Chapter 4. In chapter 4, the singly housed C57BL/6J, A/J, and B6-Chr 6A mice had similar trends in body weight as group housed mice, and thus, food intake in singly housed mice could be used to assess food consumption in these strains. In contrast, in the present studies involving 92-A and 62-BL, the expected weight differences were not observed in singly housed mice, so food consumption studies were repeated in a group setting. The smaller weight difference between 92-A and 62-BL congenic strains or differences in the stress response in these strains (relative to the strains used in the previous studies) may explain our inability to detect weight differences in these singly housed mice. For the group studies (n=8-9 cages per strain), food consumption studies were performed after approximately 80-90 days of HFSC diet consumption. Between two and four mice were placed in a clean cage and a known amount of food was provided 245

on day one. Then, 24 hours later, the remaining food in the cage (including large pieces of food in the bedding) was weighed and recorded. If large portions of food had crumbled at the bottom of the cage, the data for that day was discarded because the small pieces on the bottom could not be accurately measured. The procedure was repeated for

three consecutive days. For these studies, the food intake per day per cage was divided

by the number of mice per cage. Then, the average food consumption per mouse in each

cage was calculated. Similarly, the mean body weight for the cage was calculated, and

the average food consumption per body weight (mean for the cage) was calculated.

4. Blood chemistry: All mice were assigned to either blood chemistry or gene expression

studies at weaning. Blood chemistry studies were performed after 28 and 100 days of

HFSC diet exposure. For these studies, the mice were fasted overnight (~16 hours). The

following morning, the mice were anesthetized with Avertin (see Chapter 4). Five

minutes later, tail glucose measurements were obtained (One Touch Ultra blood glucose

monitoring system, LifeScan, Milpitas, CA), and a blood sample was collected from the

retro-orbital sinus. For the blood studies, ~250 uL (measured using gradations on tube)

of whole blood was collected in an un-coated microcapillary tube into a microtainer tube

coated with EDTA (Microtainer tube with EDTA, Becton Dickinson, Franklin Lanes,

NJ). Then, ~400 uL (measured using gradations on tube) of whole blood was collected in

a heparinized microcapillary tube and placed in a heparinized microtainer tube

(Microtainer plasma separator tube with heparin, Becton Dickinson, Franklin Lanes, NJ).

246

Table AV-1. Composition of LabDiet 5010. For more information see www.labdiet.com.

Nutrients Minerals Protein % 23.5 Ash, % 7.2 Arginine, % 1.4 Calcium, % 1 Cystine, % 0.34 Phosphorus, % 0.67 Glycine, % 1.2 Phosphorus (non-phytate), % 0.43 Histidine, % 0.58 Potassium, % 0.92 Isoleucine, % 1.24 Magnesium, % 0.22 Leucine, % 1.87 Sulfur, % 0.24 Lysine, % 1.42 Sodium, % 0.28 Methionine, % 0.49 Chlorine, % 0.39 Phenylalanine, % 1.08 Fluorine, ppm 35 Tyrosine, % 0.64 Iron, ppm 240 Threonine, % 0.94 Zinc, ppm 124 Tryptophan, % 0.29 Manganese, ppm 115 Valine, % 1.22 Copper, ppm 18 Serine, % 1.23 Cobalt, ppm 0.44 Aspartic Acid, % 2.68 Iodine, ppm 1.2 Glutamic Acid, % 5.02 Chromium, ppm 2 Alanine, % 1.49 Selenium, ppm 0.31 Proline, % 1.73 Taurine, % 0.03

Vitamins Fat (ether exract), % 5.1 Carotene, ppm 4.5 Fat (acid hydrolysis), % 6.2 Vitamin K, ppm 3.4 Cholesterol, ppm 275 Thaimin, Hydrochloride, ppm 90 Linoleic Acid, % 1.182 Riboflavin, ppm 8 Linolenic Acid, % 0.12 Niacin, ppm 128 Arachidonic Acid, % <0.01 Pantothenic Acid, ppm 25 Omega-3 Fatty Acids, % 0.42 Choline Chloride, ppm 2200 Total Saturated Fatty Acids, % 1.4 Total Monounsaturated Fatty Acids, % 1.52 Folic Acid, ppm 6 Fiber (Crude), % 3.9 Pyridoxine, ppm 17 Neutral Detergent Fiber (cellulose, hemi- cellulose, lignin), % 12.7 Biotin, ppm 0.35 Acid Detergent Fiber (cellulose and lignin), % 4.5 Vitamin B12, mcg/kg 33 Nitrogen-Free Extract (by difference), % 50.3 Vitamin A, IU/gm 44 Starch, % 36.2 Vitamin D, IU/gm 4.4 Glucose, % 0.26 Vitamin E, IU/kg 66 Fructose, % 0.3 Ascorbic Acid, mg/gm --- Sucrose, % 1.02 Lactose, % 0

Total Digestible Nutrients, % 76 Calories: Gross Energy, kcal/gm 4.06 Protein 27.56% Physiological Fuel Value, kcal/gm 3.41 Fat (ether extract) 13.46% Metabolizable Energy, kcal/gm 3.17 Carbohydrates 58.99%

247

The samples were centrifuged at 2000g for 5 minutes, and plasma was removed and

stored at -80°C until further use.

The plasma collected in the EDTA coated tube was used for free fatty acid and

triglyceride measurements, and the plasma collected in the heparinized tube was used for total cholesterol, glucose, and ß-hydroxybutyrate measurements. ß-hydroxybutyrate was

measured only after 100 days of diet exposure. All reagent sets except free fatty acids

were obtained from Pointe Scientific (Canton, MI). The free fatty acid kit (NEFA C,

ACS-ACOD Method) was obtained from WAKO (Richmond, VA).

5. Gene expression analysis: Gene expression studies were performed in liver, skeletal

muscle, and gonadal fat pads after 28 days and 100 days of high-fat diet exposure. For the

gene expression analysis, 14-15 mice per strain and time point were fasted for 16 hours,

and cervical dislocation (without anesthesia, IACUC approval obtained) was used to

euthanize the mice. Sections of liver and skeletal muscle were carefully dissected and

placed in 5-7 mL RNAlater (Ambion, Austin, TX). Gonadal fat pads were also carefully

dissected and snap frozen in liquid nitrogen. All dissections and tissue handling were

performed in less than 3 minutes. Liver and skeletal muscle were stored in RNAlater at

4ºC overnight and then at -20 ºC until used. Gonadal fat pads were stored at -80ºC until

used. For each strain and time point, four individual mice from separate cages were

randomly selected for analysis (after the heaviest and leanest mouse from each strain and

time point was removed).

RNA isolations were performed using the appropriate Qiagen RNeasy Kit

(Qiagen, Valencia, CA) depending on the tissue type (liver: RNeasy Mini Kit; skeletal 248

muscle: RNeasy Mini Kit using protocol for fibrous tissues; adipose: RNeasy Lipid

Tissue Mini Kit). To increase yield/concentration for adipose samples (100 day time point only), two sections of tissue per mouse were used, and the column eluate

(containing RNA) from one section was used to elute the column of the second section

(per Qiagen directions). The concentration of RNA was determined by diluting each sample in RNAse-free water and measuring the absorption at 260 nm using a spectrophotometer. All liver samples and one simultaneously extracted sample of RNA from skeletal muscle and adipose tissues (RNA concentrations were too low to run each sample submitted to the CWRU gene expression array core) were run on a 1% agarose gel and visualized with ethidium bromide under ultra-violet light (Gel Doc 2000, Bio-

Rad, Hercules, CA) to assess the quality of the RNA. A minimum of 3.5 µg of total

RNA per sample (no pooling of RNA was performed) was submitted to the array core for analysis using the Bioanalyzer, cRNA preparation, and array hybridization.

6. Subcongenic panel construction: A total of 8 subcongenic strains were constructed

(Figure AIV-1). Subcongenic panel construction was begun at the same time that the 62-

BL substrains were generated (see Chapter 3). Therefore, the subcongenic strains were constructed using 62-B. 62-B female mice were crossed to C57BL/6J male mice to generate F1 progeny. The F1 offspring were intercrossed, and the offspring were genotyped to determine if recombinant A/J segments were inherited. Mice with the desired A/J segments were backcrossed to C57BL/6J. The offspring which inherited identical segments intact were identified by genotyping and intercrossed to homozygose the line. Brother-sister mating was used to maintain each subcongenic line. 249

s a homozygouss a A/J-

isted are from Build 35 of the 35 of the from Build isted are

6-34541420 6-35919902 rs3024195 6-38783944 6-43395279

and het denotes andheterozygous locationsAll marker alleles. l

A BBBB B BA ABBB B B B B B BA B B AAAA A B AA B AAAA A B A B A B A B A B

D6MIT138 D6Mit264 #23 SNP D6MIT159 #32 SNP 6-32961136

Mb 4.5 16.6 23.4 29.7 32.2 33 34.6 35.9 36.5 38.8 43.5

-BBA eBB B B BBAA A hetB 4-AB 9A BBAA A A BAAA A A A A hetB B B B89-AB B 97-AB 92-A 7-AAABB B B B B B B AAAA A A A104-A A hetB 270-AA BBBA A A A A hetB 316-AB 111-AB 6-BBBBA A A B B B B 161-AB Strain Marker 62-BL 62-BS Figure AIV-1. 62-B subcongenic panel. The eight new subcongenic strains and 92-A, 62-BS, and 62-BL are shown below. A denote A below. shown are 62-BL and 62-BS, 92-A, and strains subcongenic new eight The panel. subcongenic 62-B AIV-1. Figure derivedhomozygous alleles, B denotes a C57BL/6J-derived alleles, constructed. strains subcongenic new the are bold in listed Strains (NCBI). sequence genome mouse 250

7. Genotyping: For subcongenic strain analysis, both microsatellite markers and

SNPs were used for subcongenic panel construction and recombination breakpoint analyses. All genotyping was performed as described in Chapter 3. For subcongenic panel construction the following genetic markers were used: D6Mit138, D6Mit159,

D6Mit159, SNP #23, SNP#32, rs3024195 (90159). To refine recombination breakpoints, the following genetic markers were used: 6-32961136, 6-34541420, 6-35919902, 6-

38783944, and 6-43395279. Details about markers are provided in Table III-1. For the

62-BL interaction cross, the microsatellite markers used in the original F2 mapping study described in Chapter 2 were genotyped in the F2 progeny. The primers and reaction conditions are provided in Chapter 2.

8. Statistical Analyses:

Diet studies and blood chemistry: An unpaired t-test was used to compare the body weight, blood chemistry, and food consumption between 62-BL and 92-A (GraphPad

Prism version 3.0). If the variance of the samples differed significantly based on an F test, a Welch t-test was used (GraphPad Prism, version 3.0).

Gene expression studies: The Affymetrix GeneChip Mouse Genome 430 2.0 array, a whole genome mouse array, was used (Affymetrix, Santa Clara, CA). For all arrays in the analysis, the .CEL files, which have raw intensity values for all probes, are loaded into the R software environment (http://www.r-project.org/) using R/AFFY and

R/AFFYPLM packages (GAUTIER et al. 2004). The RMA (robust multi-array average)

method for background adjustment, data normalization (quantile normalization), and 251 summarization (Tukey median polish) of probe level data is used (IRIZARRY et al. 2003) to pre-process the data before statistical analysis. Prior to running RMA, signal intensity histograms and box-plots, array image reconstructions, % mismatch, and MA plots are analyzed to assess the quality of the raw data. MA plots and box-plots after RMA are compared to the non-processed “raw” data to be certain that normalization occurred as expected, and scatterplots of arrays can be used to identify potential outlier arrays (or genes). The pre-processed data is then used for all further analysis.

C. RESULTS AND DISCUSSION

1. 62-BL congenic male mice are resistant to high-fat, diet-induced obesity

To confirm that the weight gain difference in 62-BL vs. 92-A male mice was specific to the HFSC diet, weight gain studies were performed using both 5010 rodent diet, our maintenance diet, and the HFSC diet (Figure AIV-2, Table AIV-2). If the weight difference was not specific to the HFSC diet, we hypothesized that the mice would have a similar weight difference on the 5010 diet. After 100 days of consuming 5010 diet, the

92-A males were slightly heavier than 62-BL males (p=0.01). The weight difference was approximately 1.15 grams which represents 3-4% of the body weight at that time point.

In contrast, after consuming the HFSC diet for 100 days, the 92-A males were also heavier than 62-BL males (p<0.0001) but the difference in body weight between the two strains on the HFSC diet was 4.66 grams, which represents 11% of the body weight.

Thus, because the percentage of body weight difference between these strains is greater on the HFSC vs. 5010 diet, this result confirms that 62-BL is resistant to diet-induced obesity relative to 92-A male mice. 252

Figure AIV-2. 92-A vs. 62-BL growth curves on 5010 and HFSC diets. The mean weight gain and standard error for each strain at each time point is plotted.

Body weight (grams) 20 92-A fed HFSC 62-B fed HFSC 92-A fed 5010 62-B fed 5010 10 0 25 50 75 100 125 150

Age (days)

253

aired t-test. aired t-test.

62-BL 17 0.01 0.27 +/- 2.344 32 0.03 62-BL 21 0.03 +/- 0.35 2.237 48 0.03 )

) 2 2

(grams) 62-BL 17 1.48 28.68 +/- 2.678 32 0.01 (grams) 62-BL 58 36.86 +/- 4.26 5.904 123 <0.0001 (grams/cm (grams/cm

Diet Trait Strain size Sample Trait +/- Std Dev Value Mean T statistic df value P Diet Trait Strain size Sample Dev Std +/- TraitValue Mean T statistic df value P 5010 BMI 92-A 17 0.01 +/- 0.28 ** ** ** 5010 FW 92-A 17 0.99 29.83 +/- ** ** ** HFSC FWHFSC 92-A BMI 67 92-A 29 41.63 +/- 4.70 0.03 +/- 0.37 ** ** ** ** ** ** Table AIV-2.TableFW and for BMI 92-Avs. 62-BL after100 days of 5010 HFSC or diet consumption. Traits were compared using an unp Abbreviation: Std Dev = standard deviation deviation Dev = Std standard Abbreviation: 254

2. Food consumption does not explain obesity resistance conferred by Obrq2

We hypothesized that differences in food consumption may explain the weight difference in 92-A vs. 62B-L males. To test this hypothesis, we performed food consumption experiments in group housed mice after 80-90 days of HFSC diet exposure.

At this time point, 92-A was heavier than 62-BL (92-A: 38.02 +/- 3.61 grams, 62-BL:

33.20 +/- 2.86 grams, t=2.661, df=13, p=0.02). Despite the difference in body weight, neither food consumption nor food consumption per gram body weight (per cage) differed between the two strains (Figure AIV-3, Table AIV-3). Therefore, food consumption does not explain the body weight difference. Analyses of energy expenditure, body activity, and fat absorption may provide more clues to the mechanism of resistance.

3. 62-BL males are resistant to high-fat, diet-induced hyperglycemia

We hypothesized that the resistance to hyperglycemia and hypercholesterolemia maps to the same region as the obesity resistance in 62-BL. To test this hypothesis, we analyzed blood chemistry in 92-A and 62-BL males. Although no significant differences in blood chemistry were detected after 28 days of high-fat diet consumption, several differences were observed after 100 days of HFSC diet consumption. After 100 days of

HFSC diet exposure, both tail blood glucose and plasma glucose measurements were significantly lower in 62-BL males relative to 92-A males (Table AIV-4). Thus, resistance to high-fat, diet-induced hyperglycemia maps to the same region as the obesity resistance in 62-BL. Investigation of insulin levels will reveal whether the 92-A strain is insulin resistant relative to 62-BL. 255

Figure AIV-3. HFSC diet consumption in 92-A vs. 62-BL male mice. For all traits, the mean value is indicated with a bar, and trait comparisons were performed using an unpaired t-test or Welch’s t-test if variance significantly differed between the strains. Significant (<0.05) p values are provided. A. The body weights in singly housed males were not significantly different. Therefore, food intake could not be evaluated. B. The body weights in group housed males were significantly different after 80-90 days of HFSC diet consumption, so group housed mice were used for the food consumption studies.

A. 50

(grams) Weight Body 30 92-A 62-BL Strain B.

P=0.02 50

(grams) Cage 10 Weight Body Per Mean 0 92-A 62-BL Strain

3.00 0.10

2.75 0.09

0.08 2.50 (grams) 0.07 2.25 Mean Food Intake Weight (grams) Weight

Mean Food Intake 0.06 (grams) / Mean Body / Mean (grams) 2.00 92-A 62-BL 0.05 Strain 92-A 62-BL Strain

256

Table AIV-3. HFSC diet consumption in 92-A vs. 62-BL males. Grouped housed males were used. The mean body weight per cage, the mean food consumption per cage, and the food consumption/body weight was compared between the two strains using an unpaired t-test (Welch’s t-test was used if the variance was significantly different based on an F test). *indicates Welch’s t-test used.

Trait 92-A n 62-BL n T statistic df P Value Body weight per cage (grams) 38.04 +/- 3.99 8 33.20 +/- 2.86 7 2.661 13 0.0196 Food consumption per cage (grams) 2.57 +/- 0.25 8 2.42 +/- 0.08 7 1.622 8* 0.1434 Food consumption/body weight 0.068 +/- 0.01 8 0.073 +/- 0.009 7 1.007 13 0.3321

257

Table AIV-4. Blood chemistry in 92-A vs. 62-BL male mice fed the HFSC diet for 28 or 100 days. Comparisons were made using an unpaired t-test (Welch’s t-test if variance significantly differed using an F test). *indicates Welch’s t-test.

28 Days Trait 92-A n 62-BL n T statistic df P Value Body weight (grams) 27.60 +/- 1.12 16 24.82 +/- 2.07 16 4.719 23* <0.0001 BMI (grams/cm2) 0.28 +/- 0.01 16 0.26 +/- 0.01 16 6.032 24* <0.0001 Tail glucose (mg/dl) 124 +/- 20 16 138 +/- 27 16 1.544 30 0.1332 Glucose (mg/dl) 184 +/- 42 16 201 +/- 59 16 0.9794 30 0.3352 Total cholesterol (mg/dl) 111 +/- 14 16 110 +/- 16 16 0.2702 30 0.7888 Triglyceride (mg/dl) 69 +/- 15 16 57 +/- 21 15 1.851 29 0.0744 Free fatty acids (mEq) 0.45 +/- 0.09 16 0.40 +/- 0.13 12 1.335 26 0.1934

100 Days Trait 92-A n 62-BL n T statistic df P Value Body weight (grams) 41.79 +/- 3.85 31 38.02 +/- 5.31 21 2.651 33* 0.0122 BMI (grams/cm2) 0.37 +/- 0.03 29 0.35 +/- 0.04 21 2.237 48 0.0299 Tail glucose (mg/dl) 206 +/- 35 29 174 +/- 37 21 3.181 48 0.0026 Glucose (mg/dl) 282 +/- 43 29 247 +/- 40 19 2.858 46 0.0064 Total cholesterol (mg/dl) 155 +/- 19 27 152 +/- 22 19 0.3526 44 0.726 Triglyceride (mg/dl) 60 +/- 17 29 65 +/- 15 21 1.181 48 0.2432 Free fatty acids (mEq) 0.42 +/- 0.09 28 0.50 +/- 0.09 20 2.854 46 0.0065 β-hydroxybutyrate (mg/dl) 6.96 +/- 2.47 21 7.83 +/- 3.00 18 1.002 37 0.3228

258

In contrast, plasma free fatty acid levels were significantly higher in 62-BL males

relative to 92-A males. Perhaps, increased triglyceride catabolism is associated with

resistance in 62-BL. Surprisingly, no differences in plasma β-hydroxybutyrate levels

were detected in 92-A vs. 62-BL which suggests that β-oxidation or free fatty acid metabolism does not differ among the strains. Further investigations of triglyceride turnover during a fast (using stable isotopes) may provide clues to the differences in triglyceride metabolism in these strains. No significant differences were detected in total cholesterol or plasma triglycerides. Thus, the resistance to hypocholesterolemia observed in the CSS and in 62-BL may map to the centromeric portion of 62-BL that is shared with

92-A. Further investigations of cholesterol are necessary because our previous studies demonstrated that the trait difference was small in 62-B vs. C57BL/6J and may be difficult to map (if reproducible).

In both strains, relative to the 28 day time point, glucose and cholesterol were higher at the 100 day time point which suggests that these values may rise as body weight increases during the time course. In contrast, triglyceride levels did not change over time. Interestingly, free fatty acid levels were higher at the later time point only in 62-BL whereas the free fatty acid levels in 92-A did not change during the time course. This strain-specific, free fatty acid effect provides evidence suggesting that the two strains respond differently to the HFSC diet. Perhaps, as mentioned previously, this differential free fatty acid response is associated with the mechanism for obesity resistance.

259

4. 92-A vs. 62-BL gene expression studies

To test whether genes within the candidate region were differentially expressed in

adipose tissue, skeletal muscle, and liver of 92-A vs. 62-BL males and to test whether

differences in pathway expression could be detected, microarray analyses were pursued in the two strains. These studies are a work in progress and have not been completed.

Planned studies include analysis with R/MAANOVA (WU 2003) and SAM (significance

analysis of microarray) (TUSHER et al. 2001) to identify differentially expressed genes in

92-A vs. 62-BL in each of the tissues at both the early and late time point. The sets of

differentially expressed genes may then be analyzed using EASE (HOSACK et al. 2003) to

test whether genes with a particular function or cellular localization are over-represented

in the differentially expressed gene list. In addition, GSEA (gene set enrichment

analysis) (SUBRAMANIAN et al. 2005) may be used to identify pathways that are up- or

down-regulated in 92-A vs. 62-BL tissues at each time point.

5. Subcongenic strains suggest complexity associated with Obrq2

To localize Obrq2, a panel of subcongenic strains derived from 62-BL was

constructed and screened for weight gain on the HFSC diet. We hypothesized that the

regions of the chromosome shared by the obesity resistant strains would narrow the

candidate interval. Of the six subcongenic strains tested, two (161-A and 4-A) were

obesity resistant (Figure AIV-4, Table AIV-5). The FW from the remaining four strains

was not significantly different from 92-A. Although 161-A and 4-A share a common

A/J-derived segment, this same segment is also present in 89-A and 111-A, which are

both obese strains. Thus, the subcongenic data does not provide evidence for a single 260

obesity resistant QTL in this region. Larger sample sizes, replicate analyses, and

additional strains are necessary to confirm this result.

6. Suppression of Obrq2 on chromosome 6

In the congenic strains discussed in Chapter 3, Obrq3 was discovered to suppress

Obrq2. To test whether we can discover suppressors or enhancers of Obrq2 on

chromosome 6, F2 progeny derived from 62-BL and B6-Chr 6A were analyzed. The F2 progeny have the 62-BL A/J-derived segment with the remainder of the chromosome segregating. We hypothesized that if no additional QTLs on chromosome 6 suppressed or enhanced the Obrq2 phenotype the FW of the F2 progeny would be similar to 62-BL.

Instead, the FW in the F2 progeny was significantly higher than 62-BL (F2: 39.24 +/-

5.31 grams, 62-BL: 36.86 +/- 4.26 grams, Welch’s t=3.026, df=140, p=0.0029) (Figure

AIV-5). Furthermore, as expected if other QTLs are segregating and influencing the 62-

BL phenotype, the variance is larger in the F2 vs. 62-BL males. Thus, we concluded that a suppressor of Obrq2 was detectable in the cross. The genotyping for the cross is underway and mapping studies should reveal the likely location(s) of the suppressor (s)

of Obrq2.

261

Figure AIV-4. 62-BL subcongenic HFSC diet survey. The mean FW of each strain is indicated with a bar. The strains were compared to 92-A using an unpaired t-test or Welch’s t-test if the variances differed. * indicates p < 0.009.

(grams) weight Final

*** 92-A 97-A 111-A 104-A 89-A 4-A 161-A 62-BL Strain

262

Table AIV-5. 62-BL subcongenic HFSC diet survey. The FW +/- one standard deviation is presented for each subcongenic strain. All subcongenic strains were compared to 92-A using an unpaired t-test (Welch’s t-test if the variance significantly differed based on an F test). *indicates Welch’s t-test was used.

Strain n Mean FW (grams) Tstatistic df P value 92-A 67 41.63 +/- 4.70 *** *** *** 62-BL 58 36.86 +/- 4.26 5.904 123 <0.0001 104-A 26 40.30 +/- 6.69 0.926 35* 0.3609 89-A 24 40.52 +/- 3.83 1.039 89 0.3018 111-A 15 41.22 +/- 5.87 0.292 80 0.7707 97-A 46 42.79 +/- 6.53 1.036 76* 0.3033 161-A 30 37.44 +/- 3.82 4.636 68 <0.0001 4-A 32 38.29 +/- 6.17 2.706 48* 0.0094

263

Figure AIV-5. F2 male progeny derived from B6-Chr 6A and 62-BL mice. The F2 male progeny were heavier than 62-BL male mice. The mean body weight of each population is indicated with a bar. A Welch’s t –test was used to compare the traits.

P=0.003

40 (grams) 30 Final weight body

20 62-BL 62-BL F2 Strain

264

REFERENCES

265

1969 The American Heritage Dictionary. American Heritage Publishing and Houghton Mifflin Company, Boston. ABU-ELHEIGA, L., W. OH, P. KORDARI and S. J. WAKIL, 2003 Acetyl-CoA carboxylase 2 mutant mice are protected against obesity and diabetes induced by high-fat/high- carbohydrate diets. Proc Natl Acad Sci U S A 100: 10207-10212. ACKERMAN, K. G., H. HUANG, H. GRASEMANN, C. PUMA, J. B. SINGER et al., 2005 Interacting genetic loci cause airway hyperresponsiveness. Physiol Genomics 21: 105-111. ALBERTI, K. G., and P. Z. ZIMMET, 1998 Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med 15: 539- 553. ARITA, Y., S. KIHARA, N. OUCHI, M. TAKAHASHI, K. MAEDA et al., 1999 Paradoxical decrease of an adipose-specific protein, adiponectin, in obesity. Biochem Biophys Res Commun 257: 79-83. ARSOV, T., D. G. SILVA, K. O'BRYAN M, A. SAINSBURY, N. J. LEE et al., 2006 Fat Aussie - a New Alstrom Syndrome Mouse Showing a Critical Role for Alms1 in Obesity, Diabetes and Spermatogenesis. Mol Endocrinol. Electronic publication ahead of print. ARYA, R., J. BLANGERO, K. WILLIAMS, L. ALMASY, T. D. DYER et al., 2002 Factors of insulin resistance syndrome--related phenotypes are linked to genetic locations on chromosomes 6 and 7 in nondiabetic mexican-americans. Diabetes 51: 841-847. ASAKAWA, A., A. INUI, H. YUZURIHA, N. UENO, G. KATSUURA et al., 2003 Characterization of the effects of pancreatic polypeptide in the regulation of energy balance. Gastroenterology 124: 1325-1336. BALKAU, B., and M. A. CHARLES, 1999 Comment on the provisional report from the WHO consultation. European Group for the Study of Insulin Resistance (EGIR). Diabet Med 16: 442-443. BATTERHAM, R. L., A. W. COWLEY, JR., C. J. SMALL, H. HERZOG, M. A. COHEN et al., 2004 Physiology: Does gut hormone PYY 3-36 decrease food intake in rodents? (reply). Nature 430: 3-4. BATTERHAM, R. L., M. A. COWLEY, C. J. SMALL, H. HERZOG, M. A. COHEN et al., 2002 Gut hormone PYY(3-36) physiologically inhibits food intake. Nature 418: 650- 654. BAYONA-BAFALUY, M. P., R. ACIN-PEREZ, J. C. MULLIKIN, J. S. PARK, R. MORENO- LOSHUERTOS et al., 2003 Revisiting the mouse mitochondrial DNA sequence. Nucleic Acids Res 31: 5349-5355. BEECHEY, C. V., 2000 Peg1/Mest locates distal to the currently defined imprinting region on mouse proximal chromosome 6 and identifies a new imprinting region affecting growth. Cytogenet Cell Genet 90: 309-314. BELKNAP, J. K., 2003 Chromosome substitution strains: some quantitative considerations for genome scans and fine mapping. Mamm Genome 14: 723-732. BELL, C. G., A. J. WALLEY and P. FROGUEL, 2005 The genetics of human obesity. Nat Rev Genet 6: 221-234. 266

BENOIT, S. C., E. L. AIR, L. M. COOLEN, R. STRAUSS, A. JACKMAN et al., 2002 The catabolic action of insulin in the brain is mediated by melanocortins. J Neurosci 22: 9048-9052. BERGE, K. E., H. TIAN, G. A. GRAF, L. YU, N. V. GRISHIN et al., 2000 Accumulation of dietary cholesterol in sitosterolemia caused by mutations in adjacent ABC transporters. Science 290: 1771-1775. BERMAN, Y., N. MZHAVIA, A. POLONSKAIA and L. A. DEVI, 2001 Impaired prohormone convertases in Cpe(fat)/Cpe(fat) mice. J Biol Chem 276: 1466-1473. BILLINGTON, C. J., J. E. BRIGGS, M. GRACE and A. S. LEVINE, 1991 Effects of intracerebroventricular injection of neuropeptide Y on energy metabolism. Am J Physiol 260: R321-327. BLACK, B. L., J. CROOM, E. J. EISEN, A. E. PETRO, C. L. EDWARDS et al., 1998 Differential effects of fat and sucrose on body composition in A/J and C57BL/6 mice. Metabolism 47: 1354-1359. BOTSTEIN, D., and N. RISCH, 2003 Discovering genotypes underlying human phenotypes: past successes for mendelian disease, future approaches for complex disease. Nat Genet 33 Suppl: 228-237. BOUCHARD, C., A. TREMBLAY, J. P. DESPRES, A. NADEAU, P. J. LUPIEN et al., 1990 The response to long-term overfeeding in identical twins. N Engl J Med 322: 1477- 1482. BRAY, G. A., 1990 Obesity: historical development of scientific and cultural ideas. Int J Obes 14: 909-926. BROCKMANN, G. A., and M. R. BEVOVA, 2002 Using mouse models to dissect the genetics of obesity. Trends Genet 18: 367-376. BROCKMANN, G. A., E. KARATAYLI, C. S. HALEY, U. RENNE, O. J. ROTTMANN et al., 2004 QTLs for pre- and postweaning body weight and body composition in selected mice. Mamm Genome 15: 593-609. BROCKMANN, G. A., J. KRATZSCH, C. S. HALEY, U. RENNE, M. SCHWERIN et al., 2000 Single QTL effects, epistasis, and pleiotropy account for two-thirds of the phenotypic F(2) variance of growth and obesity in DU6i x DBA/2 mice. Genome Res 10: 1941-1957. BROMAN, K. W., H. WU, S. SEN and G. A. CHURCHILL, 2003 R/qtl: QTL mapping in experimental crosses. Bioinformatics 19: 889-890. BROWN, C. D., M. HIGGINS, K. A. DONATO, F. C. ROHDE, R. GARRISON et al., 2000 Body mass index and the prevalence of hypertension and dyslipidemia. Obes Res 8: 605-619. BROWNLOW, B. S., A. PETRO, M. N. FEINGLOS and R. S. SURWIT, 1996 The role of motor activity in diet-induced obesity in C57BL/6J mice. Physiol Behav 60: 37-41. BRUNENGRABER, D. Z., B. J. MCCABE, J. KATANIK and S. F. PREVIS, 2002 Gas chromatography-mass spectrometry assay of the (18)o enrichment of water as trimethyl phosphate. Anal Biochem 306: 278-282. BULTMAN, S. J., E. J. MICHAUD and R. P. WOYCHIK, 1992 Molecular characterization of the mouse agouti locus. Cell 71: 1195-1204. BUTLER, A. A., and R. D. CONE, 2001 Knockout models resulting in the development of obesity. Trends Genet 17: S50-54. 267

CASTLE, W. E., 1941 Influence of certain coat color mutations on body size in mice, rats, and rabbits. Genetics 26: 177-191. CHAVEZ, M., K. KAIYALA, L. J. MADDEN, M. W. SCHWARTZ and S. C. WOODS, 1995 Intraventricular insulin and the level of maintained body weight in rats. Behav Neurosci 109: 528-531. CHEN, A. S., D. J. MARSH, M. E. TRUMBAUER, E. G. FRAZIER, X. M. GUAN et al., 2000 Inactivation of the mouse melanocortin-3 receptor results in increased fat mass and reduced lean body mass. Nat Genet 26: 97-102. CHEN, H., O. CHARLAT, L. A. TARTAGLIA, E. A. WOOLF, X. WENG et al., 1996 Evidence that the diabetes gene encodes the leptin receptor: identification of a mutation in the leptin receptor gene in db/db mice. Cell 84: 491-495. CHEVERUD, J. M., T. H. EHRICH, T. HRBEK, J. P. KENNEY, L. S. PLETSCHER et al., 2004 Quantitative trait loci for obesity- and diabetes-related traits and their dietary responses to high-fat feeding in LGXSM recombinant inbred mouse strains. Diabetes 53: 3328-3336. CHEVERUD, J. M., E. J. ROUTMAN, F. A. DUARTE, B. VAN SWINDEREN, K. COTHRAN et al., 1996 Quantitative trait loci for murine growth. Genetics 142: 1305-1319. CHEVERUD, J. M., T. T. VAUGHN, L. S. PLETSCHER, A. C. PERIPATO, E. S. ADAMS et al., 2001 Genetic architecture of adiposity in the cross of LG/J and SM/J inbred mice. Mamm Genome 12: 3-12. CHITTURI, S., S. ABEYGUNASEKERA, G. C. FARRELL, J. HOLMES-WALKER, J. M. HUI et al., 2002 NASH and insulin resistance: Insulin hypersecretion and specific association with the insulin resistance syndrome. Hepatology 35: 373-379. CHRISTOPOULOU-ALETRA, H., and N. PAPAVRAMIDOU, 2004 Methods used by the hippocratic physicians for weight reduction. World J Surg 28: 513-517. CHUNG, W. K., J. GOLDBERG-BERMAN, L. POWER-KEHOE and R. L. LEIBEL, 1996 Molecular mapping of the tubby (tub) mutation on mouse chromosome 7. Genomics 32: 210-217. CLARK, J. T., P. S. KALRA, W. R. CROWLEY and S. P. KALRA, 1984 Neuropeptide Y and human pancreatic polypeptide stimulate feeding behavior in rats. Endocrinology 115: 427-429. CLEMENT, K., C. VAISSE, N. LAHLOU, S. CABROL, V. PELLOUX et al., 1998 A mutation in the human leptin receptor gene causes obesity and pituitary dysfunction. Nature 392: 398-401. COHEN, J., A. PERTSEMLIDIS, I. K. KOTOWSKI, R. GRAHAM, C. K. GARCIA et al., 2005 Low LDL cholesterol in individuals of African descent resulting from frequent nonsense mutations in PCSK9. Nat Genet 37: 161-165. COHEN, J. C., E. BOERWINKLE, T. H. MOSLEY, JR. and H. H. HOBBS, 2006 Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. N Engl J Med 354: 1264-1272. COLEMAN, D. L., 1982 Diabetes-obesity syndromes in mice. Diabetes 31: 1-6. COLEMAN, D. L., and E. M. EICHER, 1990 Fat (fat) and tubby (tub): two autosomal recessive mutations causing obesity syndromes in the mouse. J Hered 81: 424- 427. COLEMAN, D. L., and K. P. HUMMEL, 1973 The influence of genetic background on the expression of the obese (Ob) gene in the mouse. Diabetologia 9: 287-293. 268

COLLIN, G. B., J. D. MARSHALL, A. IKEDA, W. V. SO, I. RUSSELL-EGGITT et al., 2002 Mutations in ALMS1 cause obesity, type 2 diabetes and neurosensory degeneration in Alstrom syndrome. Nat Genet 31: 74-78. COLLINS, S., T. L. MARTIN, R. S. SURWIT and J. ROBIDOUX, 2004 Genetic vulnerability to diet-induced obesity in the C57BL/6J mouse: physiological and molecular characteristics. Physiol Behav 81: 243-248. COMUZZIE, A. G., J. E. HIXSON, L. ALMASY, B. D. MITCHELL, M. C. MAHANEY et al., 1997 A major quantitative trait locus determining serum leptin levels and fat mass is located on human chromosome 2. Nat Genet 15: 273-276. CONSIDINE, R. V., M. K. SINHA, M. L. HEIMAN, A. KRIAUCIUNAS, T. W. STEPHENS et al., 1996 Serum immunoreactive-leptin concentrations in normal-weight and obese humans. N Engl J Med 334: 292-295. COOL, D. R., E. NORMANT, F. SHEN, H. C. CHEN, L. PANNELL et al., 1997 Carboxypeptidase E is a regulated secretory pathway sorting receptor: genetic obliteration leads to endocrine disorders in Cpe(fat) mice. Cell 88: 73-83. CORVA, P. M., S. HORVAT and J. F. MEDRANO, 2001 Quantitative trait loci affecting growth in high growth (hg) mice. Mamm Genome 12: 284-290. COWLEY, A. W., JR., M. LIANG, R. J. ROMAN, A. S. GREENE and H. J. JACOB, 2004a Consomic rat model systems for physiological genomics. Acta Physiol Scand 181: 585-592. COWLEY, A. W., JR., R. J. ROMAN and H. J. JACOB, 2004b Application of chromosomal substitution techniques in gene-function discovery. J Physiol 554: 46-55. CUENOT, L., 1905 Les races pures et leurs combinaisons chez les souris. Arch. Zool. Exp. Gen. 3: 123-132. DAKIN, C. L., I. GUNN, C. J. SMALL, C. M. EDWARDS, D. L. HAY et al., 2001 Oxyntomodulin inhibits food intake in the rat. Endocrinology 142: 4244-4250. DAKIN, C. L., C. J. SMALL, R. L. BATTERHAM, N. M. NEARY, M. A. COHEN et al., 2004 Peripheral oxyntomodulin reduces food intake and body weight gain in rats. Endocrinology 145: 2687-2695. DANFORTH, C. H., 1927 Hereditary adiposity in mice. J Hered 18: 153-162. DE FOURMESTRAUX, V., H. NEUBAUER, C. POUSSIN, P. FARMER, L. FALQUET et al., 2004 Transcript profiling suggests that differential metabolic adaptation of mice to a high fat diet is associated with changes in liver to muscle lipid fluxes. J Biol Chem 279: 50743-50753. DEGAWA-YAMAUCHI, M., J. E. BOVENKERK, B. E. JULIAR, W. WATSON, K. KERR et al., 2003 Serum resistin (FIZZ3) protein is increased in obese humans. J Clin Endocrinol Metab 88: 5452-5455. DELLA-FERA, M. A., and C. A. BAILE, 1979 Cholecystokinin octapeptide: continuous picomole injections into the cerebral ventricles of sheep suppress feeding. Science 206: 471-473. DENG, H. W., H. DENG, Y. J. LIU, Y. Z. LIU, F. H. XU et al., 2002 A genomewide linkage scan for quantitative-trait loci for obesity phenotypes. Am J Hum Genet 70: 1138- 1151. DENKE, M. A., C. T. SEMPOS and S. M. GRUNDY, 1993 Excess body weight. An underrecognized contributor to high blood cholesterol levels in white American men. Arch Intern Med 153: 1093-1103. 269

DENKE, M. A., C. T. SEMPOS and S. M. GRUNDY, 1994 Excess body weight. An underrecognized contributor to dyslipidemia in white American women. Arch Intern Med 154: 401-410. DEWALD, T., L. KHAODHIAR, M. P. DONAHUE and G. BLACKBURN, 2006 Pharmacological and surgical treatments for obesity. Am Heart J 151: 604-624. DIAMENT, A. L., and C. H. WARDEN, 2004 Multiple linked mouse chromosome 7 loci influence body fat mass. Int J Obes Relat Metab Disord 28: 199-210. DICKERSON, G. E., and J. W. GOWEN, 1947 Hereditary obesity and efficient food utilization in mice. Science 105: 496. DICKIE, M. M., and G. W. WOOLLEY, 1946 The Age Factor in Weight of Yellow Mice. J Hered 37: 365-368. DIETSCHY, J. M., and S. D. TURLEY, 2002 Control of cholesterol turnover in the mouse. J Biol Chem 277: 3801-3804. DREWNOWSKI, A., and S. E. SPECTER, 2004 Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr 79: 6-16. DUHL, D. M., H. VRIELING, K. A. MILLER, G. L. WOLFF and G. S. BARSH, 1994 Neomorphic agouti mutations in obese yellow mice. Nat Genet 8: 59-65. EXPERT PANEL ON DETECTION, E., AND TREATMENT OF HIGH BLOOD CHOLESTEROL IN ADULTS, 2001 Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 285: 2486-2497. FAN, W., B. A. BOSTON, R. A. KESTERSON, V. J. HRUBY and R. D. CONE, 1997 Role of melanocortinergic neurons in feeding and the agouti obesity syndrome. Nature 385: 165-168. FAROOQI, I. S., J. M. KEOGH, G. S. YEO, E. J. LANK, T. CHEETHAM et al., 2003 Clinical spectrum of obesity and mutations in the melanocortin 4 receptor gene. N Engl J Med 348: 1085-1095. FAROOQI, I. S., G. MATARESE, G. M. LORD, J. M. KEOGH, E. LAWRENCE et al., 2002 Beneficial effects of leptin on obesity, T cell hyporesponsiveness, and neuroendocrine/metabolic dysfunction of human congenital leptin deficiency. J Clin Invest 110: 1093-1103. FAROOQI, I. S., and S. O'RAHILLY, 2005 Monogenic obesity in humans. Annu Rev Med 56: 443-458. FLEGAL, K. M., M. D. CARROLL, C. L. OGDEN and C. L. JOHNSON, 2002 Prevalence and trends in obesity among US adults, 1999-2000. JAMA 288: 1723-1727. FLEMYNG, M., 1760 A discourse on the nature, causes, and cure of corpulency. L. Davis and C. Reymers, London. FORD, E. S., W. H. GILES and W. H. DIETZ, 2002 Prevalence of the metabolic syndrome among US adults: findings from the third National Health and Nutrition Examination Survey. JAMA 287: 356-359. FORTIN, A., E. DIEZ, D. ROCHEFORT, L. LAROCHE, D. MALO et al., 2001 Recombinant congenic strains derived from A/J and C57BL/6J: a tool for genetic dissection of complex traits. Genomics 74: 21-35. FOSTER, L. J., C. L. DE HOOG, Y. ZHANG, X. XIE, V. K. MOOTHA et al., 2006 A Mammalian organelle map by protein correlation profiling. Cell 125: 187-199. 270

FRIGERI, L. G., G. L. WOLFF and G. ROBEL, 1983 Impairment of glucose tolerance in yellow (Avy/A) (BALB/c X VY) F-1 hybrid mice by hyperglycemic peptide(s) from human pituitary glands. Endocrinology 113: 2097-2105. FRIGERI, L. G., G. L. WOLFF and C. TEGUH, 1988 Differential responses of yellow Avy/A and agouti A/a (BALB/c X VY) F1 hybrid mice to the same diets: glucose tolerance, weight gain, and adipocyte cellularity. Int J Obes 12: 305-320. FRIIS-HANSEN, L., K. A. LACOURSE, L. C. SAMUELSON and J. J. HOLST, 2001 Attenuated processing of proglucagon and glucagon-like peptide-1 in carboxypeptidase E- deficient mice. J Endocrinol 169: 595-602. GAUTIER, L., L. COPE, B. M. BOLSTAD and R. A. IRIZARRY, 2004 affy--analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 20: 307-315. GIBBS, J., R. C. YOUNG and G. P. SMITH, 1973 Cholecystokinin elicits satiety in rats with open gastric fistulas. Nature 245: 323-325. GILL, A. M., and T. T. YEN, 1991 Effects of ciglitazone on endogenous plasma islet amyloid polypeptide and insulin sensitivity in obese-diabetic viable yellow mice. Life Sci 48: 703-710. GOODALE, H., 1938 A study of inheritance of body weight in the albino mouse by selection. J Hered 29: 101-112. GUAN, X. M., H. YU and L. H. VAN DER PLOEG, 1998 Evidence of altered hypothalamic pro-opiomelanocortin/ neuropeptide Y mRNA expression in tubby mice. Brain Res Mol Brain Res 59: 273-279. HAGER, J., C. DINA, S. FRANCKE, S. DUBOIS, M. HOUARI et al., 1998 A genome-wide scan for human obesity genes reveals a major susceptibility locus on chromosome 10. Nat Genet 20: 304-308. HEALTH, N. I. O., 1998 Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: the evidence report, National Institutes of Health, Bethesda. HEARN, T., G. L. RENFORTH, C. SPALLUTO, N. A. HANLEY, K. PIPER et al., 2002 Mutation of ALMS1, a large gene with a tandem repeat encoding 47 amino acids, causes Alstrom syndrome. Nat Genet 31: 79-83. HEARN, T., C. SPALLUTO, V. J. PHILLIPS, G. L. RENFORTH, N. COPIN et al., 2005 Subcellular localization of ALMS1 supports involvement of centrosome and basal body dysfunction in the pathogenesis of obesity, insulin resistance, and type 2 diabetes. Diabetes 54: 1581-1587. HEDLEY, A. A., C. L. OGDEN, C. L. JOHNSON, M. D. CARROLL, L. R. CURTIN et al., 2004 Prevalence of overweight and obesity among US children, adolescents, and adults, 1999-2002. JAMA 291: 2847-2850. y HESTON, W. E., and M. K. DERINGER, 1947 Relationship between the lethal yellow (A ) gene of the mouse and susceptibility to spontaneous pumonary tumors. J Natl Cancer Inst 7: 463-465. HESTON, W. E., and G. VLAHAKIS, 1961 Influence of the Ay gene on mammary-gland tumors, hepatomas, and normal growth in mice. J Natl Cancer Inst 26: 969-983. HESTON, W. E., and G. VLAHAKIS, 1968 C3H-Avy--a high hepatoma and high mammary tumor strain of mice. J Natl Cancer Inst 40: 1161-1166. HILL, J. O., and J. C. PETERS, 1998 Environmental contributions to the obesity epidemic. Science 280: 1371-1374. 271

HILL, J. O., H. R. WYATT, G. W. REED and J. C. PETERS, 2003 Obesity and the environment: where do we go from here? Science 299: 853-855. HILL, R. J., and P. S. DAVIES, 2001 The validity of self-reported energy intake as determined using the doubly labelled water technique. Br J Nutr 85: 415-430. HIPPOCRATES, 1939 in The Genuine Works of Hippocrates. The Williams and Wilkins Company, Baltimore, MD. HOIT, B. D., S. KIATCHOOSAKUN, J. RESTIVO, D. KIRKPATRICK, K. OLSZENS et al., 2002 Naturally occurring variation in cardiovascular traits among inbred mouse strains. Genomics 79: 679-685. HOLCOMB, I. N., R. C. KABAKOFF, B. CHAN, T. W. BAKER, A. GURNEY et al., 2000 FIZZ1, a novel cysteine-rich secreted protein associated with pulmonary inflammation, defines a new gene family. EMBO J 19: 4046-4055. HOSACK, D. A., G. DENNIS, JR., B. T. SHERMAN, H. C. LANE and R. A. LEMPICKI, 2003 Identifying biological themes within lists of genes with EASE. Genome Biol 4: R70. HU, E., P. LIANG and B. M. SPIEGELMAN, 1996 AdipoQ is a novel adipose-specific gene dysregulated in obesity. J Biol Chem 271: 10697-10703. HUMMEL, K. P., D. L. COLEMAN and P. W. LANE, 1972 The influence of genetic background on expression of mutations at the diabetes locus in the mouse. I. C57BL-KsJ and C57BL-6J strains. Biochem Genet 7: 1-13. HUMMEL, K. P., M. M. DICKIE and D. L. COLEMAN, 1966 Diabetes, a new mutation in the mouse. Science 153: 1127-1128. HUSZAR, D., C. A. LYNCH, V. FAIRCHILD-HUNTRESS, J. H. DUNMORE, Q. FANG et al., 1997 Targeted disruption of the melanocortin-4 receptor results in obesity in mice. Cell 88: 131-141. IHAKA, R., and R. GENTLEMAN, 1996 R: A language for data analysis and graphics. Journal of Computational and Graphical Statistics 5: 299-314. INGALLS, A. M., M. M. DICKIE and G. D. SNELL, 1950 Obese, a new mutation in the house mouse. J Hered 41: 317-318. IRIZARRY, R. A., B. HOBBS, F. COLLIN, Y. D. BEAZER-BARCLAY, K. J. ANTONELLIS et al., 2003 Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4: 249-264. ISHIMORI, N., R. LI, P. M. KELMENSON, R. KORSTANJE, K. A. WALSH et al., 2004 Quantitative trait loci that determine plasma lipids and obesity in C57BL/6J and 129S1/SvImJ inbred mice. J Lipid Res 45: 1624-1632. JACKSON, R. S., J. W. CREEMERS, I. S. FAROOQI, M. L. RAFFIN-SANSON, A. VARRO et al., 2003 Small-intestinal dysfunction accompanies the complex endocrinopathy of human proprotein convertase 1 deficiency. J Clin Invest 112: 1550-1560. JACKSON, R. S., J. W. CREEMERS, S. OHAGI, M. L. RAFFIN-SANSON, L. SANDERS et al., 1997 Obesity and impaired prohormone processing associated with mutations in the human prohormone convertase 1 gene. Nat Genet 16: 303-306. JARVIS, J. P., J. KENNEY-HUNT, T. H. EHRICH, L. S. PLETSCHER, C. F. SEMENKOVICH et al., 2005 Maternal genotype affects adult offspring lipid, obesity, and diabetes phenotypes in LGXSM recombinant inbred strains. J Lipid Res 46: 1692-1702. KAHN, R., J. BUSE, E. FERRANNINI and M. STERN, 2005 The metabolic syndrome: time for a critical appraisal: joint statement from the American Diabetes Association 272

and the European Association for the Study of Diabetes. Diabetes Care 28: 2289- 2304. KELLEY, R. I., 2000 Inborn errors of cholesterol biosynthesis. Adv Pediatr 47: 1-53. KERSHAW, E. E., N. M. MORTON, H. DHILLON, L. RAMAGE, J. R. SECKL et al., 2005 Adipocyte-specific glucocorticoid inactivation protects against diet-induced obesity. Diabetes 54: 1023-1031. KIM, K. H., K. LEE, Y. S. MOON and H. S. SUL, 2001 A cysteine-rich adipose tissue- specific secretory factor inhibits adipocyte differentiation. J Biol Chem 276: 11252-11256. KIM, S., I. SOHN, J. I. AHN, K. H. LEE and Y. S. LEE, 2004 Hepatic gene expression profiles in a long-term high-fat diet-induced obesity mouse model. Gene 340: 99- 109. KISLINGER, T., B. COX, A. KANNAN, C. CHUNG, P. HU et al., 2006 Global survey of organ and organelle protein expression in mouse: combined proteomic and transcriptomic profiling. Cell 125: 173-186. KISSEBAH, A. H., G. E. SONNENBERG, J. MYKLEBUST, M. GOLDSTEIN, K. BROMAN et al., 2000 Quantitative trait loci on chromosomes 3 and 17 influence phenotypes of the metabolic syndrome. Proc Natl Acad Sci U S A 97: 14478-14483. KISSILEFF, H. R., F. X. PI-SUNYER, J. THORNTON and G. P. SMITH, 1981 C-terminal octapeptide of cholecystokinin decreases food intake in man. Am J Clin Nutr 34: 154-160. KLEIN, P. D., W. P. JAMES, W. W. WONG, C. S. IRVING, P. R. MURGATROYD et al., 1984 Calorimetric validation of the doubly-labelled water method for determination of energy expenditure in man. Hum Nutr Clin Nutr 38: 95-106. KOJIMA, M., H. HOSODA, Y. DATE, M. NAKAZATO, H. MATSUO et al., 1999 Ghrelin is a growth-hormone-releasing acylated peptide from stomach. Nature 402: 656-660. KOTOWSKI, I. K., A. PERTSEMLIDIS, A. LUKE, R. S. COOPER, G. L. VEGA et al., 2006 A spectrum of PCSK9 alleles contributes to plasma levels of low-density lipoprotein cholesterol. Am J Hum Genet 78: 410-422. KRAUSS, R. M., R. H. ECKEL, B. HOWARD, L. J. APPEL, S. R. DANIELS et al., 2000 AHA Dietary Guidelines: revision 2000: A statement for healthcare professionals from the Nutrition Committee of the American Heart Association. Circulation 102: 2284-2299. KREWSON, T. D., P. J. SUPELAK, A. E. HILL, J. B. SINGER, E. S. LANDER et al., 2004 Chromosomes 6 and 13 harbor genes that regulate pubertal timing in mouse chromosome substitution strains. Endocrinology 145: 4447-4451. KRUDE, H., H. BIEBERMANN, W. LUCK, R. HORN, G. BRABANT et al., 1998 Severe early- onset obesity, adrenal insufficiency and red hair pigmentation caused by POMC mutations in humans. Nat Genet 19: 155-157. KRUDE, H., H. BIEBERMANN, D. SCHNABEL, M. Z. TANSEK, P. THEUNISSEN et al., 2003 Obesity due to proopiomelanocortin deficiency: three new cases and treatment trials with thyroid hormone and ACTH4-10. J Clin Endocrinol Metab 88: 4633- 4640. KRUDE, H., D. SCHNABEL, W. LUCK and A. GRUTERS, 1999 Implications of the phenotype of POMC deficiency for the role of POMC-derived peptides in skin physiology. Ann N Y Acad Sci 885: 419-421. 273

LANDER, E., and L. KRUGLYAK, 1995 Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 11: 241-247. LANDER, E. S., and D. BOTSTEIN, 1989 Mapping mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185-199. LEE, G. H., R. PROENCA, J. M. MONTEZ, K. M. CARROLL, J. G. DARVISHZADEH et al., 1996 Abnormal splicing of the leptin receptor in diabetic mice. Nature 379: 632- 635. LEE, M. H., K. LU, S. HAZARD, H. YU, S. SHULENIN et al., 2001 Identification of a gene, ABCG5, important in the regulation of dietary cholesterol absorption. Nat Genet 27: 79-83. LEVINE, J. A., 2004 Non-exercise activity thermogenesis (NEAT). Nutr Rev 62: S82-97. LI, B., L. A. NOLTE, J. S. JU, D. H. HAN, T. COLEMAN et al., 2000 Skeletal muscle respiratory uncoupling prevents diet-induced obesity and insulin resistance in mice. Nat Med 6: 1115-1120. LIU, R., W. A. PAXTON, S. CHOE, D. CERADINI, S. R. MARTIN et al., 1996 Homozygous defect in HIV-1 coreceptor accounts for resistance of some multiply-exposed individuals to HIV-1 infection. Cell 86: 367-377. LU, D., D. WILLARD, I. R. PATEL, S. KADWELL, L. OVERTON et al., 1994 Agouti protein is an antagonist of the melanocyte-stimulating-hormone receptor. Nature 371: 799- 802. LUEDI, P. P., A. J. HARTEMINK and R. L. JIRTLE, 2005 Genome-wide prediction of imprinted murine genes. Genome Res 15: 875-884. MACARTHUR, J., 1944 Genetics of body size and related characters. American Naturalist 78: 142-157. MACDONALD, L. E., K. E. WORTLEY, L. C. GOWEN, K. D. ANDERSON, J. D. MURRAY et al., 2005 Resistance to diet-induced obesity in mice globally overexpressing OGH/GPB5. Proc Natl Acad Sci U S A 102: 2496-2501. MAEDA, K., K. OKUBO, I. SHIMOMURA, T. FUNAHASHI, Y. MATSUZAWA et al., 1996 cDNA cloning and expression of a novel adipose specific collagen-like factor, apM1 (AdiPose Most abundant Gene transcript 1). Biochem Biophys Res Commun 221: 286-289. MAES, H. H., M. C. NEALE and L. J. EAVES, 1997 Genetic and environmental factors in relative body weight and human adiposity. Behav Genet 27: 325-351. MALEK, R. L., H. Y. WANG, A. E. KWITEK, A. S. GREENE, N. BHAGABATI et al., 2006 Physiogenomic resources for rat models of heart, lung and blood disorders. Nat Genet 38: 234-239. MANTEY, C., G. A. BROCKMANN, E. KALM and N. REINSCH, 2005 Mapping and Exclusion Mapping of Genomic Imprinting Effects in Mouse F2 Families. J Hered 96: 329-338. MARCHESINI, G., M. BRIZI, G. BIANCHI, S. TOMASSETTI, E. BUGIANESI et al., 2001 Nonalcoholic fatty liver disease: a feature of the metabolic syndrome. Diabetes 50: 1844-1850. MARCHESINI, G., E. BUGIANESI, G. FORLANI, F. CERRELLI, M. LENZI et al., 2003 Nonalcoholic fatty liver, steatohepatitis, and the metabolic syndrome. Hepatology 37: 917-923. 274

MARKEL, P., P. SHU, C. EBELING, G. A. CARLSON, D. L. NAGLE et al., 1997 Theoretical and empirical issues for marker-assisted breeding of congenic mouse strains. Nat Genet 17: 280-284. MARSHALL, J. D., J. L. MU, Y. C. CHEAH, M. N. NESBITT, W. N. FRANKEL et al., 1992 The AXB and BXA set of recombinant inbred mouse strains. Mamm Genome 3: 669-680. MATIN, A., G. B. COLLIN, Y. ASADA, D. VARNUM and J. H. NADEAU, 1999 Susceptibility to testicular germ-cell tumours in a 129.MOLF-Chr 19 chromosome substitution strain. Nat Genet 23: 237-240. MCLAUGHLIN, C. L., C. A. BAILE and F. C. BUONOMO, 1985 Effect of CCK antibodies on food intake and weight gain in Zucker rats. Physiol Behav 34: 277-282. MCQUEEN, M. B., L. BERTRAM, E. B. RIMM, D. BLACKER and S. L. SANTANGELO, 2003 A QTL genome scan of the metabolic syndrome and its component traits. BMC Genet 4 Suppl 1: S96. MEHRABIAN, M., H. ALLAYEE, J. STOCKTON, P. Y. LUM, T. A. DRAKE et al., 2005 Integrating genotypic and expression data in a segregating mouse population to identify 5-lipoxygenase as a susceptibility gene for obesity and bone traits. Nat Genet 37: 1224-1233. MICHAUD, E. J., S. J. BULTMAN, L. J. STUBBS and R. P. WOYCHIK, 1993 The embryonic lethality of homozygous lethal yellow mice (Ay/Ay) is associated with the disruption of a novel RNA-binding protein. Genes Dev 7: 1203-1213. MIETTINEN, T. A., R. S. TILVIS and Y. A. KESANIEMI, 1990 Serum plant sterols and cholesterol precursors reflect cholesterol absorption and synthesis in volunteers of a randomly selected male population. Am J Epidemiol 131: 20-31. MILLER, M. W., D. M. DUHL, H. VRIELING, S. P. CORDES, M. M. OLLMANN et al., 1993 Cloning of the mouse agouti gene predicts a secreted protein ubiquitously expressed in mice carrying the lethal yellow mutation. Genes Dev 7: 454-467. MILLS, E., C. M. KUHN, M. N. FEINGLOS and R. SURWIT, 1993 Hypertension in CB57BL/6J mouse model of non-insulin-dependent diabetes mellitus. Am J Physiol 264: R73-78. MIZUNO, T. M., S. P. KLEOPOULOS, H. T. BERGEN, J. L. ROBERTS, C. A. PRIEST et al., 1998 Hypothalamic pro-opiomelanocortin mRNA is reduced by fasting and [corrected] in ob/ob and db/db mice, but is stimulated by leptin. Diabetes 47: 294- 297. MIZUNO, T. M., and C. V. MOBBS, 1999 Hypothalamic agouti-related protein messenger ribonucleic acid is inhibited by leptin and stimulated by fasting. Endocrinology 140: 814-817. MOKDAD, A. H., J. S. MARKS, D. F. STROUP and J. L. GERBERDING, 2004 Actual causes of death in the United States, 2000. JAMA 291: 1238-1245. MOKDAD, A. H., J. S. MARKS, D. F. STROUP and J. L. GERBERDING, 2005 Correction: actual causes of death in the United States, 2000. JAMA 293: 293-294. MOOTHA, V. K., J. BUNKENBORG, J. V. OLSEN, M. HJERRILD, J. R. WISNIEWSKI et al., 2003 Integrated analysis of protein composition, tissue diversity, and gene regulation in mouse mitochondria. Cell 115: 629-640. MORAES, R. C., A. BLONDET, K. BIRKENKAMP-DEMTROEDER, J. TIRARD, T. F. ORNTOFT et al., 2003 Study of the alteration of gene expression in adipose tissue of diet- 275

induced obese mice by microarray and reverse transcription-polymerase chain reaction analyses. Endocrinology 144: 4773-4782. MORRIS, K. H., A. ISHIKAWA and P. D. KEIGHTLEY, 1999 Quantitative trait loci for growth traits in C57BL/6J x DBA/2J mice. Mamm Genome 10: 225-228. NADEAU, J. H., 2001 Modifier genes in mice and humans. Nat Rev Genet 2: 165-174. NADEAU, J. H., 2003 Modifier genes and protective alleles in humans and mice. Curr Opin Genet Dev 13: 290-295. NADEAU, J. H., J. B. SINGER, A. MATIN and E. S. LANDER, 2000 Analysing complex genetic traits with chromosome substitution strains. Nat Genet 24: 221-225. NAGGERT, J. K., L. D. FRICKER, O. VARLAMOV, P. M. NISHINA, Y. ROUILLE et al., 1995 Hyperproinsulinaemia in obese fat/fat mice associated with a carboxypeptidase E mutation which reduces enzyme activity. Nat Genet 10: 135-142. NAKANO, Y., T. TOBE, N. H. CHOI-MIURA, T. MAZDA and M. TOMITA, 1996 Isolation and characterization of GBP28, a novel gelatin-binding protein purified from human plasma. J Biochem (Tokyo) 120: 803-812. NEEL, J. V., 1962 Diabetes mellitus: a "thrifty" genotype rendered detrimental by "progress"? Am J Hum Genet 14: 353-362. NESBITT, M. N., and E. SKAMENE, 1984 Recombinant inbred mouse strains derived from A/J and C57BL/6J: a tool for the study of genetic mechanisms in host resistance to infection and malignancy. J Leukoc Biol 36: 357-364. OHLEMILLER, K. K., R. M. HUGHES, J. MOSINGER-OGILVIE, J. D. SPECK, D. H. GROSOF et al., 1995 Cochlear and retinal degeneration in the tubby mouse. Neuroreport 6: 845-849. OLLMANN, M. M., B. D. WILSON, Y. K. YANG, J. A. KERNS, Y. CHEN et al., 1997 Antagonism of central melanocortin receptors in vitro and in vivo by agouti- related protein. Science 278: 135-138. O'RAHILLY, S., H. GRAY, P. J. HUMPHREYS, A. KROOK, K. S. POLONSKY et al., 1995 Brief report: impaired processing of prohormones associated with abnormalities of glucose homeostasis and adrenal function. N Engl J Med 333: 1386-1390. PAPAVRAMIDOU, N. S., S. T. PAPAVRAMIDIS and H. CHRISTOPOULOU-ALETRA, 2004 Galen on obesity: etiology, effects, and treatment. World J Surg 28: 631-635. PAREKH, P. I., A. E. PETRO, J. M. TILLER, M. N. FEINGLOS and R. S. SURWIT, 1998 Reversal of diet-induced obesity and diabetes in C57BL/6J mice. Metabolism 47: 1089-1096. PERUSSE, L., T. RANKINEN, A. ZUBERI, Y. C. CHAGNON, S. J. WEISNAGEL et al., 2005 The human obesity gene map: the 2004 update. Obes Res 13: 381-490. PETRYSHEN, T. L., A. KIRBY, R. P. HAMMER, JR., S. PURCELL, S. B. O'LEARY et al., 2005 Two quantitative trait loci for prepulse inhibition of startle identified on mouse chromosome 16 using chromosome substitution strains. Genetics 171: 1895-1904. PRENTICE, A. M., P. RAYCO-SOLON and S. E. MOORE, 2005 Insights from the developing world: thrifty genotypes and thrifty phenotypes. Proc Nutr Soc 64: 153-161. PRPIC, V., P. M. WATSON, I. C. FRAMPTON, M. A. SABOL, G. E. JEZEK et al., 2003 Differential mechanisms and development of leptin resistance in A/J versus C57BL/6J mice during diet-induced obesity. Endocrinology 144: 1155-1163. 276

QUILLENT, C., E. OBERLIN, J. BRAUN, D. ROUSSET, G. GONZALEZ-CANALI et al., 1998 HIV-1-resistance phenotype conferred by combination of two separate inherited mutations of CCR5 gene. Lancet 351: 14-18. RASHID, S., D. E. CURTIS, R. GARUTI, N. N. ANDERSON, Y. BASHMAKOV et al., 2005 Decreased plasma cholesterol and hypersensitivity to statins in mice lacking Pcsk9. Proc Natl Acad Sci U S A 102: 5374-5379. RAVUSSIN, E., M. E. VALENCIA, J. ESPARZA, P. H. BENNETT and L. O. SCHULZ, 1994 Effects of a traditional lifestyle on obesity in Pima Indians. Diabetes Care 17: 1067-1074. RAZANI, B., T. P. COMBS, X. B. WANG, P. G. FRANK, D. S. PARK et al., 2002 Caveolin-1- deficient mice are lean, resistant to diet-induced obesity, and show hypertriglyceridemia with adipocyte abnormalities. J Biol Chem 277: 8635-8647. REAVEN, G. M., 1988 Banting lecture 1988. Role of insulin resistance in human disease. Diabetes 37: 1595-1607. REBUFFE-SCRIVE, M., R. SURWIT, M. FEINGLOS, C. KUHN and J. RODIN, 1993 Regional fat distribution and metabolism in a new mouse model (C57BL/6J) of non- insulin-dependent diabetes mellitus. Metabolism 42: 1405-1409. REED, D. R., X. LI, A. H. MCDANIEL, K. LU, S. LI et al., 2003 Loci on chromosomes 2, 4, 9, and 16 for body weight, body length, and adiposity identified in a genome scan of an F2 intercross between the 129P3/J and C57BL/6ByJ mouse strains. Mamm Genome 14: 302-313. REIDELBERGER, R. D., and T. E. SOLOMON, 1986 Comparative effects of CCK-8 on feeding, sham feeding, and exocrine pancreatic secretion in rats. Am J Physiol 251: R97-105. REIFSNYDER, P. C., G. CHURCHILL and E. H. LEITER, 2000 Maternal environment and genotype interact to establish diabesity in mice. Genome Res 10: 1568-1578. ROCHA, J. L., E. J. EISEN, L. D. VAN VLECK and D. POMP, 2004 A large-sample QTL study in mice: I. Growth. Mamm Genome 15: 83-99. ROVERE, C., A. VIALE, J. NAHON and P. KITABGI, 1996 Impaired processing of brain proneurotensin and promelanin-concentrating hormone in obese fat/fat mice. Endocrinology 137: 2954-2958. SALMON, D. M., and J. P. FLATT, 1985 Effect of dietary fat content on the incidence of obesity among ad libitum fed mice. Int J Obes 9: 443-449. SAMSON, M., F. LIBERT, B. J. DORANZ, J. RUCKER, C. LIESNARD et al., 1996 Resistance to HIV-1 infection in caucasian individuals bearing mutant alleles of the CCR-5 chemokine receptor gene. Nature 382: 722-725. SANTAGATA, S., T. J. BOGGON, C. L. BAIRD, C. A. GOMEZ, J. ZHAO et al., 2001 G-protein signaling through tubby proteins. Science 292: 2041-2050. SCHERER, P. E., S. WILLIAMS, M. FOGLIANO, G. BALDINI and H. F. LODISH, 1995 A novel serum protein similar to C1q, produced exclusively in adipocytes. J Biol Chem 270: 26746-26749. SCHOELLER, D. A., E. RAVUSSIN, Y. SCHUTZ, K. J. ACHESON, P. BAERTSCHI et al., 1986 Energy expenditure by doubly labeled water: validation in humans and proposed calculation. Am J Physiol 250: R823-830. SCHWARTZ, M. W., D. G. BASKIN, T. R. BUKOWSKI, J. L. KUIJPER, D. FOSTER et al., 1996a Specificity of leptin action on elevated blood glucose levels and 277

hypothalamic neuropeptide Y gene expression in ob/ob mice. Diabetes 45: 531- 535. SCHWARTZ, M. W., J. L. MARKS, A. J. SIPOLS, D. G. BASKIN, S. C. WOODS et al., 1991 Central insulin administration reduces neuropeptide Y mRNA expression in the arcuate nucleus of food-deprived lean (Fa/Fa) but not obese (fa/fa) Zucker rats. Endocrinology 128: 2645-2647. SCHWARTZ, M. W., R. J. SEELEY, L. A. CAMPFIELD, P. BURN and D. G. BASKIN, 1996b Identification of targets of leptin action in rat hypothalamus. J Clin Invest 98: 1101-1106. SCHWARTZ, M. W., R. J. SEELEY, S. C. WOODS, D. S. WEIGLE, L. A. CAMPFIELD et al., 1997 Leptin increases hypothalamic pro-opiomelanocortin mRNA expression in the rostral arcuate nucleus. Diabetes 46: 2119-2123. SCHWARTZ, M. W., A. J. SIPOLS, J. L. MARKS, G. SANACORA, J. D. WHITE et al., 1992 Inhibition of hypothalamic neuropeptide Y gene expression by insulin. Endocrinology 130: 3608-3616. SCHWARTZ, M. W., S. C. WOODS, D. PORTE, JR., R. J. SEELEY and D. G. BASKIN, 2000 Central nervous system control of food intake. Nature 404: 661-671. SHMULEWITZ, D., S. C. HEATH, M. L. BLUNDELL, Z. HAN, R. SHARMA et al., 2006 Inaugural Article: Linkage analysis of quantitative traits for obesity, diabetes, hypertension, and dyslipidemia on the island of Kosrae, Federated States of Micronesia. Proc Natl Acad Sci U S A 103: 3502-3509. SHORT, T., 1727 Discourse on the courses and effects of corpulency together with the method for its prevention and cure. J. Roberts, London. SILVER, L. M., 1995 Mouse Genetics: Concepts and Applications, pp. Oxford University Press, New York. SINGER, J. B., A. E. HILL, L. C. BURRAGE, K. R. OLSZENS, J. SONG et al., 2004 Genetic dissection of complex traits with chromosome substitution strains of mice. Science 304: 445-448. SINGER, J. B., A. E. HILL, J. H. NADEAU and E. S. LANDER, 2005 Mapping quantitative trait loci for anxiety in chromosome substitution strains of mice. Genetics 169: 855-862. SNELL, G. D., 1948 Methods for the study of histocompatability genes. J Genet 49: 87- 103. SOUKAS, A., N. D. SOCCI, B. D. SAATKAMP, S. NOVELLI and J. M. FRIEDMAN, 2001 Distinct transcriptional profiles of adipogenesis in vivo and in vitro. J Biol Chem 276: 34167-34174. SPARKS, L. M., H. XIE, R. A. KOZA, R. MYNATT, M. W. HULVER et al., 2005 A high-fat diet coordinately downregulates genes required for mitochondrial oxidative phosphorylation in skeletal muscle. Diabetes 54: 1926-1933. STANLEY, B. G., S. E. KYRKOULI, S. LAMPERT and S. F. LEIBOWITZ, 1986 Neuropeptide Y chronically injected into the hypothalamus: a powerful neurochemical inducer of hyperphagia and obesity. Peptides 7: 1189-1192. STEPHENS, T. W., M. BASINSKI, P. K. BRISTOW, J. M. BUE-VALLESKEY, S. G. BURGETT et al., 1995 The role of neuropeptide Y in the antiobesity action of the obese gene product. Nature 377: 530-532. 278

STEPPAN, C. M., S. T. BAILEY, S. BHAT, E. J. BROWN, R. R. BANERJEE et al., 2001 The hormone resistin links obesity to diabetes. Nature 409: 307-312. STROBEL, A., T. ISSAD, L. CAMOIN, M. OZATA and A. D. STROSBERG, 1998 A leptin missense mutation associated with hypogonadism and morbid obesity. Nat Genet 18: 213-215. STUBDAL, H., C. A. LYNCH, A. MORIARTY, Q. FANG, T. CHICKERING et al., 2000 Targeted deletion of the tub mouse obesity gene reveals that tubby is a loss-of-function mutation. Mol Cell Biol 20: 878-882. STUNKARD, A. J., T. T. FOCH and Z. HRUBEC, 1986 A twin study of human obesity. JAMA 256: 51-54. STUNKARD, A. J., J. R. HARRIS, N. L. PEDERSEN and G. E. MCCLEARN, 1990 The body- mass index of twins who have been reared apart. N Engl J Med 322: 1483-1487. STYLIANOU, I. M., R. KORSTANJE, R. LI, S. SHEEHAN, B. PAIGEN et al., 2006 Quantitative trait locus analysis for obesity reveals multiple networks of interacting loci. Mamm Genome 17: 22-36. SUBRAMANIAN, A., P. TAMAYO, V. K. MOOTHA, S. MUKHERJEE, B. L. EBERT et al., 2005 Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102: 15545-15550. SURWIT, R. S., C. L. EDWARDS, S. MURTHY and A. E. PETRO, 2000 Transient effects of long-term leptin supplementation in the prevention of diet-induced obesity in mice. Diabetes 49: 1203-1208. SURWIT, R. S., M. N. FEINGLOS, J. RODIN, A. SUTHERLAND, A. E. PETRO et al., 1995 Differential effects of fat and sucrose on the development of obesity and diabetes in C57BL/6J and A/J mice. Metabolism 44: 645-651. SURWIT, R. S., C. M. KUHN, C. COCHRANE, J. A. MCCUBBIN and M. N. FEINGLOS, 1988 Diet-induced type II diabetes in C57BL/6J mice. Diabetes 37: 1163-1167. SURWIT, R. S., A. E. PETRO, P. PAREKH and S. COLLINS, 1997 Low plasma leptin in response to dietary fat in diabetes- and obesity-prone mice. Diabetes 46: 1516- 1520. TAKAHASHI, M., Y. KAMEI and O. EZAKI, 2005 Mest/Peg1 imprinted gene enlarges adipocytes and is a marker of adipocyte size. Am J Physiol Endocrinol Metab 288: E117-124. TANG, W., M. B. MILLER, S. S. RICH, K. E. NORTH, J. S. PANKOW et al., 2003 Linkage analysis of a composite factor for the multiple metabolic syndrome: the National Heart, Lung, and Blood Institute Family Heart Study. Diabetes 52: 2840-2847. TANG-CHRISTENSEN, M., P. J. LARSEN, R. GOKE, A. FINK-JENSEN, D. S. JESSOP et al., 1996 Central administration of GLP-1-(7-36) amide inhibits food and water intake in rats. Am J Physiol 271: R848-856. TARTAGLIA, L. A., M. DEMBSKI, X. WENG, N. DENG, J. CULPEPPER et al., 1995 Identification and expression cloning of a leptin receptor, OB-R. Cell 83: 1263- 1271. THORNTON, J. E., C. C. CHEUNG, D. K. CLIFTON and R. A. STEINER, 1997 Regulation of hypothalamic proopiomelanocortin mRNA by leptin in ob/ob mice. Endocrinology 138: 5063-5066. 279

TSCHOP, M., T. R. CASTANEDA, H. G. JOOST, C. THONE-REINEKE, S. ORTMANN et al., 2004 Physiology: does gut hormone PYY3-36 decrease food intake in rodents? Nature 430: 1 p following 165; discussion 162 p following 165. TSCHOP, M., and M. L. HEIMAN, 2001 Rodent obesity models: an overview. Exp Clin Endocrinol Diabetes 109: 307-319. TSCHOP, M., D. L. SMILEY and M. L. HEIMAN, 2000 Ghrelin induces adiposity in rodents. Nature 407: 908-913. TSUJII, S., and G. A. BRAY, 1989 Acetylation alters the feeding response to MSH and beta-endorphin. Brain Res Bull 23: 165-169. TURTON, M. D., D. O'SHEA, I. GUNN, S. A. BEAK, C. M. EDWARDS et al., 1996 A role for glucagon-like peptide-1 in the central regulation of feeding. Nature 379: 69-72. TUSHER, V. G., R. TIBSHIRANI and G. CHU, 2001 Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98: 5116- 5121. UDUPI, V., P. GOMEZ, L. SONG, O. VARLAMOV, J. T. REED et al., 1997 Effect of carboxypeptidase E deficiency on progastrin processing and gastrin messenger ribonucleic acid expression in mice with the fat mutation. Endocrinology 138: 1959-1963. VAISSE, C., K. CLEMENT, B. GUY-GRAND and P. FROGUEL, 1998 A frameshift mutation in human MC4R is associated with a dominant form of obesity. Nat Genet 20: 113- 114. VAN SCHOTHORST, E. M., N. FRANSSEN-VAN HAL, M. M. SCHAAP, J. PENNINGS, B. HOEBEE et al., 2005 Adipose gene expression patterns of weight gain suggest counteracting steroid hormone synthesis. Obes Res 13: 1031-1041. VAZQUEZ, J. A., E. L. MORSE and S. A. ADIBI, 1985 Effect of dietary fat, carbohydrate, and protein on branched-chain amino acid catabolism during caloric restriction. J Clin Invest 76: 737-743. VLAHAKIS, G., and W. E. HESTON, 1963 Increase of Induced Skin Tumors in the Mouse by the Lethal Yellow Gene (Ay). J Natl Cancer Inst 31: 189-195. VREKEN, P., A. E. VAN LINT, A. H. BOOTSMA, H. OVERMARS, R. J. WANDERS et al., 1999 Quantitative plasma acylcarnitine analysis using electrospray tandem mass spectrometry for the diagnosis of organic acidaemias and fatty acid oxidation defects. J Inherit Metab Dis 22: 302-306. WAKELAND, E., L. MOREL, K. ACHEY, M. YUI and J. LONGMATE, 1997 Speed congenics: a classic technique in the fast lane (relatively speaking). Immunol Today 18: 472- 477. WANLESS, I. R., and J. S. LENTZ, 1990 Fatty liver hepatitis (steatohepatitis) and obesity: an autopsy study with analysis of risk factors. Hepatology 12: 1106-1110. WARDEN, C. H., N. YI and J. FISLER, 2004 Epistasis among genes is a universal phenomenon in obesity: evidence from rodent models. Nutrition 20: 74-77. WATSON, P. M., S. P. COMMINS, R. J. BEILER, H. C. HATCHER and T. W. GETTYS, 2000 Differential regulation of leptin expression and function in A/J vs. C57BL/6J mice during diet-induced obesity. Am J Physiol Endocrinol Metab 279: E356-365. WEST, D. B., D. FEY and S. C. WOODS, 1984 Cholecystokinin persistently suppresses meal size but not food intake in free-feeding rats. Am J Physiol 246: R776-787. 280

WEST, D. B., J. GOUDEY-LEFEVRE, B. YORK and G. E. TRUETT, 1994a Dietary obesity linked to genetic loci on chromosomes 9 and 15 in a polygenic mouse model. J Clin Invest 94: 1410-1416. WEST, D. B., J. WAGUESPACK, B. YORK, J. GOUDEY-LEFEVRE and R. A. PRICE, 1994b Genetics of dietary obesity in AKR/J x SWR/J mice: segregation of the trait and identification of a linked locus on chromosome 4. Mamm Genome 5: 546-552. WESTERTERP, K. R., W. H. SARIS, M. VAN ES and F. TEN HOOR, 1986 Use of the doubly labeled water technique in humans during heavy sustained exercise. J Appl Physiol 61: 2162-2167. WHO, 1997 Obesity: Preventing and Managing the Global Epidemic - Report of a WHO Consultation on Obesity, 3-5 June 1997, Geneva, WHO/NUT/NCD/98.1. WILLETT, W. C., W. H. DIETZ and G. A. COLDITZ, 1999 Guidelines for healthy weight. N Engl J Med 341: 427-434. WILSON, B. D., D. BAGNOL, C. B. KAELIN, M. M. OLLMANN, I. GANTZ et al., 1999 Physiological and anatomical circuitry between Agouti-related protein and leptin signaling. Endocrinology 140: 2387-2397. WILTSHIRE, T., M. T. PLETCHER, S. BATALOV, S. W. BARNES, L. M. TARANTINO et al., 2003 Genome-wide single-nucleotide polymorphism analysis defines haplotype patterns in mouse. Proc Natl Acad Sci U S A 100: 3380-3385. WILUND, K. R., L. YU, F. XU, G. L. VEGA, S. M. GRUNDY et al., 2004 No association between plasma levels of plant sterols and atherosclerosis in mice and men. Arterioscler Thromb Vasc Biol 24: 2326-2332. WOODS, S. C., E. C. LOTTER, L. D. MCKAY and D. PORTE, JR., 1979 Chronic intracerebroventricular infusion of insulin reduces food intake and body weight of baboons. Nature 282: 503-505. WU, H., KERR, M.K., CUI, X., CHURCHILL, G.A., 2003 MAANOVA: A software package for the analysis of spotted cDNA microarray experiments, pp. 313-341 in The Analysis of Gene Exression Data: Methods and Software, edited by G. PARMIGIANI, GARETT, E.S., IRIZARRY, R.A., ZEGER, S.L. Springer-Verlag, New York. YAN, L. L., M. L. DAVIGLUS, K. LIU, J. STAMLER, R. WANG et al., 2006 Midlife body mass index and hospitalization and mortality in older age. JAMA 295: 190-198. YEN, T. T., M. M. MCKEE and N. B. STAMM, 1984 Thermogenesis and weight control. Int J Obes 8 Suppl 1: 65-78. YEO, G. S., I. S. FAROOQI, S. AMINIAN, D. J. HALSALL, R. G. STANHOPE et al., 1998 A frameshift mutation in MC4R associated with dominantly inherited human obesity. Nat Genet 20: 111-112. YI, N., A. DIAMENT, S. CHIU, K. KIM, D. B. ALLISON et al., 2004 Characterization of epistasis influencing complex spontaneous obesity in the BSB model. Genetics 167: 399-409. YI, N., D. K. ZINNIEL, K. KIM, E. J. EISEN, A. BARTOLUCCI et al., 2006 Bayesian analyses of multiple epistatic QTL models for body weight and body composition in mice. Genet Res 87: 45-60. YORK, B., K. LEI and D. B. WEST, 1996 Sensitivity to dietary obesity linked to a locus on chromosome 15 in a CAST/Ei x C57BL/6J F2 intercross. Mamm Genome 7: 677- 681. 281

YORK, B., K. LEI and D. B. WEST, 1997 Inherited non-autosomal effects on body fat in F2 mice derived from an AKR/J x SWR/J cross. Mamm Genome 8: 726-730. YOUNGREN, K. K., J. H. NADEAU and A. MATIN, 2003 Testicular cancer susceptibility in the 129.MOLF-Chr19 mouse strain: additive effects, gene interactions and epigenetic modifications. Hum Mol Genet 12: 389-398. YU, S., O. GAVRILOVA, H. CHEN, R. LEE, J. LIU et al., 2000 Paternal versus maternal transmission of a stimulatory G-protein alpha subunit knockout produces opposite effects on energy metabolism. J Clin Invest 105: 615-623. ZHANG, J. V., P. G. REN, O. AVSIAN-KRETCHMER, C. W. LUO, R. RAUCH et al., 2005 Obestatin, a peptide encoded by the ghrelin gene, opposes ghrelin's effects on food intake. Science 310: 996-999. ZHANG, S., and H. K. GERSHENFELD, 2003 Genetic contributions to body weight in mice: relationship of exploratory behavior to weight. Obes Res 11: 828-838. ZHANG, Y., R. PROENCA, M. MAFFEI, M. BARONE, L. LEOPOLD et al., 1994 Positional cloning of the mouse obese gene and its human homologue. Nature 372: 425-432. ZHOU, Z., S. YON TOH, Z. CHEN, K. GUO, C. P. NG et al., 2003 Cidea-deficient mice have lean phenotype and are resistant to obesity. Nat Genet 35: 49-56.