University of Southern California Dissertations and Theses

Genetic variations in gene from the cytochrome P450 family may account for differential response to pioglitazone therapy in the Hispanic women

(USC Thesis Other)

Genetic variations in gene from the cytochrome P450 family may account for differential response to pioglitazone therapy in the Hispanic women

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i
GENETIC V ARIATIONS IN GENE FROM THE CYTOCHROME P450 FAMILY MAY
ACCOUNT FOR DIFFERENTIAL RESPONSE TO PIOGLITAZONE THERAPY IN
THE HISPANIC WOMEN

by

Yen-Fu Chen

____________________________________________________________________

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(BIOSTATISTICS)

December 2009

Copyright 2009 Yen-Fu Chen

ii
Acknowledgments
I would like to express my appreciation to Dr. Richard Watanabe, chair of my thesis
committee, for his guidance and support throughout the research and writing. I would like
to express my gratitude to Dr. Hooman Allayee and Dr. James Gauderman for serving as
my committee members and their valuable suggestions. My heart and earnest thanks to
my wife, Katy, the most important person in my life.

And welcome Scarlett to this world.
iii
Table of Contents
Acknowledgments ii
List of Tables iv
Abbreviations vi
Abstract vii
Introduction 1
Research Design and Methods 6
Subjects 6
Experimental Protocol for Phenotypes 7
Genotyping Protocol 7
Identification of Genetic Variant and SNP picking 7
Quality control of genotyping data 10
Tag SNP selection and genotyping imputation 11
Statistical Analysis 12
Results 14
Basic Characteristics 14
PIO response – Principal Component Analysis 16
PC Association with Multiple Phenotypes 17
PIO response – SNP association 23
SNP Association with Multiple Phenotypes 23
Discussion 28
References 33
Appendix: Table A1: List of Imputed Genotypes 36
iv
List of Tables
Table1: Summary of SNP sets selected by SNPwalk 10
Table2: Phenotype values at baseline and after one year of PIO
treatment in responders and non-responders. Values are median
and interquartile range 15
Table3: One-year change in phenotypes in responders and
non-reponders to PIO. Values are median and interquartile range 16
Table4: Principal Component Analysis of tagSNPs in the CYP3A4 gene
region (N=49) 19
Table5: Principal Component Analysis of tagSNPs in the CYP2C8 gene
region (N=50) 19
Table6: Principal Component Analysis of tagSNPs in the CYP2A9 gene
region (N=54) 20
Table7: Association between PCs in CYP3A4 and PIO Response 20
Table8: Association of PCs in CYP2C8 and PIO Response 21
Table9: Association of PCs in CYP2C9 and PIO Response 21
Table10: Association Between PCs and 1-year change in Phenotypes 22
Table11: Association Between SNPs in CYP3A4 and PIO Response 25
Table12: Association Between SNPs in the CYP2C8 and PIO Response 25
Table13: Association Between SNPs in the CYP2C9 and PIO Response 26
v
Table14: Association Between SNPs in CYP3A4 and 1-year Change in
Phenotypes clinical responses 26
Table 15: Association Between SNPs in CYP2C8 and 1-year Change in
Phenotypes 27
Table 16: Association Between SNP in CYP2C9 and 1-year Change in
Phenotypes 27

vi
Abbreviations

AIR Acute insulin response to glucose
BMI Body Mass Index
CYP2C8 Cytochrome P450, Family2, Subfamily C, Polypeptide 8
CYP2C9 Cytochrome P450, Family2, Subfamily C, Polypeptide 9
CYP3A4 Cytochrome P450, Family3, Subfamily A, Polypeptide 4
DI Disposition Index
GDM Gestational Diabetes Mellitus
IVGTT Intravenous Glucose Tolerance Test
OGTT Oral Glucose Tolerance Test
LD Linkage Disequilibrium
PIO Pioglitazone
PIPOD Pioglitazone in the Prevention of Diabetes
PPARG peroxisome proliferator-activated receptor-γ
S
I
Insulin Sensitivity
SNP Single Nucleotide Polymorphism
TRIPOD Troglitazone in the Prevention of Diabetes
vii
TRO Troglitazone
T2DM Type 2 Diabetes Mellitus
viii
Abstract
Studies have shown that in addition to treating T2DM, TZDs may be effective in the
prevention of T2DM as first demonstrated in the TRIPOD and PIPOD studies. However,
both studies showed that 30%-40% subjects do not respond to TZDs. To determine
whether genetic variants in CYP gene family contribute to non-response, 102 SNPs in
CYP2C8, CYP2C9, and CYP3A4 were genotyped and examined for the association with
PIO response and clinical phenotypes. Our results show SNP rs4646440 in CYP3A4 is
associated with PIO response. Two additional SNPs, rs1934980 and rs7916699 also
showed trend for association with PIO response. rs1934980 in CYP2C8 and rs10509679
in CYP2C9 showed trend of association with 1-year change in phenotypes. Our study
suggests that polymorphisms in CYP gene family are associated with differential PIO
response. These results support a polygenic model regarding differential response to TZD
therapy in Hispanic women.
1
Introduction
Thiazolidinediones (TZDs) are an insulin-sensitizing class of drug introduced in
last decade to treat type 2 diabetes mellitus (T2DM) and initially included three
compound derivatives Pioglitazone (PIO), Troglitazone (TRO), and Rosiglitazone.
TZDs are agonists for peroxisome proliferators-activated receptor γ (PPARG) and
are thought to improve insulin sensitivity by stimulating adipogenesis and reducing
circulating triglycerides. The amelioration of insulin resistant reduces insulin
secretory demands and preserve pancreatic β-cell function[1] [2]. The Troglitazone
in the Prevention of Diabetes (TRIPOD) study tested whether the amelioration of
insulin resistance through TZD therapy would preserve pancreatic β-cell function,
and delay, or ultimately prevent the onset of T2DM in Hispanic women [1].
Two-hundred-sixty-six Hispanic women with history of gestational diabetes mellitus
(GDM) were recruited in the TRIPOD study and randomized to treatment and
placebo groups. Incidence of T2DM and measure of insulin sensitivity (S
I
), β-cell
function and other phenotypes were followed for a median of 42 months and
measured. The study showed that S
I
increased an average of 88% after only three
months of treatment and T2DM incidence was reduced 55% in women who received
2
TRO. The preventive effect persisted eight months after treatment had been
discontinued, despite the fact that S
I
decreased after termination of TRO therapy [1].
These findings suggest that TRO may delay or even prevent the onset of T2DM.
However, in March 2000, TRO was withdrawn from the market due to
increased risk for liver tumors. A follow-up study, called the Pioglitazone in
Prevention of Diabetes (PIPOD) was conducted to examine if PIO could also prevent
or delay T2DM in individuals who previously participated in the TRIPOD study [2].
The results show that PIO stopped the decline of β-cell function in women from the
placebo group, as well as maintain the S
I
in women from treatment group from the
TRIPOD study. The risk of diabetes was lowest in women with the largest reduction
in IVGTT insulin area after one year of treatment. These two results hint that TZDs
could have effects to reduce insulin resistance and insulin secretory demands,
thereby preserving β-cell function. Subsequent multi-center trials have replicated
the findings from the TRIPOD and PIPOD studies [1, 3-8].
Though the results from previous studies are promising in terms of long-term
treatment and prevention of T2DM, there are still a significant portion of people who
do not respond to TRO or PIO treatment [1, 3, 4, 9]. After receiving three months of
3
treatment in the TRIPOD study, one-third of women in treatment group show no
statistically significant change in S
I
. Overall, about 30% of women did not respond
to TRO treatment during the entire trial period. Subject characteristics including
baseline BMI, weight, fasting glucose, insulin, systolic blood pressure (SBP),
diastolic blood pressure (DBP), acute insulin response to glucose (AIR), S
I
,
disposition index (DI) did not predict who would be a responder or non-responder[3]
[4]. Given the inability of demographic or baseline clinical measure to predict
response, we hypothesized that one or more genetic variants may contribute to
differential response to TZDs [9]. Since TZDs act as agonists to PP ARG and
stimulate adipogenesis [10], a convenient and reasonable hypothesis is that a genetic
coding variation in PP ARG may be the responsible for the differential TZD response.
The common Pro12Ala variant in PP ARG, a known T2DM susceptibility variant,
was tested in TRIPOD subjects [11]. The results showed that this particular
polymorphism was not associated with response to TRO or any TRO-induced change
in insulin sensitivity [11].
However, this study could not rule out the possibility that other polymorphisms
in PP ARG or polymorphisms in other genes could contribute to the differential
4
response to TRO. Therefore, a detailed sequencing effort on PP ARG was performed
to identify additional variants that may contribute to differential response to TRO [9].
One-hundred-thirty-three variants including 127 single nucleotide polymorphism
(SNPs) and 6 insertion deletions were identified and 8 variants showed evidence for
associated with TRO response (p<0.05). Three polymorphisms were also associated
with 3-month changes in S
I
. The results suggest that the genetic variation in PP ARG
may partly underlie response to TRO in Hispanic women at high risk for T2DM. A
similar study examining response to PIO showed that 5 out of the 8 variants showing
association with response to TRO also show association with PIO response (p<0.05)
[2]. 7 variants showed evidence of association with changes in clinical phenotypes
after 1-year study period [12]. These results show that the differential response to
PIO or TRO may be associated with more than one genetic variant and that genetic
variants in genes other than PP ARG are associated with the differential response to
TZDs.
Cytochrome P450 is a very large and diverse family enzymes, which includes
more than 15 sub-families and 57 proteins encoded by different genes and is
primarily responsible for drug metabolism in humans. Cytochrome P450, Family2,
5
Subfamily C, Polypeptide 8 (CYP2C8), Cytochrome P450, Family2, Subfamily C,
Polypeptide 9 (CYP2C9), and Cytochrome P450, Family3, Subfamily A, Polypeptide
4 (CYP3C4) are related to TZDs metabolism [13, 14]. CYP2C8 and CYP3A4 have
been found to metabolize PIO and TRO in vivo[13], while CYP2C8 possesses the
ability to catalyze Rosiglitazone [15, 16]. Therefore, members of cytochrome P450
family serve as legitimate target genes, which may account for differential response
to PIO or TRO. The goal of this study is to examine whether polymorphisms in these
three genes are associated with response to PIO and corresponding changes in
phenotypes.
6
Research Design and Methods
Subjects
The PIPOD design and subject recruitment have been described previously in
detail [1]. In brief, 277 Hispanic women of Mexican, Guatemalan or Salvadoran
descent with the history of GDM were recruited for TRIPOD study between August
1995 and May 1998. Two-hundred-seventeen were randomized to receive TRO
treatment or placebo. Ninety-five women who completed and did not have diabetes
at the end of TRIPOD study, and whose HbA
1C
(A1C) was lower than 7% were
eligible to participate in the PIPOD study, regardless of whether they were in the
placebo or treatment arms in TRIPOD.
Subjects in PIPOD were treated with 30 mg/day PIO for two months, and then
increased to 45mg/day for the reminder of a 3-year treatment period. If during the
course of treatment their A1C exceeded 7%, then final testing was performed and the
subject was referred to treatment of diabetes. For the current study, we only include
subjects who have complete phenotypes at baseline and at one year of treatment and
have available DNA (n=56). Written informed consent was obtained from all
subjects prior to participation in the TRIPOD or PIPOD study. Both studies were
7
approved by the Institutional Review Board of University of Southern California.
Experimental Protocol for Phenotypes
Subjects in PIPOD study received PIO (30 mg/day for first two months, then 45
mg/day for three years). OGTTs and IVGTTs were repeated after one year of
treatment. DNA for these subjects was collected during TRIPOD study.
Genotyping Protocol
The detail genotyping protocols are described in detail elsewhere [17]. In Brief,
all SNPs were assayed with the Illumina GoldenGate Assay (Illumina Inc, San Deigo,
CA) . Primers and multiplex conditions were designed to use the Assay Design Tool
(ADT). Reaction products were deployed on the BeadArray
TM
platform [18] [19].
Identification of Genetic Variant and SNP picking
Because Hispanics were not represented in the HapMap resource [20] at the
time of this study, SNPs were selected using a procedure we call SNPwalk. The goal
of SNPwalk is to select polymorphism SNPs at regular intervals across a defined
region for genotyping. Prior to SNP selection using SNPwalk, the following criteria
were established:
8

1. Select SNPs that yield an average density of 2 to 2.5Kb per SNP.
2. Select SNPs within the known coding region of the gene (as defined by NCBI
genome build 36 assembly) and also 40Kb upstream and 10Kb downstream from
the coding region to include regions (ex: promoter, enhancer) regulate the gene.
3. Select bi-allelic SNPs only.
4. Include all non-synonymous SNPs within the defined region.
5. Preferentially select SNPs geneotyped by HapMap and are polymorphic in all
four HapMap populations (Yoruba in Ibadan, Nigeria (YRI), Japanese in Tokyo,
Japan (JPT), Han Chinese in Beijing, China (CHB), Utah residents with ancestry
from northern and western Europe (CEU)) at the time of this study.
6. Because of our limited sample size, exclude SNPs having a minor allele
frequency less than 0.05 (see below).
7. In region with no genotype HapMap SNPs available, use other public SNP
resources, like SeattleSNPs (http://pga.mbt.washington.edu) or SNP500Cancer
Database (http://snp500cancer.nci.nih.gov) to select SNPs.
The following procedure was applied using the criteria set forth above:
9
1. Information on SNPs within in our target gene region are downloaded from the
HapMap website (http://www.hapmap.org).
2. Monomorphic SNPs and SNPs having minor allele frequency (MAF) less than
0.05 in all four HapMap populations are excluded.
3. Select SNPs starting 40Kb upstream of gene region; a new SNP will be picked if
its distance to the previous selected SNP is 2 to 2.5 Kb.
4. If there are no SNPs located within 2 to 2.5Kb our current SNP, then the next
furthest SNP is selected.
5. If there is more than one SNP located 2 to 2.5 Kb from the previous SNP, select
the SNP that has the higher MAF or is non-synonymous polymorphism.
6. After we the first round of SNP picking, we calculate all gap sizes among two
adjacent selected SNPs (defined as the distance between two adjacent selected
SNPs). For all gaps > 3.0 Kb, additional HapMap SNPs are selected in an attempt
to fill the gap to the set in order to shrink the gap size. The goal is to keep every
gap size under 3.0 Kb.
7. Once all gaps are filled, all remaining unpicked non-synonymous SNPs are
selected.
10
8. For all remaining gaps >10 Kb, additional resources other than HapMap are
utilized to identify gap-filling SNPs. The following public online resources were
used: SeattleSNPs(http://pga.mbt.washington.edu/), SNP 500 Cancer Database
(http://snp500cancer.nci.nih.gov/) and NCBI dbSNP database
(http://www.ncbi.nlm.nih.gov/SNP/index.html).
A summary of the SNPwalk selection procedure for CYP3A4, CYP2C8, and CYP2C9
is shown in Table 1.
Table 1. Summary of SNP sets selected by SNPwalk
Gene Gene
Region
(Kb)
SNPs
Picked
Average
Gap Size
(Kb)
Non-HapMap
SNPs
Density
(Kb/SNP
)
Maximum
Gap Size
(Kb)
CYP3A4 77.21 26 2.96 8 2.97 7.71
CYP2C8 82.73 33 2.37 2 2.51 5.69
CYP2C9 100.71 43 2.38 0 2.34 6.94
Quality control of genotyping data
Using the SNPwalk procedure, we select 26, 33, and 43 SNPs, respectively, in
CYP3A4, CYP2C8 and CYP2C9 for genotyping. Due to the technical restrictions of
the Illumina platform, SNPs having low prior probability of successful genotyping
11
were excluded from genotyping. As a result, 14, 12 and 23 SNPs were dropped from
CYP3A4, CYP2C8 and CYP2C9, respectively.
We check the quality of genotyping by performing Hardy-Weinburg
Equilibrium (HWE) tests on each SNP. Two SNPs in CYP2C8 and one SNP in
CYP2C9 failed HWE tests and were excluded from the analysis. Additionally,
monomorphic SNPs and SNPs have MAF less than 0.05 were also excluded: 3 SNPs
in CYP3A4, 3 SNPs in CYP2C8 and 5 SNPs in CYP2C9. The final average SNP
densities were 8.58, 5.17 and 7.19 SNP/Kb for CYP3A4, CYP2C8, and CYP2C9,
respectively.
TagSNP selection and genotype imputation
We used Tagger [21] as implemented in Haploview [22] to select tag SNPs for
each gene. We use the aggressive tagging method, including pairwise r
2
selection and
multi-marker haplotype approaches. Tagger selected 4 tag SNPs for CYP3A4, 7 for
CYP2C8 and 6 for CYP2C9.
Because of the limited sample size in our study, a small number of individual
genotypes were imputed to minimize the number of missing genotypes at the subject
level. Missing genotypes were imputed by taking advantage of r
2
and D’ values
12
among other SNPs where genotype data were available. Individual genotypes were
only imputed from other SNPs which have a D’=1 and r
2
greater than 0.9 with our
imputed SNP. In CYP3A4, we imputed one individual genotype in rs2177179. In
CYP2C8, we imputed two individual genotypes in rs1934980 and rs1057910, and in
CYP2C9 we imputed two individual genotypes in rs12778431. Details regarding the
imputation are found in Appendix 1.
Statistical Analysis
IVGTTs were analyzed using the Bergman Minimal Model to estimate AIR, S
I

and DI. Response to PIO was defined as the change in S
I
higher than the known
coefficient of variation (0.3 units) in S
I
. Pill compliance was defined as the total
percent of pills taken over the course of PIPOD.
Because of the relative small sample size of the study, median and quartile
range were calculated instead of mean and variance for sample characteristics. Under
the null hypothesis of no difference between responders and non-reponders, the
non-parametric Kruskal-Wallis test was used to compare the median of phenotypes
between baseline and after one year of PIO treatment, and also the change of
phenotypes medians from baseline to after one year of PIO treatment between
13
responders and non-responders.
We used the Principal Component (PC) Analysis approach of Gauderman et al.
to initially test for association between response to PIO or 1-year change in
phenotype jointly with all SNPs [23]. We use 80% gene variation explained as the
cut point to decide the number of PCs in the model. We ran a basic model testing the
association between PCs and PIO response only and a second model adjusting for
age, BMI, and the change in BMI from baseline to after one year as covariates. For
PCs showing evidence for a “trend” for association (p<0.25), we then examined the
association between the individual SNPs and PIO response adjusting for age,
baseline BMI and the change in BMI between baseline and after one year under a
multiple logistic regression framework. The association between key phenotypes and
PCs was also examined by using the methods described above, except that linear
regression methods were used as our test of association. All regression modeling and
test of association were conducted using SAS V .9.1.
14
Results
Basic Characteristics
Table 2 shows subject demographics and clinical characteristics at baseline and
after one year of PIO treatment in 37 responders and 19 non-responders with DNA
and complete phenotypes. There are no statistically significant differences in age,
weight, BMI, AIR, disposition index, SBP, DBP, triglycerides, cholesterol and pill
compliance between responders and non-responders at both baseline and after one
year of PIO treatment. However, baseline fasting glucose and S
I
are higher in
responders (p=0.04) compared to non-responders. Fasting insulin is marginally
higher (p=0.05) and S
I
is significantly lower in non-responders (p=0.01) after one
year of PIO treatment.
Table 3 shows 1-year change in phenotypes in responders and non-responders
after treatment with PIO. There are significant differences in fasting insulin (p<0.05),
as well as fasting glucose, S
I
and disposition index (p<0.001), while there are no
differences in BMI and AIR.

15
Table 2 Phenotype values at baseline and after one year of PIO treatment in
responders and non-responders. Values are median and interquartile range.
Phenotype
Responders (N=37) Non-Responders (N=19)
Baseline 1-Year Baseline 1-Year
Median Qrange Median QRange Median QRange Median QRange
Age (years) 41.8 6.3 --- --- 38.3 11.8 --- ---
Weight (kg) 72.3 13.6 75.5 15.5 70.5 14.5 72.7 15.5
BMI
(Kg/m
2
)
30.1 4.7 31.2 6.0 30.2 6.6 31.3 6.6
Fasting
Glucose
(mg/dl)
95 20 92 10 88* 27 91 20
Fasting
Insulin
(mg/dl)
13 10 8 4 11 28 12* 23
S
I

(x10
3
min
-1
per
pM)
1.5 1.2 3.9 2.5 3.1* 3.6 2.2* 2.5
AIR
(pMx10min)
306.0 575.0 286.0 411.0 417.0 489.0 348.0 492.0
Disposition
Index
480.4 905.7 1045.9 1003.3 1207.9 1201.5 794.6 951.8
Systolic
Blood
Pressure (mm
Hg)
109.0 12 165.0 30.5 113.5 16.5 164.5 29.5
Diastolic
Blood
Pressure (mm
Hg)
68.5 10.5 99.0 13.0 69.5 17.5 104.5 18.6
16

Triglycerides
(mg/dL)
96.0 74.0 104.0 95.0 147.0 94.0 130.0 119.0
Cholesterol
(mg/dL)
200.0 49.0 184.0 59.0 185.0 67.0 194.0 52.0
Pill
Compliance
(%)
--- --- 91.8 20.5 --- --- 100.0 23.3
* Statistically significantly between responders and non-responders (p<0.05)
Table 3 One-year change in phenotypes in responders and non-reponders to PIO.
Values are median and interquartile range.
Phenotypes
Responders (N=37) Non-Responders (N=19)
Median QRange Median Qrange
BMI (Kg/m
2
) 1.45 2.16 0.79 1.39
Fasting Glucose
(mM)
-6 9 3** 11
Fasting
Insulin (mM)
-4 7 -2* 8
S
I

(x10
3
min
-1
per pM)
1.62 2.07 -0.46** 1.42
AIR
(pMx10min)
-17.0 248 43.0 271.7
Disposition
Index
373.8 751.2 -102.6** 673.0
* Statistically significantly between responders and non-responders (p<0.05)
** Statistically significantly between responders and non-responders (p<0.001)
PIO response -- Principal Component Analysis
For all three genes we found that only three PCs were required to explain more
than 80% of gene variation (Table 4-6). The PC represents a weighted average of
Table 2: Continued
17
SNPs in the region. For example, PC1 in CYP3A4 consisted of:
PC1 = rs6956344*0.56 + rs4646440*(-0.22) + rs4646437*0.64 +
rs2177179*0.48

We then tested the association between the PCs and PIO response in simple and
adjusted models (Table 7-9). We found evidence of association between PC2 in
CYP3A4 and PIO response (p=0.03; OR=2.15, 95% CI=1.09-4.15; adjusted model),
but no evidence for association with PC1 (p= 0.29) or PC3 (p= 0.32). PC2 in
CYP3A4 weights heavily on rs4646440 (Table 4), which we found it associated with
PIO response and phenotype and will be discussed later. PC2 and PC3 in CYP2C8
showed trends for association with PIO response (p=0.11 and 0.12, respectively;
adjusted model), but not with PC1 (p=0.42; adjusted model). PC2 in CYP2C9
showed trend for association (p-0.15; adjusted model), but not PC1 or PC3 (p=0.47
and 0.40, respectively; adjusted model).
PC Association with Multiple Phenotypes
We next tested whether the PCs were associated with PIO-induced changes in
T2DM-related quantitative traits; BMI, S
I
, AIR, DI, and TG (Table 10). In CYP3A4,
18
PC2 showed a trend for association with AIR (p=0.09), as well as PC3 with DI
(p=0.17), and PC1 with TG (p=0.10). In CYP2C8, we found significant association
between PC2 with S
I
and AIR (p=0.02, 0.04). No other PCs show evidences
ofassociation or trend of association. In CYP2C9, PC3 showed significant
association with AIR (p=0.02), and PC1 showed trend for association with AIR also
(p=0.18).
19
Table 4. Principal Component Analysis of tagSNPs in the CYP3A4 gene region.
(N=49)
SNP PC1 PC2 PC3 PC4
rs6956344 0.56 -0.23 0.56 0.55
rs4646440 -0.22 0.85 0.46 0.12
rs4646437 0.64 0.19 0.16 -0.73
rs2177179 0.48 0.43 -0.66 0.38
Eigenvalue 2.07 1.03 0.70 0.19
Variance
Explained (%)
52 26 18 5
Cumulative
Variance (%)
52 78 95 100

Table 5. Principal Component Analysis of tagSNPs in the CYP2C8 gene region.
(N=50)
SNP PC1 PC2 PC3 PC4
rs1934955 0.38 0.40 -0.35 -0.23
rs10509681 0.34 0.12 0.79 -0.04
rs2275620 0.48 -0.20 0.11 0.07
rs1934980 0.13 -0.78 -0.07 0.33
rs7916699 -0.36 0.35 0.11 0.75
rs4553304 0.43 0.19 0.13 0.37
rs11188183 0.43 0.12 0.45 0.35
Eigenvalue 3.70 1.40 0.80 0.48
Variance
Explained (%)
53 20 11 7
Cumulative
Variance (%)
53 73 84 91

20
Table 6. Principal Component Analysis of tagSNPs in the CYP2A9 gene region.
(N=54)
SNP PC1 PC2 PC3 PC4
rs4086116 0.59 -0.11 -0.02 0.00
rs10509679 -0.07 0.29 0.86 0.00
rs1856908 0.40 0.58 0.17 0.00
rs1057910 0.49 -0.15 0.01 0.70
rs2185570 0.49 -0.15 0.01 -0.70
rs12778431 -0.0004 0.72 -0.48 0.00
Eigenvalue 2.79 1.37 1.15 0.56
Variance
Explained (%)
46 23 19 9
Cumulative
Variance (%)
46 69 88 97

Table 7. Association between PCs in CYP3A4 and PIO Response.

PC Simple
Association
p-value
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

PC1 0.29 0.29 0.75 (0.45-1.27)
PC2 0.03 0.03 2.15 (1.09-4.25)
PC3 0.41 0.32 1.52 (0.66-3.50)
* Model adjusted for age, baseline BMI and △BMI
21
Table 8. Association of PCs in CYP2C8 and PIO Response.

PC Simple
Association
p-value
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

PC1 0.27 0.42 1.16 (0.80-1.69)
PC2 0.13 0.11 0.66 (0.39-1.11)
PC3 0.13 0.12 0.56 (0.27-1.17)
* Model adjusted for age, baseline BMI and △BMI

Table 9. Association of PCs in CYP2C9 and PIO Response.

PC Simple
Association
p-value
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

PC1 0.56 0.47 0.85 (0.56-1.31)
PC2 0.16 0.15 1.44 (0.88-2.35)
PC3 0.48 0.40 1.25 (0.74-2.10)
* : Model adjusted for age, baseline BMI and △BMI
22
Table 10. Association Between PCs and 1-year change in Phenotypes.
PC △
BMI(Kg/m
2
)
1

△
S
I
(x10
3
min
-1
per
pM)
2

△
AIR
(pMx10min)
2

△
Disposition
Index
2

△
Triglyceride
(mg/dL)
2

CYP3A4
PC1 0.28 0.82 0.45 0.74 0.10
3

PC2 0.56 0.97 0.09 0.60 0.52
PC3 0.94 0.56 0.49 0.17 0.39
CYP2C8
PC1 0.18 0.85 0.50 0.40 0.60
PC2 0.90 0.02 0.04 0.54 0.34
PC3 0.58 0.57 0.95 0.64 0.41
CYP2C9
PC1 0.21 0.75 0.18 0.54 0.51
PC2 0.86 0.86 0.43 0.46 0.97
PC3 0.91 0.61 0.02 0.90 0.89
1. Model adjusted for age, baseline BMI, nominal p-values
2. Model adjusted for age, baseline BMI and △BMI, nominal p-values
23
PIO response – SNP association
In order to examine the PC-based association further, we performed
individual SNP association with PIO response assuming an additive genetic model
(Table 11-13). SNP rs4646440 in CYP3A4 showed evidence for association with PIO
response (Bonferroni p=0.009, OR=5.27 (CI=1.50-18.44)) and remained significant
under a dominant model (p=0.01, OR=5.18 (CI=1.41-19.23)). SNP rs4646440 is
located in intron ten of CYP3A4. No other SNPs in CYP3A4 showed evidence for
association with response to PIO.
SNPs rs1934980 and rs7916699 in CYP2C8 showed trends for association with
PIO response (Bonferroni p=0.12 and 0.14 respectively). We found no evidence for
an association between variation in CYP2C9 and PIO response.
SNP Association with Multiple Phenotypes
We further examine the association between SNPs with 1-year change in
phenotypes for SNPs that showed significant or trends for association with PIO
response (Table 14-16). None of the SNPs in CYP3A4 showed evidence for
association with 1-year change in AIR, DI, S
I
, TG or BMI. rs1856908 in CYP2C8
showed marginally association with △S
I
(p=0.01), but was not significant after
24
Bonferroni correction. rs10509679 in CYP2C9 showed marginally association with
△AIR (p=0.03), but also did not remain significant after Bonferroni correction.
25
Table 11. Association Between SNPs in CYP3A4 and PIO Response.
SNP Major/
Minor
Allele

MAF in
Responders/
Non-responders
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

rs6956344 G/A 0.14/0.05 0.11 0.24 (0.04-1.37)
rs4646440 G/A 0.18/0.38 0.009
**
5.27 (1.50-18.44)
rs4646437 G/A 0.20/0.16 0.56 0.67 (0.20-2.40)
rs2177179 A/G 0.09/0.05 0.34 2.34 (0.41-13.30)
*
Model adjusted for age, baseline BMI and △BMI, nominal p-values
**
Significant after bonferonni correction of p value

Table 12. Association Between SNPs in the CYP2C8 and PIO Response.
SNP Major/
Minor
Allele

MAF in
Responders/
Non-responders
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

rs1934955 A/G 0.13/0.14 0.98 1.02 (0.26-3.94)
rs10509681 A/G 0.06/0.03 0.33 3.29 (0.30-36.00)
rs2275620 A/T 0.31/0.39 0.71 1.19 (0.48-2.99)
rs1934980 A/G 0.09/0.18 0.14 0.36 (0.10-1.30)
rs7916699 A/G 0.46/0.29 0.14 0.52 (0.22-1.23)
rs4553304 C/G 0.30/0.32 0.86 0.93 (0.40-2.14)
rs11188183 A/G 0.13/0.19 0.40 0.58 (0.17-2.04)
*
Model adjusted for age, baseline BMI and △BMI, nominal p-values
26
Table 13. Association Between SNPs in the CYP2C9 and PIO Response.
SNP Major/
Minor
Allele

MAF in
Responders/
Non-responders
Adjusted
Association
p-value
*

Odds Ratio
(95%CI)
*

rs4086116 G/A 0.14/0.09 0.47 0.63 (0.18-2.17)
rs10509679 G/A 0.05/0.11 0.19 2.84 (0.59-13.67)
rs1856908 A/C 0.31/0.42 0.37 1.45 (0.64-.328)
rs1057910 A/C 0.08/0.05 0.38 2.25 (0.36-13.89)
rs2185570 A/G 0.05/0.09 0.85 0.85 (0.15-4.75)
rs12778431 G/C 0.07/0.11 0.79 1.21(0.30-4.86)
*
Model adjusted for age, baseline BMI and △BMI, nominal p-values
Table 14. Association Between SNPs in CYP3A4 and 1-year Change in Phenotypes.
SNPs △
BMI(Kg/m
2
)
1

△
S
I
(x10
3
min
-1

per pM)
2

△ △ △ △
AIR(pMx10min)

2

△ △ △ △
Disposition
Index
2

△ △ △ △
Triglycerides
(mg/dL)
2s

rs6956344 0.86 0.92 0.65 0.49 0.13
rs4646440 0.61 0.75 0.57 0.17 0.66
rs4646437 0.31 0.69 0.13 0.62 0.10
rs2177179 0.25 0.66 0.16 0.37 0.68
1. Model adjusted for age, baseline BMI, nominal p-values
2. Model adjusted for age, baseline BMI and △BMI, nominal p-values
27
Table 15. Association Between SNPs in CYP2C8 and 1-year Change in Phenotypes.
SNPs △
BMI(Kg/m
2
)
1

△
S
I
(x10
3
min
-1

per pM)
2

△ △ △ △
AIR(pMx10min)

2

△ △ △ △
Disposition
Index
2

△ △ △ △
Triglycerides
(mg/dL)
2s

rs1934980 0.55 0.01 0.13 0.25 0.38
rs7916699 0.52 0.42 0.51 0.33 0.13
rs1934955 0.42 0.52 0.12 0.29 0.11
rs10509681 0.14 0.63 0.51 0.81 0.96
rs2275620 0.12 0.41 0.79 0.41 0.91
rs4553304 0.59 0.17 0.10 0.87 0.14
rs11188183 0.50 0.90 0.22 0.47 0.23
1. Model adjusted for age, baseline BMI, nominal p-values
2. Model adjusted for age, baseline BMI and △BMI, nominal p-values

Table 16 Association Between SNP in CYP2C9 and 1-year Change in Phenotypes.
SNPs △
BMI(Kg/m
2
)
1

△
S
I
(x10
3
min
-1

per pM)
2

△ △ △ △
AIR(pMx10min)

2

△ △ △ △
Disposition
Index
2

△ △ △ △
Triglycerides
(mg/dL)
2s

rs10509679 0.99 0.67 0.03 0.98 0.87
rs4086116 0.19 0.78 0.28 0.60 0.49
rs1856908 0.34 0.87 0.40 0.28 0.77
rs1057910 0.38 0.50 0.34 0.81 0.07
rs2185570 0.24 0.70 0.63 0.42 0.72
rs12778431 0.95 0.65 0.08 0.52 0.94
1. Model adjusted for age, baseline BMI, nominal p-values
2. Model adjusted for age, baseline BMI and △BMI, nominal p-values

28
Discussion
Our study examined the characteristic differences between responders and
non-responders, as well as the effect of genetic variants in specific cytochrome p450
genes on response to PIO therapy in Hispanic women at risk for T2DM. In terms of
subject phenotypes/characteristics, there were few differences between responders
and non-responders. Age, weight, BMI, fasting glucose, fasting insulin, AIR, DI, SBP,
DBP, TG and cholesterol were similar between responders and non-responders.
Baseline S
I
in non-responders was marginally higher than responders (3.1 vs. 1.5
x10
3
min
-1
per pM, p=0.04, data not shown). However, after one year of PIO
treatment, S
I
(3.9 vs. 2.2 x10
3
min
-1
per pM, p<0.05) and fasting insulin (12 vs. 8
mg/dl, p<0.05) in responders was different compared to non-responders. This finding
is similar to previous findings showing that the one-year fasting insulin was
significantly lower in responders and one year S
I
was significantly higher in
responders [1, 2, 4]. We found that the one-year change for fasting glucose, fasting
insulin, S
I
, and DI were all significantly different between responders and
non-responders (p<0.05), which were also similar to previous studies.
Our power to detect association between polymorphisms and differential PIO
response is limited due to sample size. In order to identify potential trends for future
study with larger sample sizes, we chose to increase our significant threshold. We
estimated an appropriate significance cut-off using Quanto [24-27] and assuming 1)
a 1:2 unmatched case-control design 2) An additive genetic model, and 3) minor
29
allele frequency of 0.25, and 4) an odds ratio of 2.0. The criterion of OR=2.0 was
derived from our previous study examining association between PPARG and
response to TRO [9]. Base on these assumptions, a significance threshold of p=0.25
gave us 70% power to detect odds ratio ≥2.0.
The focus of our study is to examine the association between polymorphisms in
three CYP genes with PIO response and various T2DM-related phenotypes. Principal
component analysis was performed to conduct an association test between the linear
combination of SNPs in one gene with phenotypes. This gave us the methodological
advantage to screen for evidence for association, but without paying the price of
multiple test correction due to multiple SNPs. However, since our sample size is
limited, the trend of association threshold had to be set to a relatively high (p<0.25),
in order to achieve adequate statistical power. When we examined the association
between PCs and PIO response, PC analysis showed some level of association
between all three genes and PIO response. The association test between PCs and
characteristics/phenotypes also showed there are different degrees of association
between different phenotypes and PCs. All results hint that the polymorphisms in
CYP genes are associated with PIO response.
PC analysis showed PC2 in CYP3A4 has the strongest association with PIO
response (p=0.03). Our SNP association tests with PIO responses confirmed that
SNP rs4646440 in CYP3A4 was strongly associated with differential response to PIO
treatment, showin significant association (nominal p=0.009, 0.01). We also found
30
that rs1934980 in CYP2C8 was associated with response to PIO. rs1934980 showed
trend for association with PIO response and nominally significant association with S
I

(p=0.01, S
I
decreased by -2.18 x10
3
min
-1
per pM per G allele) and AIR (p=0.13, AIR
increased 183.98 pMx10 min per G allele). Additionally, PC2 in CYP2C8 was also
associated with S
I
(p=0.02) and AIR (p=0.04) with additional analysis suggesting
rs1934980 as the SNP contributing to the evidence for association.
In both the TRIPOD and PIPOD study, TZDs were all able to improve S
I
and
reduce endogenous insulin secretion. However, despite being drugs of the same
family and having similar non-response rates, the concordance of response to TRO
and PIO was only 50% [12]. This suggests that the mechanism underlying response
to TZDs differs even among individual’s medications from this class. Interestingly, in
different TZDs related trials or studies, most of observed a TZD non-response rate
ranging from 30-40% [1, 3, 4, 9-11, 28, 29]. This raises the question of whether the
underlying cause of differential response is purely genetic, purely environmental
effect, or due to interaction between gene and environment. Results from our studies
and from others suggest a purely environmental explanation is highly unlikely. This
suggests differential response to PIO maybe due to purely genetic effects or possibly
interactions between genetic variants and specific environmental; possibilities that
still require examination. Because of our limited sample size, we did not have
sufficient power to formally test the gene-environment interaction hypothesis. Thus,
we cannot exclude the possibility that gene-environmental interactions may
contribute to differential response.
31
From previous studies and this study, we found SNPs in PP ARG, CYP3A4 and
CYP2C8 that showed association with PIO response. It is reasonable to assume that
the genetic effect component of characteristic differences between responders and
non-responders on response to PIO therapy may be a polygenic effect, and this
polygenic effect could be contributed by genetic polymorphism in other genes other
than the ones aforementioned. Future efforts will be focused on the genes related to
biochemical pathway corresponding to PPARG, including Retinoid X receptor
(RXRA), which forms a heterodimer with PP ARG; Nuclear Receptor Corepressor 1
(NCOR1), which is the co-repressor to RXRA-PPARG complex; Adiponectin, CqQ
and collagen domain containing (ADIPOQ), which is down-regulated by PP ARG,
and involved in glucose and lipid metabolism. In terms of summarizing the
polygenic effect from different genetic variants, method of PC analysis could be
introduced again to construct linear weighting combination of SNPs from different
genes.
In summary, our study found that the SNP rs4646440 in CYP3A4 are associated
with PIO response. Other two SNPs, rs1934980 and rs7916699 also show trend of
significant association with PIO response. PCs constructed from three genes all show
different levels of trend of association with AIR, though none of the SNPs show
association with any phenotypes after correction of Bonferroni. We speculate the
differential response to TZDs may be a polygenic effect because of the sums of
effects from multiple genetic variants located in different genes. Much is still
unknown about the mechanism to differential response to TZDs, and the genetic
32
variants which contribute to it. A study with large sample size and the genotyping
efforts on other genes may help us to understand the detailed mechanism more. The
association and the trends of association found in our study are worthy of further
investigation.
33
References
28. Antonucci, T., et al., Impaired glucose tolerance is normalized by treatment with
the thiazolidinedione troglitazone. Diabetes Care, 1997. 20(2): p. 188-93.
3. Aronoff, S., et al., Pioglitazone hydrochloride monotherapy improves glycemic
control in the treatment of patients with type 2 diabetes: a 6-month randomized
placebo-controlled dose-response study. The Pioglitazone 001 Study Group. Diabetes
Care, 2000. 23(11): p. 1605-11.
22. Barrett, J.C., et al., Haploview: analysis and visualization of LD and haplotype
maps. Bioinformatics, 2005. 21(2): p. 263-5.
12. Black, M.H., Genetic Variation in Peroxisome Proliferator-activated
receptor-gamma (PPARG) may account for differential response to Pioglitazone, in
Department of Preventive Medicine. 2005, University of Southern California: Los
Angeles. p. 26.
29. Bluher, M., G. Lubben, and R. Paschke, Analysis of the relationship between the
Pro12Ala variant in the PP AR-gamma2 gene and the response rate to therapy with
pioglitazone in patients with type 2 diabetes. Diabetes Care, 2003. 26(3): p. 825-31.
16. Bozkurt, O., et al., Pharmacogenetics of glucose-lowering drug treatment: a
systematic review. Mol Diagn Ther, 2007. 11(5): p. 291-302.
1. Buchanan, T.A., et al., Preservation of pancreatic beta-cell function and
prevention of type 2 diabetes by pharmacological treatment of insulin resistance in
high-risk hispanic women. Diabetes, 2002. 51(9): p. 2796-803.
4. Buchanan, T.A., et al., Response of pancreatic beta-cells to improved insulin
sensitivity in women at high risk for type 2 diabetes. Diabetes, 2000. 49(5): p. 782-8.
19. Butler, H. and J. Ragoussis, BeadArray-based genotyping. Methods Mol Biol,
2008. 439: p. 53-74.
34
21. de Bakker, P.I., et al., Efficiency and power in genetic association studies. Nat
Genet, 2005. 37(11): p. 1217-23.
6. Durbin, R.J., Thiazolidinedione therapy in the prevention/delay of type 2 diabetes
in patients with impaired glucose tolerance and insulin resistance. Diabetes Obes
Metab, 2004. 6(4): p. 280-5.
23. Gauderman, W.J., et al., Testing association between disease and multiple SNPs
in a candidate gene. Genet Epidemiol, 2007. 31(5): p. 383-95.
27. Gauderman, W. and J. Morrison, QUANTO 1.1: A computer program for power
and sample size calculations for genetic-epidemiology studies,
http://hydra.usc.edu/gxe. 2006.
26. Gauderman, W., Candidate gene association analysis for a quantitative trait,
using parent-offspring trios. Genet Epidemiol, 2003. 25: p. 327-38.
24. Gauderman, W., Sample size requirements for association studies of gene-gene
interaction. Am J Epidemiol, 2002. 155: p. 478-484.
25. Gauderman, W.J., Sample size calculations for matched case-control studies of
gene-environment interaction. Statistics in Medicine, 2002. 21(1): p. 35-50.
17. Hanson, R.L., et al., Identification of PVT1 as a candidate gene for end-stage
renal disease in type 2 diabetes using a pooling-based genome-wide single
nucleotide polymorphism association study. Diabetes, 2007. 56(4): p. 975-83.
13. Jaakkola, T., et al., Pioglitazone is metabolised by CYP2C8 and CYP3A4 in vitro:
potential for interactions with CYP2C8 inhibitors. Basic Clin Pharmacol Toxicol,
2006. 99(1): p. 44-51.
15. Kirchheiner, J., et al., Impact of genetic polymorphisms in CYP2C8 and
rosiglitazone intake on the urinary excretion of dihydroxyeicosatrienoic acids.
Pharmacogenomics, 2008. 9(3): p. 277-288.

35
8. Rajagopalan, R., S. Iyer, and M. Khan, Effect of pioglitazone on metabolic
syndrome risk factors: results of double-blind, multicenter, randomized clinical trials.
Curr Med Res Opin, 2005. 21(1): p. 163-72.
14. Scheen, A.J., Pharmacokinetic interactions with thiazolidinediones. Clin
Pharmacokinet, 2007. 46(1): p. 1-12.
11. Snitker, S., et al., Changes in insulin sensitivity in response to troglitazone do not
differ between subjects with and without the common, functional Pro12Ala
peroxisome proliferator-activated receptor-gamma2 gene variant: results from the
Troglitazone in Prevention of Diabetes (TRIPOD) study. Diabetes Care, 2004. 27(6):
p. 1365-8.
10. Spiegelman, B.M., PP AR-gamma: adipogenic regulator and thiazolidinedione
receptor. Diabetes, 1998. 47(4): p. 507-14.
18. Steemers, F.J. and K.L. Gunderson, Illumina, Inc. Pharmacogenomics, 2005. 6(7):
p. 777-82.
20. The International Hapmap Consortium, The International HapMap Project.
Nature, 2003. 426(6968): p. 789-96.
7. Wallace, T.M., J.C. Levy, and D.R. Matthews, An increase in insulin sensitivity
and basal beta-cell function in diabetic subjects treated with pioglitazone in a
placebo-controlled randomized study. Diabet Med, 2004. 21(6): p. 568-76.
9. Wolford, J.K., et al., Sequence variation in PP ARG may underlie differential
response to troglitazone. Diabetes, 2005. 54(11): p. 3319-25.
2. Xiang, A.H., et al., Effect of pioglitazone on pancreatic beta-cell function and
diabetes risk in Hispanic women with prior gestational diabetes. Diabetes, 2006.
55(2): p. 517-22.
5. Xiang, A.H., et al., Pharmacological treatment of insulin resistance at two
different stages in the evolution of type 2 diabetes: impact on glucose tolerance and
beta-cell function. J Clin Endocrinol Metab, 2004. 89(6): p. 2846-51.
36
Appendix
Table A1. List of Imputed Genotypes
SNP Subject
ID
Genotype
Imputed
SNP
Imputed
From
Genotype
in source
R-sq
uare
D’
(Confidence
Bound)
rs2177179 133 AA rs2687091 GG 1.0 1.0 (1.0-0.94)
rs1934980 82 AA rs1058932 GG 0.91 1.0 (1.0-0.94)
rs1057910 258 AA rs9332214 AA 1.0 1.0 (1.0-0.92)
rs12778431 145 GG rs12251688 CC 0.96 1.0 (1.0-0.92)
rs12778431 244 CG rs12255222 AT 1.0 1.0 (1.0-0.93)

Abstract (if available)

Abstract Studies have shown that in addition to treating T2DM, TZDs may be effective in the prevention of T2DM as first demonstrated in the TRIPOD and PIPOD studies. However, both studies showed that 30%-40% subjects do not respond to TZDs. To determine whether genetic variants in CYP gene family contribute to non-response, 102 SNPs in CYP2C8, CYP2C9, and CYP3A4 were genotyped and examined for the association with PIO response and clinical phenotypes. Our results show SNP rs4646440 in CYP3A4 is associated with PIO response. Two additional SNPs, rs1934980 and rs7916699 also showed trend for association with PIO response. rs1934980 in CYP2C8 and rs10509679 in CYP2C9 showed trend of association with 1-year change in phenotypes. Our study suggests that polymorphisms in CYP gene family are associated with differential PIO response. These results support a polygenic model regarding differential response to TZD therapy in Hispanic women.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Variation in insulin-like growth factor-2 binding protein 2 interacts with adiposity to alter insulin sensitivity in Mexican Americans

PDF

Association of single nucleotide polymorphisms in GCK, GCKR and PNPLA3 with type 2 diabetes related quantitative traits in Mexican-American population

PDF

Novel roles for Maf1 in embryonic stem cell differentiation and adipogenesis

PDF

Detecting joint interactions between sets of variables in the context of studies with a dichotomous phenotype, with applications to asthma susceptibility involving epigenetics and epistasis

PDF

From mesenchymal stem cell therapy to discovery of drug therapy for systemic sclerosis

PDF

The role of endoplasmic reticulum chaperones in adipogenesis, liver cancer and mammary gland development

Asset Metadata

Creator Chen, Yen-Fu (author)

Core Title Genetic variations in gene from the cytochrome P450 family may account for differential response to pioglitazone therapy in the Hispanic women

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 11/25/2009

Defense Date 10/20/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag CYP2C8,CYP2C9,CYP3A4,cytochrome P450,Diabetes,OAI-PMH Harvest,pioglitazone,PIPOD,principal component analysis,SNP,T2DM,TZD

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Watanabe, Richard M. (committee chair), Allayee, Hooman (committee member), Gauderman, W. James (committee member)

Creator Email frank.usc@gmail.com,yenfuche@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2765

Unique identifier UC1281138

Identifier etd-Chen-3382 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-278329 (legacy record id),usctheses-m2765 (legacy record id)

Legacy Identifier etd-Chen-3382.pdf

Dmrecord 278329

Document Type Thesis

Rights Chen, Yen-Fu

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu