IA

HSSXT*�

F

Bfic(wtMfomtvCbr[battcScbcsbrDitc

IiS

GASTROENTEROLOGY 2006;130:1679–1687

dentification of Two Gene Variants Associated With Risk ofdvanced Fibrosis in Patients With Chronic Hepatitis C

ONGJIN HUANG,* MITCHELL L. SHIFFMAN,‡ RAMSEY C. CHEUNG,§ THOMAS J. LAYDEN,�

COTT FRIEDMAN,¶ OLIVIA T. ABAR,* LINDA YEE,# ANAND P. CHOKKALINGAM,*TEVEN J. SCHRODI,* JASON CHAN,* JOSEPH J. CATANESE,* DIANE U. LEONG,* DAVID ROSS,*IAOLAN HU,* ALEXANDER MONTO,# LINDA B. MCALLISTER,* SAMUEL BRODER,**HOMAS WHITE,* JOHN J. SNINSKY,* and TERESA L. WRIGHT#

Celera Diagnostics, Alameda, California; ‡Virginia Commonwealth University, Richmond, Virginia; §Stanford University, Stanford, California;University of Illinois, Chicago, Illinois; ¶Mount Sinai School of Medicine, New York, New York; #University of California San Francisco, San

rancisco, California; and **Celera Genomics, Rockville, Maryland

ttpPaftiitota

HcBtvpap

smafiggdw

n

ackground & Aims: Previously identified clinical riskactors such as sex, alcohol consumption, and age atnfection do not accurately predict which patients withhronic hepatitis C (CHC) will develop advanced fibrosisbridging fibrosis and cirrhosis). The aim of this studyas to identify genetic polymorphisms that can predict

he risk of advanced fibrosis in patients with CHC.ethods: A total of 916 subjects with CHC was enrolled

rom 2 centers. A gene-centric disease association studyf 24,832 putative functional, single nucleotide poly-orphisms (SNPs) was performed. Of the 1609 SNPs

hat were significantly associated (P < .05) with ad-anced fibrosis in the discovery cohort (University ofalifornia San Francisco [UCSF], N � 433), the firstatch of 100 SNPs were selected for validation in theeplication cohort (Virginia Commonwealth UniversityVCU], N � 483). Results: A missense SNP in the DEADox polypeptide 5 (DDX5) gene was significantly associ-ted with an increased risk of advanced fibrosis in bothhe UCSF and the VCU cohorts (OR, 1.8 and 2.2, respec-ively). Two diplotype groups, carrying the haplotypesomposed of the DDX5 SNP and 2 neighboring POLG2NPs were also significantly associated with an in-reased risk of advanced fibrosis and had comparable oretter risk estimates. In addition, a missense SNP in thearnitine palmitoyltransferase 1A (CPT1A) gene was as-ociated with a decreased risk of advanced fibrosis inoth the UCSF and the VCU cohorts (OR, 0.3 and 0.6,espectively). Conclusions: Subjects with CHC carryingDX5 minor allele or DDX5-POLG2 haplotypes are at an

ncreased risk of developing advanced fibrosis, whereashose carrying the CPT1A minor allele are at a de-reased risk.

nfection with hepatitis C virus (HCV) is a major causeof cirrhosis and hepatocellular carcinoma and the lead-

ng indication for liver transplantation in the United

tates and many Western countries. The progression rate

o cirrhosis varies widely among HCV patients, andreatment with antiviral therapy is most cost-effective inatients with evidence of progressive liver disease.1,2

reviously identified risk factors such as sex, alcohol, aget infection, obesity, and hepatic steatosis have beenound to correlate with fibrosis progression.3–5 However,hese factors have not been useful in accurately predict-ng which patients will progress to cirrhosis and indentifying which patients require treatment. The iden-ification of genetic factors, alone or in combination withther clinical risk factors, could more accurately identifyhose patients at increased risk for fibrosis progressionnd thereby improve treatment strategies.

Previous studies of host genetic factors involved inCV fibrosis have been limited to the analysis of a few

andidate genes in relatively small sample sets (seeataller et al6 for review and papers cited herein). Fur-

hermore, very few published genetic markers have beenalidated in independent sample sets.6 This is a commonroblem observed in other disease association studies7–9

nd precludes these markers from being usefully incor-orated into clinical practice.To address these issues, we carried out a genome-wide

can with putative functional, single nucleotide poly-orphisms (SNPs)10 on a group of 433 HCV patients

nd validated those markers associated with advancedbrosis (bridging fibrosis and cirrhosis) in a separateroup of 483 patients. Two novel markers, located in theenes DDX5 (17q21) and CPT1A (11q13.1), and 2iplotypes including the DDX5 marker were associatedith advanced fibrosis in both cohorts. These genetic

Abbreviations used in this paper: HCV, hepatitis C virus; SNP, singleucleotide polymorphism.© 2006 by the American Gastroenterological Association Institute

0016-5085/06/$32.00

doi:10.1053/j.gastro.2006.02.032

fati

Ha(MpttpRiahcvaa

tbibgbsdBaprieusfdccbtLo

S

npluigtaeicmstSwfbg

fw2pifcVmwo

t

1680 HUANG ET AL GASTROENTEROLOGY Vol. 130, No. 6

actors had moderate risk allele frequencies (9%–38%)nd significant odds ratios (1.7–3.4), suggesting thathey could play a useful role in predicting risk of fibrosisn HCV patients.

Materials and Methods

Study Population

A total of 916 patients with well-documented chronicCV were prospectively recruited from the hepatology clinics

t 2 large centers, the University of California San FranciscoUCSF; N � 433) and the Virginia Commonwealth University

edical Center (VCU; N � 483), over a 22 consecutive montheriod. Patients were either new referrals for evaluation and/orreatment of chronic HCV or existing patients being moni-ored on an ongoing basis at either of these centers. Allatients were adults (ages, 18–75 years), anti-HCV and HCVNA positive, and had a baseline liver biopsy prior to receiv-

ng any treatment for HCV. Patients were excluded if they hadny other chronic liver disease (Wilson’s disease, hereditaryemochromatosis, autoimmune hepatitis, steatohepatitis),oinfection with human immunodeficiency virus or hepatitis Birus, or presence of hepatocellular carcinoma. The study waspproved by the institutional review boards of both centers,nd written informed consent was obtained from all subjects.

Data on Clinical Risk Factors

All participants completed a questionnaire to recordheir risk factors for HCV acquisition (injection drug use,lood transfusion, and others), and alcohol consumption. Med-cal records were reviewed regarding demographics (date ofirth, sex, and ethnicity), histologic data on liver biopsy (dates,rade, and stage), and virologic features (HCV genotype). Liveriopsies were performed by hepatologists at each site andcored by local experienced pathologists using the systemescribed by Knodell et al (VCU) or its modification byatts-Ludwig (UCSF).11 At VCU, all biopsy specimens werelso reviewed by one of the authors (M.L.S.). Only thoseatients with biopsy specimens that were considered by theeading pathologists or hepatologists as adequately sized werencluded in the study. Age at infection was estimated as thearlier time point between the first exposure to injection drugse or blood transfusion prior to 1990, as defined in othertudies.3,4,12 Estimated duration of infection was calculatedrom the age of infection to the date of liver biopsy. When theuration of infection could be estimated, a fibrosis rate wasalculated as fibrosis stage divided by duration of infection. Toalculate progression rate on VCU patients, stage as measuredy Knodell score was converted into METAVIR units (similaro Batts-Ludwig score) as described in previous studies.3,13

ifetime and daily alcohol consumption were calculated basedn previously validated questionnaires.14

Functional Genome Scan

The genome scan consisted of 24,823 gene-centric

NPs composed of 68.3% coding functional SNPs (missense, a

onsense, acceptor, and donor splice sites); 24.9% noncodingutative regulatory SNPs that may effect RNA stability (SNPsocated at putative transcription factor binding sites, or 5=/3=ntranslated regions); and 6.8% other types of SNPs, includ-ng 1.8% intergenic (including human mouse conserved re-ions), 1.7% silent, and 3.3% intronic SNPs. These latter SNPypes were the results of continuous chromosome assembly andnnotation updates changing their original SNP types. Eighty-ight percent of the SNPs were identified in public and 12%n proprietary databases. A total of 12,248 genes were directlyovered by the scan. Direct coverage of genes on each chro-osome ranged from 23% (X chromosome) to 52% (chromo-

ome 10), with an average coverage of 39%. Most genes wereested with only 1 SNP; however, 1418 genes had 4 or moreNPs. The distribution of the marker minor allele frequency inhite subjects was as follows: 27% markers with a minor allele

requency less than 5%; 14% with a minor allele frequencyetween 5% and 10%; and 59% with a minor allele frequencyreater than 10%.

Genotyping

As illustrated in Figure 1, the genome scan was per-ormed on pooled DNA made from the UCSF samples. DNAere pooled according to minimal (stage 0 or 1), mild (stage), and advanced fibrosis stages (stage 3 or 4). These initialools were further divided by sex (male vs female) and ethnic-ty (white subjects vs others). The first 100 consecutive SNPsound to be associated with advanced fibrosis in the UCSFohort were then validated by individual genotyping in theCU cohort to confirm the observed associations. “Replicatedarkers” were defined as those SNPs significantly associatedith advanced fibrosis in both cohorts, with odds ratios (ORs)f similar magnitude and direction.

From each subject, 24 mL blood was collected in 3 EDTAubes; 3.0 ng of individual or pooled DNA was used in each

Figure 1. Strategy and work flow of association study.

mplification reaction. Pools of DNA were generated by mix-

isw1A7Fapaea

arafggwf(wa2btds

llHjgsel(ewdOwhttieora

tssawwhfivgctwvtancV

T

AE

S

D

F

D

F

Uwa

Kb

s

May 2006 FIBROSIS GENETIC MARKERS 1681

ng equal volumes of standardized DNA from each studyubject of the pool. For each SNP, 2 amplification reactionsere performed, 1 for each allele. Each PCR reaction containedof the 2 allele-specific primers and a common reverse primer.ll reactions were amplified on Applied Biosystems PRISM900HT Sequence Detection instrument (Applied Biosystems,oster City, CA). For individual DNA, the genotypes wereutomatically called by a custom-designed algorithm. Forooled DNA, duplicates were run for each DNA pool, andllele frequencies were calculated based on the average differ-nce in cutoff threshold (Ct) between the major and minorlleles.15

Statistical Analysis

For all analyses, cases were defined as subjects withdvanced fibrosis (bridging fibrosis or cirrhosis), and theemaining subjects with none to mild fibrosis were defineds controls. Fisher exact test16 was applied to data generatedrom pooled DNA (allelic association). Unconditional lo-istic regression17 was used to analyze data from individualenotyping (genotypic association). Multivariate analysesere adjusted for other risk factors, including sex (male vs

emale), daily alcohol (�50 g vs �50 g), and ethnicitywhite subjects vs others). Age at infection and steatosisere not adjusted because they were not available in 30%

nd over 50% of subjects, respectively. P values were-sided in the UCSF cohort and 1-sided in the VCU cohortecause of the narrow hypothesis used in the VCU cohorthat the effect of markers (odds ratios) had to be in the sameirection as in the UCSF cohort for the marker to beignificant.

For each “replicated marker,” flanking regions with highinkage disequilibrium were identified from the standardinkage disequilibrium display of HaploView, derived from

apMap data.18,19 Additional SNPs in or immediately ad-acent to these high linkage disequilibrium regions wereenotyped in all subjects. A “sliding window” with windowize of 3 SNPs was applied to all SNPs in the region. Forach window, haplotypes were estimated using the Hap-o.Score program, and the association global P valuethrough permutation test), adjusted for sex, alcohol, andthnicity, was calculated.20 SNPs comprising the windowsith the best P value were selected, and haplotype andiplotype of these SNPs for each subject were inferred.21

nly those haplotypes/diplotypes at over 1% frequencyere presented here, in which the error rate associated withaplotype/diplotype inference was negligible.20 Uncondi-ional logistic regression was used to estimate the associa-ion of these diplotypes with advanced fibrosis after adjust-ng for sex, alcohol, and ethnicity. For diplotype riskstimation, diplotypes were grouped according to presencef 1 or more haplotypes associated with increased fibrosisisk; the diplotype with the lowest estimated risk was used

s the reference group. w

Results

Study Subjects

Clinical characteristics of the 2 patient popula-ions are listed in Table 1. The 2 cohorts shared manyimilarities, suggesting that there was a good cross-ectional representation of HCV patients. The mean agend the proportion of white subjects within the 2 groupsere similar. A higher proportion of patients from UCSFere male, and these patients had consumed more alco-ol than the patients at VCU. Despite this, the meanbrosis score and the percentage of patients with ad-anced fibrosis stages (bridging fibrosis or cirrhosis) werereater in the VCU cohort. Table 2 compares the clinicalharacteristics of patients with advanced fibrosis withhose with none to mild fibrosis. At each site, male sexas significantly (P � .023–.004) associated with ad-anced fibrosis stage. Daily alcohol consumption of morehan 50 g was more common in the UCSF patients withdvanced fibrosis, but this association fell short of sig-ificance (P � .066). No relationship between alcoholonsumption and advanced fibrosis was present in theCU cohort. For both sites, age at infection over 40 years

able 1. Characteristics of Patients

UCSF_discoveryN � 433

VCU_replicationN � 483

ge (mean � SD) 52.7 � 7.2 50.8 � 8.3thnicity, %White 69.0 70.6African American 14.5 26.3Other ethnicity 16.6 3.1

ex% Male 83.7 57.8

aily alcohol� 50 g/day 46.4% 16.2%All: mean � SD 87 � 113 24 � 32Men: mean � SD 92 � 113 30 � 33Women: mean � SD 60 � 105 16 � 28

ibrosis scorea

mean � SD 1.7 � 1.3 2.0 � 1.5No fibrosis 21.6% 17.4%Portal/periportal fibrosis 50.6% 34.2%Bridging fibrosis 16.8% 23.6%Cirrhosis 11.0% 24.8%

uration of infectionMean duration 27.3 � 7.3 21.9 � 8.5Patients with duration 75.2% 71.8%

ibrosis rateb

mean � SD 0.08 � 0.17 0.09 � 0.45

CSF, University of California San Francisco; VCU, Virginia Common-ealth University.Fibrosis score was reported in Batts-Ludwig system for UCSF, innodell system for VCU.For rate calculation, the Knodell stage was converted to Metavirtage in VCU samples.3,13

as not a risk factor, probably because very few subjects

(Oibdw

tsswwmtmf

1v(Src(4a

ctsc(

a(taTb1gi(taa(tsDtsAmtfip

btso3dic

T

SD

A

AD

1682 HUANG ET AL GASTROENTEROLOGY Vol. 130, No. 6

�7%) were first infected when over the age of 40 years.lder age at the time of biopsy and longer duration of

nfection were significantly associated with advanced fi-rosis in the VCU but not in the UCSF cohort. Theuration of infection was similar for subjects with andithout advanced fibrosis within each cohort.

Functional Genome Scan

To increase speed and cost-efficiency, the func-ional genomic scan was performed on pooled UCSFamples. This strategy has been used successfully in othertudies.15,22 To verify the pooling accuracy, 17 markersith allele frequencies ranging from 0.3% to 48.3%ere genotyped in individual DNA as well as in poolsade of the same DNA. There was an excellent correla-

ion (r2 � 0.9862) between the allele frequencies deter-ined from individual genotyping and the calculated

requencies from pools (data not shown).Of the 24,823 SNPs genotyped in the UCSF patients,

609 SNPs showed evidence of association with ad-anced fibrosis in white subjects. Of these, 438 SNPs27.2%) had ORs greater than 2 or less than 0.5; 1133NPs (70.4%) were located in known genes, and theemaining 476 (29.6%) were located in uncharacterizedhromosomal regions. Regarding the SNP type, 105765.7%) were missense, nonsense, or splice site SNPs;60 (28.6%) were putative noncoding regulatory SNPs;nd 92 (5.7%) belonged to other categories.

Polymorphism in the DDX5 and CPT1AGene

The first 100 SNPs identified in the UCSF dis-overy cohort were then examined in the VCU replica-ion cohort. Two of these 100 SNPs were found to beignificantly associated with advanced fibrosis in bothohorts (Table 3). The first SNP was in the gene DEAD

able 2. Clinical Risk Factors Associated With Advanced Fibr

UCSF

Cases ControlsAdvanced fibrosis

(N � 121)None-mild fibros

(N � 312)

ex (% male) 90.1% 81.1%aily alcohol �50gMean � SD 91.7 � 102.5 85.8 � 116.5% �50 g 53.7% 43.9%

ge at infection �40YrsMean � SD 23.3 � 7.9 22.8 � 7.5% �40 3.3% 1.9%

ge at biopsy 50.8 � 6.6 49.8 � 7.6uration of infection 27.7 � 6.5 27.1 � 7.5

Asp-Glu-Ala-Asp) box polypeptide 5 (DDX5), causing r

n amino acid change from Ser to Ala in exon XIIIS480A). Subjects from UCSF carrying 1 or 2 copies ofhe minor allele had a significantly increased risk ofdvanced fibrosis (OR, 1.8; 95% CI: 1.0–3.3; P � .048).he association was confirmed in another equally largeut more diverse cohort from VCU (OR, 2.3; 95% CI:.3–4.3; P � .004). The second SNP was located in theene Carnitine palmitoyltransferase 1A (CPT1A), caus-ng an amino acid change from Ala to Thr in exon VIIIA275T). Subjects from UCSF carrying 1 or 2 copies ofhe minor allele had a significantly decreased risk ofdvanced fibrosis (OR, 0.3; 95% CI: 0.1–0.8; P � .015),nd the association was confirmed in the VCU cohortOR, 0.6; 95% CI: 0.3–1.1; P � .050). For both SNPs,he results did not differ significantly after adjusting forex, daily alcohol intake, and ethnicity, except for theDX5 SNP in the UCSF cohort (P � .083). However,

he combined P values for both SNPs remained highlyignificant, with or without these adjustments (Table 3).ge at time of infection was not included in the adjust-ent because this is the most unreliable of the prognos-

ic factors, was not significantly associated with advancedbrosis (Table 2), and was only available in 70% ofatients at either center (Table 1).

Additional SNPs in the DDX5 and CPT1ARegion

To characterize further the observed associationetween the DDX5 SNP and advanced fibrosis, we geno-yped 10 additional markers, located from 161 kb up-tream (SNP1) to 25 kb downstream (SNP11) of theriginal DDX5 polymorphism (Figure 2). Interestingly,SNPs (SNP 4, 5, 6) located in the Polymerase (DNA

irected), gamma 2 (POLG2) gene, adjacent to the orig-nal DDX5 marker (SNP7), were also significantly asso-iated with advanced fibrosis after adjusting for other

VCU

P value

Cases Controls

P valueAdvanced fibrosis

(N � 234)None-mild fibrosis

(N � 249)

.023 64.5% 51.4% .004

.623 23.9 � 29.0 24.0 � 34.2 .961

.066 18.0% 14.5% .297

.616 26.5 � 9.5 25.6 � 9.1 .391

.390 6.8% 4.4% .247

.213 49.3 � 7.5 46.2 � 8.7 �.0001

.508 22.9 � 8.5 21.0 � 8.4 .041

osis

is

isk factors (Figure 2: adjusted P � .05 in UCSF and

VCs

h2ft�t5tU7P(tdt

fdocaacb

ivt1a1cTaicoOVwF

fidtrHoUwa

T

Dh

Ch

Ua

b

c

d

e

f

g TT ge

May 2006 FIBROSIS GENETIC MARKERS 1683

CU cohorts). We also typed 13 SNPs in and around thePT1A polymorphism but did not find any additional

ignificant SNPs in this region (data not shown).

Haplotypes in DDX5 Region

A “sliding window” analysis indicated that theap window 4, composed of the 3 POLG2 SNPs (Figure), had the most significant global P values (calculatedrom all possible haplotypes from these 3 SNPs) in bothhe UCSF and VCU cohorts (Figure 3A, P � .02 and P

.0004, respectively). When the window shifted 1 baseo include the original DDX5 marker (SNP7 in window), the association still remained significant, althoughhe global P values were slightly larger (P � .04 inCSF, P � .0005 in VCU). Therefore, SNPs 4, 5, 6, andwere picked for further analysis. Since SNP 4 and 6 inOLG2 were in almost complete linkage disequilibrium

r2 � 0.973 and 0.969, in cases and controls, respec-ively), we focused on SNPs 5, 6, and 7 for haplotype andiplotype analysis to study further the risk allele(s) inhis region.

Figure 3B lists all diplotypes present at over 1%requency in both the UCSF and the VCU cohorts. Fouriplotype groups were categorized based on carriers of 1r 2 copies of haplotypes CAC, CAA, AAA, and AGA,orresponding from the highest to the lowest odds ratios,ssociated with advanced fibrosis (Figure 3B). Individu-ls with diplotype group D1, containing 1 or moreopies of the CAC haplotype (carrying the risk alleles of

able 3. Genetic Polymorphisms Associated with Advanced F

Gene/SNP Study Genotype

Casesa

(Advancedfibrosis)

(%)

Co(Mfi

DX5 UCSF AC � CCe 21 (0.18) 3CV7450990(rs3179861)

AA 99 (0.82) 27

VCU AC � CCe 34 (0.15) 1AA 200 (0.85) 23

UCSF � VCUPT1A UCSF CT � TTg 4 (0.03) 3CV15851335(rs17610395)

CC 117 (0.97) 27

VCU CT � TTg 20 (0.09) 3CC 214 (0.91) 21

UCSF � VCU

CSF, University of California San Francisco; VCU, Virginia CommonwCases were defined as patients with fibrosis stage 3 � 4.Controls were defined as patients with fibrosis stage 0–2 at UCSF,Model 1: unadjusted analysis.Model 2: adjusted for sex, alcohol, and ethnicity.Genotypes AC and CC were combined because of the infrequency oFisher combined P value.16

Genotypes CT and TT were combined because of the infrequency of

oth POLG2 SNP5 and DDX5 SNP7), were at a signif- a

cant 2.53- to 3.30-fold increased risk for having ad-anced fibrosis, in the UCSF and VCU cohorts, respec-ively. Individuals with diplotype group D2, containingor more copies of the CAA haplotype (carrying the risk

llele of only POLG2 SNP5), were also at a significant.88- to 2.20-fold increased risk, in the UCSF and VCUohorts, respectively, but lower than D1 (Figure 3B).he results suggest that both the original DDX5 SNP7nd its neighboring POLG2 SNP5 contributed to thencreased risk for advanced fibrosis in patients withhronic hepatitis C. Interestingly, compared with theriginal DDX5 marker, diplotype group D1 had higherRs and was present at similar frequency in the UCSF/CU cohorts; diplotype group D2 had similar ORs butas present at markedly higher frequency (Table 3,igure 3B).

Discussion

Although some HCV patients develop progressivebrosis and cirrhosis, many patients have mild liverisease for decades and have no morbidity or mortalityhat could be attributed to HCV infection.23 The mostecent NIH Consensus Conference recommended that allCV patients could be considered for therapy regardless

f serum alanine aminotransferase and degree of fibrosis.2

nfortunately, such an approach exposes many patientsith mild disease to the adverse events and high cost

ssociated with pegylated interferon and ribavirin ther-

is

sb

os)

Model 1c Model 2d

OR 95% CI P value OR 95% CI P value

0) 1.8 1.0–3.3 .048 1.7 0.9–3.1 .0830)

7) 2.3 1.3–4.3 .004 2.2 1.2–4.1 .0073)

.002f .005f

1) 0.3 0.1–0.8 .015 0.3 0.1–0.9 .0279)

3) 0.6 0.3–1.1 .050 0.6 0.3–1.1 .0487)

.006f .010f

University.

at VCU.

genotype.

notype.

ibros

ntrolild-n

brosi(%)

2 (0.15 (0.9

7 (0.02 (0.9

5 (0.17 (0.8

3 (0.16 (0.8

ealth

0–1

f CC

py. Ideally, these risks could be minimized if testing

wdirs

fiaswwtap

dbttpitgDspsg

Fec

1684 HUANG ET AL GASTROENTEROLOGY Vol. 130, No. 6

ere available to identify which HCV patients with mildisease were at risk of developing advanced fibrosis. Moremportantly, early identification and treatment of high-isk subjects, before the development of advanced fibro-is, could significantly increase treatment response rate.24

Previously identified clinical risk factors for advancedbrosis included male sex, excessive alcohol use, olderge at the time of infection, obesity, and the presence ofteatosis on liver biopsy.3–5 Unfortunately, many patientsith these characteristics have mild disease, and othersithout these factors have cirrhosis.25 Thus, although

hese factors have been associated with advanced fibrosisnd cirrhosis in large cohorts from an epidemiologic

igure 2. Additional SNPs in DDX5 region and risk of advanced fibthnicity. bHap window indicates the 3 SNPs included for each windandidates for diplotype analysis.

erspective, these factors simply cannot accurately pre- m

ict which patients with none or mild fibrosis on a liveriopsy now will develop cirrhosis in the future. We haveherefore hypothesized that host genetic factors are likelyo be the primary factor driving fibrosis progression inatients with chronic HCV. Indeed, the genetic markersdentified in this study have multiple advantages overraditional clinical risk factors (Figure 4). First, 4 novelenetic components, DDX5 and CPT1A SNPs and 2DX5-POLG2 diplotype groups, were significantly as-

ociated with advanced fibrosis in both cohorts. In com-arison, sex was the only clinical risk factor that wasignificant in both cohorts. Second, the newly identifiedenetic factors also had comparable or greater risk esti-

. *SNP7 is the original DDX5 SNP. aAdjusted for sex, alcohol, andigure 3A). cSNPs in bold are the ones in Hap windows 4 and 5 and

rosisow (F

ates than those using clinical risk factors. Third, the

phcfsaw

sbnmoiptoNfi

petmeisddmlbbi

Dws

Fivdfijn

May 2006 FIBROSIS GENETIC MARKERS 1685

ercentage of the population carrying the risk allele/aplotype for genetic factors was consistent among theohorts, in contrast to the large variance for clinical riskactors. Finally, the genotyping results were accurate,ubjective, and available for all patients. In contrast,ccurate estimates of alcohol intake and age at infectionere difficult to obtain and subject to recall bias.The SNPs associated with advanced fibrosis in this

tudy were identified by comparing HCV patients withridging fibrosis or cirrhosis with patients with eitherone or mild fibrosis. By grouping patients in thisanner, we believe that we have maximized our chances

f identifying SNPs associated with fibrosis progressionf such SNPs really existed. We purposely did not groupatients according to fibrosis progression rate becausehis is a value calculated from fibrosis stage and durationf infection: both of which are highly prone to error.umerous studies have demonstrated that assigning a

igure 3. (A) Haplotype windown DDX5 region and risk of ad-anced firbosis. (B) DDX5-POLG2iplotypes and risk of advancedbrosis. aAll analyses were ad-usted for sex, alcohol, and eth-icity.

brosis stage to a liver biopsy specimen from a HCV l

atient is based on liver biopsy size and quality and thexperience of the examining pathologist.26–29 Even underhe best of circumstances, staging of liver biopsy speci-ens may vary for up to 10% of the cohort being

valuated. However, such variation is nearly always lim-ted to a single stage.26–29 Furthermore, different stagingystems may be utilized by different centers and even byifferent pathologists within the same center. Thus, byividing our biopsies into 2 groups, those with none toild and those with bridging fibrosis-cirrhosis, we be-

ieve that we have eliminated the need to utilize the sameiopsy scoring system and the need to re-review alliopsies by a single pathologist and have minimized thempact of sampling error.

Interestingly, the published functional mechanisms ofDX5 and CPT1A, the 2 genes found to be associatedith advanced fibrosis in this study, provide some in-

ight into their potential roles in fibrogenesis, particu-

arly in patients with chronic hepatitis C. DEAD box

pADtsRethRpCfimigi

(tddtd

att(ssmlhla

tt2ahapwdaIauCo

Frf9alfwe

1686 HUANG ET AL GASTROENTEROLOGY Vol. 130, No. 6

roteins, characterized by the conserved motif Asp-Glu-la-Asp (DEAD), are putative RNA helicases.30 TheDX5 (previously called p68) gene is expressed in mul-

iple tissues including the liver.31 Goh et al demon-trated that DDX5 interacts with NS5B, the HCVNA-dependent RNA polymerase, and that depletion of

ndogenous DDX5 correlates with a reduction in theranscription of HCV RNA, suggesting that DDX5 is auman cellular factor that interacts with HCV RNA.32

ecent studies found that HCV NS5 and other viralroteins interact and activate hepatic stellate cells.33,34

ombined with our observed association of DDX5 withbrosis risk in HCV patients, it is possible that DDX5ight not only influence liver cell injury in HCV-

nfected hepatocytes but may also directly affect fibro-enesis in activated stellate cells, the key fibrogenic celln liver.35

The other risk gene, the liver isoform of CPT1CPT1A), is the key enzyme in the carnitine-dependentransport of long-chain fatty acid across the mitochon-rial inner membrane. Deficiency of CPT1A results in aecreased rate of hepatic fatty acid �-oxidation, charac-erized by hypoketotic hyperglycemia.36 Gobin et al

igure 4. Clinical and genetic factors: risk of advanced fibrosis. Oddsatios. To make the results comparable, all clinical and genetic riskactors were adjusted for sex, alcohol, and ethnicity Lines indicate5% confidence intervals of odds ratios; numbers next to each linere P values. Dotted line indicates an odds ratio equals to 1. aPopu-ation at risk indicates the percentage of subjects carrying the riskactors. bThe odds ratio of CPT1A SNP was reversed to be consistentith other factors; therefore, the odds ratio for this SNP should bexplained as protective effect.

emonstrated that A275T was located more than 20 Å

way from the catalytic site, causing 25% reduction ofhe enzymatic activity.37 On the other hand, we foundhat this marker was not associated with steatosis gradeP � .49) among the approximately 50% of UCSFubjects who had histologic steatosis grade (data nothown). Taken together, we hypothesized that A275Tildly reduced CPT1A enzymatic activity, causing the

ower rate of fatty acid �-oxidation but not affecting theepatic steatosis; thus, the net outcome of A275T wasess oxidative stress, as a result, decreasing the risk ofdvanced fibrosis.

At this time, only the first 100 consecutive SNPs ofhe 1609 found to be associated with advanced fibrosis inhe discovery cohort have undergone validation. To date,of these 100 SNPs have been found to be significantly

ssociated with advanced fibrosis in patients with chronicepatitis C. The remaining 1509 SNPs are being evalu-ted in ongoing studies. Thus, although it is certainlyossible that additional SNPs may also be associatedith fibrosis progression, the current findings clearlyemonstrate that the development of bridging fibrosisnd cirrhosis is at least in part because of host genetics.dentifying and understanding genetic markers associ-ted with fibrosis risk has the potential to enhance ournderstanding of the natural history of chronic hepatitisand improve the cost-effectiveness and global impact

f treating this disease.

References1. National Institutes of Health Consensus Development Confer-

ence Statement: Management of hepatitis C: June 10–12, 2002.Hepatology 2002;36:S3–20.

2. Strader DB, Wright W, Thomas DL, Seeff LB. AASLD practiceguideline: diagnosis, management, and treatment of hepatitis C.Hepatology 2004;39:1147–1171.

3. Poynard T, Bedossa P, Opolon P. Natural history of liver fibrosisprogression in patients with chronic hepatitis C. The OBSVIRC,METAVIR, CLINIVIR, and DOSVIRC groups. Lancet 1997;349:825–832.

4. Wright M, Goldin R, Fabre A, Lloyd J, Thomas H, Trepo C, PradatP, Thursz M. Measurement and determinants of the natural his-tory of liver fibrosis in hepatitis C virus infection: a cross-sectionaland longitudinal study. Gut 2003;52:574–579.

5. Ramesh S, Sanyal AJ. Hepatitis C and nonalcoholic fatty liverdisease. Semin Liver Dis 2004;24:399–413.

6. Bataller R, North KE, Brenner DA. Genetic polymorphisms and theprogression of liver fibrosis: a critical appraisal. Hepatology2003;37:493–503.

7. Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A compre-hensive review of genetic association studies. Genet Med 2002;4:45–61.

8. Colhoun HM, McKeigue PM, Davey Smith G. Problems of report-ing genetic associations with complex outcomes. Lancet 2003;361:865–872.

9. Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN.Meta-analysis of genetic association studies supports a contri-bution of common variants to susceptibility to common disease.

Nat Genet 2003;33:177–182.

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

nm

May 2006 FIBROSIS GENETIC MARKERS 1687

0. Risch NJ. Searching for genetic determinants in the new millen-nium. Nature 2000:405:847–856.

1. Brunt EM. Grading and staging the histopathological lesions ofchronic hepatitis: the Knodell histology activity index and beyond.Hepatology 2000;31:241–246.

2. Monto A, Patel K, Bostrom A, Pianko S, Pockros P, McHutchisonJG, Wright TL. Risks of a range of alcohol intake on hepatitisC-related fibrosis. Hepatology 2004;39:826–834.

3. Bedossa P, for the French METAVIR Cooperative Group. Inter- andintraobserver variation in the assessment of liver biopsy ofchronic hepatitis C. Hepatology 1994;20:15–20.

4. Skinner HA, Sheu WJ. Reliability of alcohol use indices. TheLifetime Drinking History and the MAST. J Stud Alcohol 1982;43:1157–1170.

5. Germer S, Holland MJ, Higuchi R. High-throughput SNP allele-frequency determination in pooled DNA samples by kinetic PCR.Genome Res 2000;10:258–266.

6. Fisher RA. Statistical methods for research workers. 12th ed.Edinburgh: Oliver & Lloyd, 1954.

7. Breslow NE, Day NE. Statistical methods in cancer research.Volume I. The analysis of case-control studies. IARC Sci Publ1980;32:5–338.

8. The International HapMap Consortium. The International Hap-Map Project. Nature 2003;426:789–796.

9. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis andvisualization of LD and haplotype maps. Bioinformatics 2005;21:263–265.

0. Schaid DJ, Rowland CM, Tines DE, Jacobson RM, Poland GA.Score tests for association between traits and haplotypes whenlinkage phase is ambiguous. Am J Hum Genet 2002;70:425–434.

1. Yoo J, Seo B, Kim Y. SNPAnalyzer: web-based workbench for theSNP analysis (http://istech21.com/en/service/snp_a01.html).Genomic Inform 2003;14:591–592.

2. Chen J, Germer S, Higuchi R, Berkowitz G, Godbold J, Wetmur JG.Kinetic polymerase chain reaction on pooled DNA: a high-through-put, high-efficiency alternative in genetic epidemiological studies.Cancer Epidemiol Biomarkers Prev 2000;11:131–136.

3. Seeff LB. Natural history of chronic hepatitis C. Hepatology 2002;36:S35–S46.

4. Zeuzem S, Feinman SV, Rasenack J, Heathcote EJ, Lai MY, GaneE, O’Grady J, Reichen J, Diago M, Lin A, Hoffman J, Brunda MJ.Peginterferon alfa-2a in patients with chronic hepatitis C. N EnglJ Med 2000;343:1666–1672.

5. Marcellin P, Asselah T, Boyer N. Fibrosis and disease progres-

sion in hepatitis C. Hepatology 2002;36:S47–S56.

6. Regev A, Berho M, Jeffers LJ, Milikowski C, Molina EG, Pyrsopou-los NT, Feng ZZ, Reddy KR, Schiff ER. Sampling error and intraob-server variation in liver biopsy in patients with chronic HCV infec-tion. Am J Gastroenterol 2002;97:2614–2618.

7. Bedossa P, Dargere D, Paradis V. Sampling variability of liverfibrosis in chronic hepatitis C. Hepatology 2003;38:1449–1457.

8. Afdal NH, Nunes D. Evaluation of liver fibrosis: a concise review.Am J Gastroenterol 2004;99:1160–1174.

9. Rousselet MC, Michalak S, Dupre F, Croue A, Bedossa P, Saint-Andre JP, Cales P. Sources of variability in histological scoring ofchronic viral hepatitis. Hepatology 2005;41:257–264.

0. Rocak S, Linder P. DEAD-box proteins: the driving forces behindRNA metabolism. Nat Rev Mol Cell Biol 2004;5:232–241.

1. Rossler OG, Hloch P, Schutz N, Weitzenegger T, Stahl H. Struc-ture and expression of the human p68 RNA helicase gene.Nucleic Acids Res 2000;28:932–939.

2. Goh PY, Tan YJ, Lim SP, Tan YH, Lim SG, Fuller-Pace F, Hong W.Cellular RNA helicase p68 relocalization and interaction with thehepatitis C virus (HCV) NS5B protein and the potential role of p68in HCV RNA replication. J Virol 2004;78:5288–5298.

3. Bataller R, Paik YH, Lindquist JN, Lemasters JJ, Brenner DA.Hepatitis C virus core and nonstructural proteins induce fibro-genic effects in hepatic stellate cells. Gastroenterology 2004;126:529–540.

4. Mazzocca A, Sciammetta SC, Carloni V, Cosmi L, Annunziato F,Harada T, Abrignani S, Pinzani M. Binding of hepatitis C virusenvelope protein E2 to CD81 up-regulates matrix metalloprotein-ase-2 in human hepatic stellate cells. J Biol Chem 2005;280:11329–11339.

5. Friedman SL. Molecular regulation of hepatic fibrosis, an inte-grated cellular response to tissue injury. J Biol Chem 2000;275:2247–2250.

6. Bonnefont JP, Djouadi F, Prip-Buus C, Gobin S, Munnich A, BastinJ. Carnitine palmitoyltransferases 1 and 2: biochemical, molec-ular and medical aspects. Mol Aspects Med 2004;25:495–520.

7. Gobin S, Thuillier L, Jogl G, Faye A, Tong L, Chi M, BonnefontJP, Girard J, Prip-Buus C. Functional and structural basis ofcarnitine palmitoyltransferase 1A deficiency. J Biol Chem2003;278:50428–50434.

Received July 20, 2005. Accepted February 8, 2006.Address requests for reprints to: Hongjin Huang, PhD, Celera Diag-

ostics, 1401 Harbor Bay Parkway, Alameda, California 94502. e-ail: hongjin.huang@celeradiagnostics.com; fax: (510) 749-6200.

Supported by Celera Diagnostics.