Transcriptome at the time of hepatitis C virus recurrence may predict the severity of fibrosis progression after liver transplantation

ORIGINAL ARTICLE

Transcriptome at the Time of Hepatitis C VirusRecurrence May Predict the Severity of FibrosisProgression After Liver TransplantationValeria Mas,1,2 Daniel Maluf,1 Kellie J. Archer,1,3 Amiee Potter,1 Jihee Suh,1 Ricardo Gehrau,1

Valeria Descalzi,4 and Federico Villamil51Division of Transplantation, Department of Surgery, 2Department of Pathology, and 3Department ofBiostatistics, Virginia Commonwealth University, Richmond, VA, 4Liver Unit, Favaloro Foundation, BuenosAires, Argentina; and 5Liver Transplantation Unit, British Hospital, Buenos Aires, Argentina

Allograft gene expression analysis may provide insights into the mechanisms involved in liver damage during hepatitis C vi-rus recurrence (HCVrec) after orthotopic liver transplantation (OLT) and allow the identification of patients who have ahigher risk of developing severe disease. Forty-three OLT recipients with hepatitis C virus (HCV) were evaluated. Genome-wide gene expression analysis was performed with formalin-fixed, paraffin-embedded (FFPE) liver biopsy samples obtainedfrom 21 OLT recipients with HCV at the time of clinical HCVrec, which was defined as increased alanine aminotransferaselevels and detectable HCV RNA levels in serum. Patients were classified into 3 groups according to the severity of the fibro-sis in the liver biopsies at 36 months post-OLT : group 1 (G1) for mild fibrosis (F0-F1), group 2 for moderate fibrosis (F2),and group 3 (G3) for severe fibrosis (F3-F4). No significant differences were observed between the groups with respect to

Additional Supporting Information may be found in the online version of this article.

Abbreviations: ACE, angiotensin-converting enzyme; AIC, Akaike information criterion; AIRE, autoimmune regulator; ARMC8,armadillo repeat containing 8; ATF, activating transcription factor; ATP8A2, ATPase, aminophospholipid transporter, class I, type8A, member 2; C3orf30, chromosome 3 open reading frame 30; C4orf33, chromosome 4 open reading frame 33; C10orf4, chromo-some 10 open reading frame 4; CCDC, coiled-coil domain containing; CEACAM3, carcinoembryonic antigen-related cell adhesionmolecule 3; CEP, centrosomal protein; CETN1, centrin EF-hand protein 1; CLIP, CAP-GLY domain containing linker protein;CLOCK, clock homolog; CSPG4, chondroitin sulfate proteoglycan 4; CTH, cystathionase; DASL, complementary DNA–mediatedannealing, selection, extension, and ligation; DCHS1, dachsous 1; DGAT1, diacylglycerol O-acyltransferase 1; DLK1, delta-like 1homolog; DNAI1, dynein, axonemal, intermediate chain 1; DST, dystonin; DYNC1H1, dynein, cytoplasmic 1, heavy chain 1;EIF2AK4, eukaryotic translation initiation factor 2 alpha kinase 4; ERBB2IP, erbb2 interacting protein; ERK, extracellular signal-regulated kinase; FFPE, formalin-fixed, paraffin-embedded; FUT8, fucosyltransferase 8; Fx, function; G1, group 1; G2, group 2;G3, group 3; GAGE6, G antigen 6; GPR37L1, G protein-coupled receptor 37 like 1; GSTCD, glutathione S-transferase, C-terminaldomain containing; HCN1, hyperpolarization activated cyclic nucleotide-gated potassium channel 1; HCV, hepatitis C virus;HCVrec, hepatitis C virus recurrence; HRH4, histamine receptor H4; HSBP1, heat shock factor binding protein 1; IFRD1, inter-feron-related developmental regulator 1; IL, interleukin; ILMN, Illumina probe identifier; INSL6, insulin-like 6; ISG, interferon-stimulated gene; KCNJ2, potassium inwardly rectifying channel, subfamily J, member 2; KCNMB2, potassium large conductancecalcium-activated channel, subfamily M, beta member 2; KRTAP6-3, keratin-associated protein 6-3; LASSO, least absoluteshrinkage and selection operator; LHFPL2, lipoma HMGIC fusion partner-like 2; LOC387856, hypothetical protein LOC387856;MGAT4A, mannosyl (alpha-1,3)-glycoprotein beta-1,4-N-acetylglucosaminyltransferase, isozyme A; MMP, matrix metalloprotei-nase; MTHFD2, methylenetetrahydrofolate dehydrogenase 2; MXI1, MAX interactor 1; N, total sample size; NDEL1, nuclear distri-bution gene E homolog (A. nidulans)-like 1; NLRP7, NLR family, pyrin domain containing 7; OLT, orthotopic liver transplantation;OR, olfactory receptor; PDE4A, phosphodiesterase 4A; PHTF2, putative homeodomain transcription factor 2; PIBF1, progesteroneimmunomodulatory binding factor 1; PLS, polymerase; POLE, polymerase (DNA directed) epsilon; QPCR, quantitative real-time po-lymerase chain reaction; RAS, renin-angiotensin system; RASD1, renin-angiotensin system dexamethasone-induced 1; RCAN1,regulator of calcineurin 1; REV3L, REV3-like; SCN2A, sodium channel, voltage-gated, type II, alpha subunit; SDCBP2, syndecanbinding protein (syntenin) 2; SEMA3A, sema domain 3A; SLC, solute carrier; SMC3, structural maintenance of chromosomes 3;SNAPC3, small nuclear RNA activating complex, polypeptide 3, 50 kDa; SOS, son of sevenless; SPATA17, spermatogenesis associ-ated 17; STIM2, stromal interaction molecule 2; STS1, suppressor of T cell receptor signaling 1; SYT3, synaptotagmin III; TMEM,transmembrane protein; TNRC6A, trinucleotide repeat containing 6A; TTLL1, tubulin tyrosine ligase-like family, member 1;UBE2C, ubiquitin-conjugating enzyme E2C; WGE, whole genome gene expression; XKR6, XK Kell blood group complex subunit-related family, member 6; ZNF, zinc finger protein.

Address reprint requests to Valeria Mas, Ph.D., Division of Transplantation, Department of Surgery, Virginia Commonwealth University, WestHospital, 9th Floor, 1200 East Broad Street, P.O. Box 980057, Richmond, VA 23298-0057. Telephone: 804-828-2364, Fax: 804-828-2462; E-mail:[email protected]

DOI 10.1002/lt.22309View this article online at wileyonlinelibrary.com.LIVER TRANSPLANTATION.DOI 10.1002/lt. Published on behalf of the American Association for the Study of Liver Diseases

LIVER TRANSPLANTATION 17:824–835, 2011

VC 2011 American Association for the Study of Liver Diseases.

donor age, histology during HCVrec, treated episodes of acute cellular rejection, or immunosuppression therapy. The resultswere validated in the remaining 22 OLT recipients with HCV using quantitative real-time polymerase chain reaction. Fifty-seven beadtypes showed significantly different expression (P < 0.001) between the groups during HCVrec. In G3, the geneexpression of interleukin-28RA (IL-28RA), IL-28, and angiotensin-converting enzyme was up-regulated. Samples from G1and G3 were used to determine whether a multigenetic classifier could be derived to predict the group class. The finalmodel included the intercept and 9 bead types. Pairwise scatter plots of these 9 bead types revealed that G1 and G3 werewell separated with respect to each gene. Our analysis has demonstrated the utility of a set of molecular markers indicatingHCVrec severity early after OLT. Liver Transpl 17:824–835, 2011. VC 2011 AASLD.

Received December 10, 2010; accepted March 18, 2011.

Cirrhosis due to hepatitis C virus (HCV) infection is theleading cause of orthotopic liver transplantation (OLT)among adults in the Western world.1-3 The natural his-tory of HCV infection after OLT has been well defined.Reinfection of the liver allograft occurs rapidly and uni-versally.3-7 The majority of patients with virological re-currence will develop histological graft injuries, withapproximately one-third of the patients developing cir-rhosis within only 5 years of follow-up.3-7 In addition,the time from HCV cirrhosis to decompensation is sig-nificantly shorter in OLT recipients versus nontrans-plant patients.3-8 Not surprisingly, in comparison withother indications for OLT, HCV is associated withdecreased patient and graft survival.8-10

The mechanisms responsible for the acceleratedliver damage in HCV-infected OLT recipients remainlargely unknown. Numerous donor, recipient, andpost-OLT factors such as the age of the donor5,11-13

and the type and intensity of the immunosuppres-sion14-16 have been proposed for identifying patientsdestined to develop severe hepatitis C virus recur-rence (HCVrec). However, no single variable appearsto be accurate enough to independently predict long-term histological outcomes.

Moreover, high variability in the severity of diseaseprogression has been observed among patients, andthis indicates the need for early biomarkers that canbe used to identify those patients at high risk foraggressive fibrosis development.

The analysis of allograft gene expression signaturesnot only may provide insights into the biology ofHCVrec but also may help us to identify patients whohave a higher risk of developing severe disease andmay thus benefit from early therapeutic interventions.The analysis of protocol liver biopsy samples obtainedduring long-term follow-up is at present the most accu-rate way of monitoring the progression of HCVrec. Inaddition, protocol biopsy samples are a unique sourceof sequentially obtained graft materials for studying themolecular pathogenesis of HCVrec.7,17-19 Formalinfixed, paraffin embedded (FFPE) liver tissue can pro-vide RNA that until recently was unfit for standardanalyses by microarray-based methods20 because ofextensive RNA degradation.21 However, a new methodfor processing FFPE samples is now available.22

In this study, we tested the complementary DNA–mediated annealing, selection, extension, and ligation(DASL) methodology for gene expression profiling inFFPE liver tissue to identify gene expression patterns

early after liver transplantation that are associatedwith the histological severity of HCVrec.

PATIENTS AND METHODS

Patients and Samples

The study included 43 adults who underwent OLT forHCV cirrhosis at the Favaloro Foundation (BuenosAires, Argentina) between 1995 and 2006. All patientscompleted at least 3 years of follow-up post-OLTand underwent liver biopsy at that time for theassessment of the severity of their recurrent disease.Post-OLT immunosuppression consisted of calci-neurin inhibitors [29 (67.4%) received cyclosporine A,and 14 (32.6%) received tacrolimus] and corticoste-roids [which were withdrawn during the first yearfrom 39 patients (91%)]. Seven patients receivedinduction therapy [OKT3 (4), anti-thymocyte globulin(1), or basiliximab (2)], and 33 (77%) received a thirdimmunosuppressive agent at some time during follow-up [mycophenolate (28) or azathioprine (5)]. Patientswith biopsy-proven acute rejection (19 or 44%)received 1 intravenous bolus of 1 g of methylpredniso-lone, which was followed by an oral recycle of predni-sone with a starting dose of 200 mg/day.

Postoperatively, all patients underwent clinical andbiochemical monitoring at regular but variable inter-vals (biweekly, weekly, every other week, and monthlythereafter) according to individual outcomes. Whenaminotransferase levels were found to be elevated,serum HCV RNA and liver biopsy samples wereobtained. Liver biopsy was performed according to pro-tocol (at the time of HCVrec, 1 year after OLT, andannually thereafter) and when it was clinically indi-cated. Liver tissue obtained at the time of HCVrec wasanalyzed using gene expression, whereas the long-termseverity of recurrent disease was based on the histolog-ical analysis of biopsy samples obtained 3 years afterOLT. The median number of biopsy procedures perpatient was 6 (range ¼ 4-14). According to the META-VIR scoring system,23 fibrosis was scored as F0(absent), F1 (portal fibrosis), F2 (portal fibrosis withfew septa), F3 (septal fibrosis), or F4 (cirrhosis). Inaddition, necroinflammation activity was graded as A0(absent), A1 (mild), A2 (moderate), or A3 (severe).According to the severity of their allograft fibrosis3 years post-OLT, the patients were placed in 3 groups:

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 825

(1) G1 for mild fibrosis (F0-F1, n ¼ 16), (2) G2 for mod-erate fibrosis (F2, n ¼ 14), and (3) G3 for severe fibrosis(F3-F4, n ¼ 13). No patient received antiviral therapyfor HCV for the duration of the study. Genome-widegene expression was initially analyzed in 21 patientsand was subsequently validated in the remaining 22patients. The study protocol conformed to the ethicalguidelines of the 1975 Declaration of Helsinki; this wasreflected by the a priori approval of the humanresearch committee of the Favaloro Foundation.

RNA Isolation From FFPE Liver Biopsy Samples

Total RNA was isolated from FFPE liver biopsy sam-ples with the RecoverAll total nucleic acid isolation kit(Ambion).

Whole Genome Gene Expression (WGE)

DASL (Illumina)

Twenty-three unique samples were hybridized to 28BeadArrays; they included 21 unique samples pro-cured during HCVrec and 2 normal liver samples. Theunique samples taken at the time of HCVrec and usedin the WGE DASL assay were taken from G1 (8), G2(5), and G3 (8).

The reactions were performed according to theinstructions of the WGE DASL manual (Illumina). Theresulting polymerase chain reaction productswere hybridized to HumanRef-8 Expression Bead-Chips (Illumina). After hybridization, the HumanRef-8BeadChips were scanned with the BeadArray reader(Illumina).

Quality Control

Before the statistical analyses, the quality of all Bead-Arrays was assessed by examinations of (1) the per-centage of present calls (defined as the percentage ofbead types with a detection call P value <0.05) and (2)the plots of signal intensities for housekeeping,cy3 hybridization, low stringency hybridization,labeling, and biotin control bead types. Also, 4 sam-ples were evaluated in duplicate to check the assayreproducibility. The RNA purity was determined with aspectrophotometer; the ratio of the absorbance at 260nm to the absorbance at 280 nm and the ratio of theabsorbance at 260 nm to the absorbance at 270 nmwere estimated. In addition, the RNA integrity wasassessed with an Agilent bioanalyzer. Moreover, 3 dif-ferent housekeeping genes were used for QPCR toidentify samples of sufficient quality for the assays.The housekeeping genes included ribosomal proteinL13a (RPL13a) (QPCR RPL13a has been reported tohave greater predictive power for the assessment ofRNA performance in the DASL assay24), glucuronidaseb, and glyceraldehyde 3-phosphate dehydrogenase.

Microarray Data Analysis

Image analysis was performed with Illumina Bead-Scan software and thereafter was processed with the

Illumina BeadStudio gene expression module (version3.4.0). Gene expression was first estimated as theaverage of the unnormalized, non–background-sub-tracted signals for beads within a bead type for the24,526 noncontrol bead types for each of the 23 sam-ples. Thereafter, quantile normalization of the non–background-subtracted bead intensities was used.For each bead type, to identify differentially expressedgenes at the time of HCVrec in the 3 groups (G1, G2,and G3), a linear model was fitted to predict geneexpression (the group was used as the independentpredictor variable). A moderated F test from the LinearModels for Microarray Data package was used fordetermining whether there was a significant differencewith respect to the groups and the P values. To con-trol for multiple-hypothesis testing, bead types with aP value <0.001 were considered significant. The sig-nificance of all possible linear contrasts was alsoexamined with the P value from each linear contrast.

Interaction Networks and Functional Analysis

Gene ontology and gene interaction analyses wereexecuted with Ingenuity Pathways Analysis tools(version 7.0, Ingenuity Systems).

Class Prediction Analysis

The least absolute shrinkage and selection operator(LASSO) is a penalized method for estimating a logis-tic regression model when number of covariates (p) isgreater than number of samples (n) and there is colli-nearity among the candidate predictors.25-27 TheLASSO model was fitted for class prediction; the finalselected model had the minimum Akaike informationcriterion (AIC) with the Generalized Linear ModelsPath package in the R programming environment. Theminimum AIC was selected over the minimum Bayes-ian information criterion on the basis of a large simu-lation study (that study concluded that although theBayesian information criterion tends to select the cor-rectly sized model, the AIC more often includes a non-zero coefficient estimate for the true predictor).28 Thepredicted class was G3 if the fitted probability was�0.50 and was G1 otherwise. To obtain an unbiasedestimate of the classification error, we used leave-one-out cross-validation, which is also called N-fold cross-validation (N is the total sample size). All analyseswere conducted in the R programming environment29

with appropriate Bioconductor packages.30 TheLASSO models were fit with the Generalized LinearModels Path package.31

QPCR

QPCR was used to confirm the WGE DASL reactionsand to validate genes identified as differentiallyexpressed (false discovery rate < 10%) by the groupsduring HCVrec (see the Supporting Information formore information on QPCR).

826 MAS ET AL. LIVER TRANSPLANTATION, July 2011

Statistical Analysis

Descriptive statistics are reported for demographicand clinical variables: proportions for categorical vari-ables and means and standard deviations for continu-ous variables.

RESULTS

Clinical Data

The mean recipient age was 53.5 6 8.0 years, and 25of the recipients (58%) were male. The majority of thepatients (36/43 or 84%) were infected with HCV geno-type 1, and the remaining 7 patients were infectedwith genotype 2 or 3. Eight patients (19%) had hepa-tocellular carcinoma, and 9 (21%) underwent livingdonor transplantation. The mean donor age was 37.56 15 years for deceased donors and 34 6 10 years forliving donors. Only 9 donors (21%) were older than50 years. The time to HCVrec was 5.0 6 3.8 monthspost-OLT. Liver biopsy samples obtained at the timeof clinical and virological recurrence showed minimalportal and/or lobular inflammation in 15 patients(35%), acute hepatitis in 8 patients (18.6%), andchronic hepatitis in 20 patients (46.5%). The preva-lence of biopsy-proven and treated acute rejection was44% (19/43 patients). The mean follow-up post-OLTwas 8.3 6 2.6 years. Seven patients (16%) underwentre-OLT, and 8 (18%) died. The histological analysisrevealed progressive increases in the fibrosis stageover time (Table 1) and showed a potential associationbetween fibrosis 1 year post-OLT and disease progres-sion. The proportion of patients with mild recurrentdisease (F0-F1) decreased from 70% at 1 year to37.5% at 3 years, whereas the proportion of patientswith severe recurrence (F3-F4) increased from 5% to30% in the same time (Table 1). As shown in Table 2,no significant differences were observed between thegroups with respect to most of the analyzed variables(ie, donor and recipient demographics, prevalence ofhepatocellular carcinoma, cold ischemia time, grafthistology at recurrence, time to HCVrec, type of calci-neurin inhibitor, and rates of treated acute cellularrejection). A possible explanation for why donor agewas not significantly associated with fibrosis progres-sion is that the mean donor age was only 37.5 years.Moreover, the lack of association between the clinicalvariables and the severity of disease progression may

be related to the relatively small sample size. Patientsfrom G2 had a significantly lower prevalence of HCVgenotype 1 and significantly higher alanine amino-transferase levels at the time of HCVrec. Significantdifferences were observed in donor/recipient gendermatching between G1 and G2 and in post-OLT mor-tality rates between G1 and G3 (Table 2). Death dueto HCVrec occurred in 1 of 3 G2 patients and in 3 of 5G3 patients.

Quality Control

After the overall quality of the data produced for theassay was evaluated, only BeadArrays with a presentcall value higher than 30% were used for the analysis(Supporting Information Fig. 1). Additional evalua-tions were performed to ensure the quality of the data(see the Supporting Information for more informationon quality control analysis).

Allograft Gene Expression Patterns

at the Time of Clinical HCVrec

Quantile normalization of the non–background-sub-tracted bead intensities was used to enforce a consis-tent distribution of intensities across all hybridizedsamples (Supporting Information Fig. 2). To analyzethe molecular pathways characterizing HCV infectionsin liver transplant recipients, we first compared geneexpression profiles for liver biopsy samples collectedduring HCVrec with those for normal livers. This anal-ysis showed significant differences (P < 0.001) in 851bead types; 250 were down-regulated, and 601 wereup-regulated (Supporting Information Table 1).

A core analysis was performed to interpret the dataset in the context of biological processes, pathways,and molecular networks. Among the 851 significantbead types at the time of clinical HCVrec, 50 networkswere identified (20 had a score � 15). The top-scoringnetwork is shown in Fig. 1.

Within the top canonical pathways, a specific analy-sis of the cellular immune response during HCVrecshowed that IL-6 signaling (P ¼ 0.0165) was affectedby the overexpression of mitogen-activated protein ki-nase kinase kinase 71, son of sevenless 1 (SOS1),SOS2, and nuclear factor kappa B2 and by the under-expression of IL-F7 and IL-1F8 in comparison withnormal livers. The overexpression of genes involved in

TABLE 1. Severity of Fibrosis During HCVrec and Follow-Up (n 5 43)

Fibrosis Stage* During Recurrence [n (%)] 1 Year [n (%)] 2 Years [n (%)] 3 Years [n (%)]

F0 22 (51.16) 9 (20.93) 2 (4.65) 2 (4.65)F1 19 (44.19) 21 (48.84) 17 (39.53) 14 (32.56)F2 1 (2.33) 11 (25.58) 13 (30.23) 14 (32.56)F3 1 (2.33) 2 (4.65) 11 (25.58) 8 (18.6)F4 0 0 0 5 (11.63)

*METAVIR score.

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 827

acute responses (IL-6R, phosphoinositide-3-kinaseregulatory subunit 1, SOS1, and SOS2) and the up-regulation of genes involved in the chemotaxis of nat-ural killer cells [chemokine (C-X-C motif) ligand 10and guanine nucleotide binding protein alpha activat-ing activity polypeptide O] were observed in samplesobtained during HCVrec.

Antigen presentation signaling (eg, human leuko-cyte antigen DQA1, human leukocyte antigen DRB4,and proteasome subunit beta type 9) and IL-15 pro-duction (interferon regulatory factor 3, nuclear factorkappa B2, and protein tyrosine kinase 7) were alsoincreased in HCV-positive samples. The proteinencoded by the IL-15 gene is a cytokine that regulatesT cell and natural killer cell activation and prolifera-tion. Moreover, IL-4 and IL-10 signalings were pre-dominant in HCVrec samples versus normal livers.

Genes involved in interferon signaling were over-expressed during HCVrec; these included interferon-induced protein 35, interferon-induced protein withtetratricopeptide repeats 3, interferon-induced protein44, and 20,50-oligoadenylate synthetase 1 (a 14-foldchange).

Genes involved in apoptosis [tumor necrosis factor(ligand) superfamily, member 10; tumor necrosis

factor receptor superfamily, member 10b; and B celllymphoma 2–like 12] were overexpressed in HCVrecsamples. Growth factor signaling was activated duringHCVrec, with HGF signaling [GRB2-associated bind-ing protein 1 (a 6.55-fold change), SOS1 (a 7.79-foldchange), and phosphoinositide-3-kinase regulatorysubunit 1 (a 13.5-fold change)] and fibroblast growthfactor signaling [activating transcription factor 4(ATF4; a 31-fold change) and fibroblast growth factorreceptor 4 (an 18-fold change)] being the most impor-tant. In comparison with normal livers, the most rele-vant down-expressed genes in HCVrec samples werethose involved in the collagen catabolic process[matrix metalloproteinase 25 (MMP25) and MMP16]and antiapoptosis (serpin peptidase inhibitor B2 andsphingosine kinase 1).

Gene Expression Signatures During HCVrec

and Its Association With Fibrosis Progression

3 Years After Liver Transplantation

From the overall moderated F tests, there were 57bead types with significantly different expression (P <0.001; Fig. 2). When we examined linear contrasts, wefound significant differences in 14 bead types for G1

TABLE 2. Donor, Recipient, and Postoperative Variables for Patients With Mild (G1), Moderate (G2),

or Severe Fibrosis (G3)

G1 G2 G3

Patients (n) 16 14 13Recipient age (years)* 51 6 8 53 6 9 56 6 7Recipient gender: male/female [n (%)/n] 11 (69)/5 6 (43)/8 8 (61.5)/5Hepatocellular carcinoma [n (%)] 2 (12.5) 4 (29) 2 (15)HCV genotype: 1/2-3 [n (%)/n] 15 (94)/1 8 (57)/6† 13 (100)/0Living donors [n (%)] 3 (19) 2 (14) 4 (31)Donor age (years)* 36 6 16 35 6 14 40 6 12Deceased donor age (years)* 35 6 17 36 6 15 43 6 12Donor older than 50 years [n (%)] 4 (25) 2 (14) 3 (23)Donor gender: male/female [n (%)/n] 8 (50)/8 11 (79)/3 10 (77)/3Donor/recipient gender match [n (%)] 9 (56) 3 (21) 11 (85)‡

Cold ischemia time (minutes)* 536 6 201 510 6 196 433 6 233Time to recurrence (months)* 5.5 6 4.8 5.7 6 3.7 3.7 6 1.9Alanine aminotransferase at recurrence (IU/L)* 223 6 118 675 6 705§ 338 6 180Histology at recurrence [n (%)]Minimal inflammation 4 (25) 4 (29) 7 (54)Acute hepatitis 5 (31) 3 (21) 0Chronic hepatitis 7 (44) 7 (50) 6 (46)

Cyclosporine/tacrolimus [n (%)] 9 (56)/7 9 (64)/5 11 (85)/2Acute rejection [n (%)] 7 (44) 9 (64) 6 (46)Treated acute rejection [n (%)] 5 (31) 8 (57) 6 (46)Steroid boluses [n (%)] 5 (31) 8 (57) 6 (46)Retransplantation [n (%)] 0 2 (14) 3 (23)Death [n (%)] 0 3 (21) 5 (38)k

*The data are presented as means and standard deviations.†P ¼ 0.03 versus G1 and P ¼ 0.02 versus G3.‡P ¼ 0.004 versus G2.§P ¼ 0.03 versus G1.kP ¼ 0.01 versus G1.

828 MAS ET AL. LIVER TRANSPLANTATION, July 2011

and G2 (Supporting Information Table 2), in 5 beadtypes for G2 and G3 (Supporting Information Table3), and in 50 bead types for G1 and G3 (Table 1). AVenn diagram displaying the overlap between the sig-nificant bead types for the 3 linear contrasts is pre-sented in Fig. 2A.

A core analysis was performed to interpret the dataset in the context of biological processes, pathways,and molecular networks. Among the 57 significantbead types during clinical HCVrec, 14 networks wereidentified. Cell-to-cell signaling and interactions and

cell death were the principal functions associatedwith the top-scoring networks.

Our major interest, however, was the analysis of thedifferences in gene expression patterns during clinicalHCVrec between G1 (F0-F1) and G3 (F3-F4). From theanalysis of these 50 significant bead types, 9 networkswere identified. The top-scoring network (score ¼ 17)was associated with an infection mechanism, cellulardevelopment, and antigen presentation. The geneexpression data set was then filtered to include the 50bead types differentially expressed between G1 and G3

Figure 1. Top-scoring network of interactions between identified and quantified genes that were significantly differentially expressedby HCV allografts during clinical HCVrec and normal liver allografts. The interconnections of significant functional networks in HCVallograft biopsy samples taken during clinical HCVrec are presented; gene nodes are presented as red (up-regulation), green (down-regulation), or white (no change). The fold changes and significance are also indicated. Apoptosis signaling is labeled. Gene identifiersmarked with an asterisk (*) indicate that multiple identifiers in the dataset file map to a single gene in the Global Molecular Network.

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 829

(Table 3). Agglomerative hierarchical clustering withWard’s method was applied to the filtered data set withPearson’s correlation as the similarity measure (Fig. 2B).

The molecular and cellular functions associatedwith the 50 differentially expressed genes includedcarbohydrate metabolism, small-molecule biochemis-try, cell death, cell morphology, and cell-to-cell signal-ing and interactions.

The expression of IL-28RA and IL-28 was up-regu-lated in G3 patients versus G1 patients. Moreover, theexpression of ACE was up-regulated in the group ofpatients with faster progression to fibrosis 3 yearspost-OLT. An increasing body of evidence suggeststhat the renin-angiotensin system (RAS) contributesto tissue injury and fibrosis in patients with chronicliver diseases.32 Also, the gene expression of suppres-sor of T cell receptor signaling 1 (STS1; ubiquitinassociated and SH3 domain containing B) and CAP-GLY domain containing linker protein 4 (CLIP4) wasincreased in G3 versus G1.

Predicting Sample Class and Fibrosis

Progression

To determine whether a multigenetic classifier couldbe derived to predict the group class, we filtered thedata set to retain only the 16 samples from G1 (n ¼ 8)and G3 (n ¼ 8). The final model included the interceptand 9 bead types corresponding to 8 unique genes(Table 1). Pairwise scatter plots of these 9 bead typesrevealed that G1 and G3 were well separated for eachgene (Fig. 3). For the training data, the model had100% sensitivity and 100% specificity. The N-fold

cross-validated error rate was 31.3%, with 3 of the 8G1 cases and 2 of the 8 G3 cases misclassified.

Selection of Genes and Number of Genes for

Assay Validation and Study Validation

We used 2 strategies to select the genes for the micro-array assay validation with QPCR: (1) the genes wereat the top of the list of differentially expressed genes,and/or (2) the genes were among the top-scoring net-works and pathways (when we used interaction net-works and functional analysis for the differential com-parison analysis) and/or were selected as part of thebest LASSO model.

To determine the number of genes that should bevalidated with a more sensitive assay, we restrictedthe data set to G1 and G3 patients. The LASSO modelwas fitted for group predictions with the intensities ofthe 24,527 bead types as covariate samples and withan N-fold cross-validation procedure. For each of theN models in the cross-validation procedure, the modelwith the smallest AIC was selected as the best model.The median number of bead types (genes) in the Nmodels was 7 (range ¼ 2-10). For our study, 11 geneswere selected for further evaluation.

Validation Analysis Using QPCR

Therefore, 11 genes were evaluated with QPCR: STS1;glutathione S-transferase, C-terminal domain con-taining (GSTCD); IL-28RA; progesterone immunomo-dulatory binding factor 1 (PIBF1); methylenetetrahy-drofolate dehydrogenase 2 (MTHFD2); regulator ofcalcineurin 1 (RCAN1); coiled-coil domain containing99 (CCDC99); ATPase, aminophospholipid transporter,

Figure 2. (A) Venn diagram and (B) supervised cluster analysis. The data set was restricted to the 50 bead types that weresignificantly differentially expressed (P < 0.001) between G1 and G3 during HCVrec. Thereafter, agglomerative hierarchical clusteringwith Ward’s method was applied to the filtered data set with Pearson’s correlation as the similarity measure.

830 MAS ET AL. LIVER TRANSPLANTATION, July 2011

class I, type 8A, member 2 (ATP8A2); immunoglobulinsuperfamily, DCC subclass, member 4 (NOPE); semadomain 3A (SEMA3A); and ACE. For this specific ana-lysis, RNA from all the samples included in the DASLassay were studied [the training set, n ¼ 23 (21 HCVsamples and 2 normal livers), and the validation set,

n ¼ 26 (22 HCV samples and 4 normal livers)]. Themicroarray results were further validated in both thetraining and validation sets (STS1, r ¼ 0.61, P ¼0.001; GSTCD, r ¼ 0.56, P ¼ 0.03; IL-28RA, r ¼ 0.75,P ¼ 0.01; PIBF1, r ¼ 0.56, P ¼ 0.035; MTHFD2, r ¼0.67, P ¼ 0.002; RCAN1, r ¼ 0.78, P ¼ 0.001;

TABLE 3. Genes Differentially Expressed by G1 and G3 During HCVrec

Probe Identifier Symbol

Mean Expression

P ValueG1 G3

ILMN_1653856 STS1 6.45 9.50 0.00037ILMN_1658632 NLRP7 8.36 7.51 0.00008ILMN_1663370 CCDC62 8.86 8.25 0.00022ILMN_1670282 AIRE 8.82 7.96 0.00012ILMN_1672121 LOC387856 8.18 6.97 0.00030ILMN_1683824 TMEM16G 9.77 9.14 0.00006ILMN_1687404 HCN1 8.00 6.75 0.00021ILMN_1693896 ATP8A2 8.50 7.60 0.00007ILMN_1695357 CCDC99 5.98 7.75 0.00004ILMN_1700310 SDCBP2 7.20 9.78 0.00096ILMN_1712112 RCAN1 7.85 7.97 0.00010ILMN_1723058 CLIP1 6.12 7.38 0.00011ILMN_1724762 XKR6 8.33 7.46 0.00004ILMN_1724774 SYT3 8.30 7.29 0.00032ILMN_1738449 STIM2 8.81 8.73 0.00035ILMN_1738849 SLC9A2 8.26 7.17 0.00007ILMN_1739640 DCHS1 9.28 8.27 0.00024ILMN_1741422 FUT8 5.76 6.27 0.00015ILMN_1743570 CEACAM3 8.83 7.78 0.00003ILMN_1745880 DNAI1 9.36 8.34 0.00005ILMN_1747744 LHFPL2 6.21 7.69 0.00001ILMN_1755643 MGAT4A 7.27 9.41 0.00034ILMN_1755869 OR51T1 8.95 8.14 0.00029ILMN_1758679 TMEM168 7.45 10.79 0.00011ILMN_1759792 CLIP4 7.24 9.98 0.00055ILMN_1765641 SEMA3A 9.39 8.11 0.00017ILMN_1780630 LOC390110 8.45 7.54 0.00004ILMN_1782545 SLC2A4 8.44 7.42 0.00008ILMN_1782944 GPR37L1 6.49 8.66 0.00010ILMN_1783832 GAGE6 8.46 7.48 0.00009ILMN_1785768 PDE4A 8.45 7.42 0.00004ILMN_1799011 CEP250 8.71 7.78 0.00005ILMN_1805024 ERBB2IP 11.93 11.61 0.00020ILMN_1806791 OR4C46 8.68 7.71 0.00018ILMN_1809951 ZNF200 6.08 8.92 0.00031ILMN_1810424 HRH4 9.17 8.26 0.00024ILMN_1812701 C4orf33 9.87 10.54 0.00024ILMN_2041463 C3orf30 8.58 7.70 0.00017ILMN_2139396 NOPE 8.91 8.24 0.00016ILMN_2161508 PHTF2 6.46 8.91 0.00058ILMN_2224300 SNAPC3 8.57 7.34 0.00037ILMN_2227473 PIBF1 6.45 7.11 0.00006ILMN_2242068 GSTCD 6.78 9.12 0.00014ILMN_2301083 UBE2C 6.33 9.14 0.00080ILMN_2332440 KCNMB2 7.46 6.72 0.00025ILMN_2339410 ACE 6.28 7.13 0.00000ILMN_2372795 TTLL1 8.35 8.18 0.00033ILMN_2375141 C10orf4 9.24 8.48 0.00002ILMN_2405324 IL-28RA 7.25 8.79 0.00017ILMN_2405521 MTHFD2 6.88 10.21 0.00005

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 831

CCDC99, r ¼ 0.52, P ¼ 0.04; ATP8A2, r ¼ 0.79, P ¼0.001; NOPE, r ¼ 0.6, P ¼ 0.02; SEMA3A, r ¼ 0.69,P ¼ 0.023; and ACE, r ¼ 0.86, P ¼ 0.001).

DISCUSSION

Recent data suggest that the number of patients withHCV cirrhosis who are referred for OLT will continueto increase until 2020.33–35 Strategies for improvingthe management of patients with HCVrec include notonly more effective antiviral agents but also the identi-fication of robust predictors of severe disease in orderto prevent or minimize their impact on allograft func-tion. The pathogenesis of accelerated fibrogenesisin HCV-infected OLT recipients remains largelyunknown. Gene expression microarray analysis hasproven to be a remarkable tool for biologists and nowhas the potential to be widely adopted in clinicalresearch. Until recently, the major challenge was theneed for high-quality RNA for standard microarraysample labeling. However, the WGE DASL assay hasbeen shown to be a sensitive and reproducible profil-ing system that can be used with FFPE tissue sam-ples20-22,35,36 (including liver biopsy samples37). The

use of microarrays for gene expression in FFPEsamples opens an unlimited set of possibilities forincreasing our knowledge about HCVrec post-OLTand for identifying disease progression biomarkers.However, because the technique was designed fordegraded and partially degraded RNA, quality controlanalysis is required. Therefore, in this study, we eval-uated the RNA integrity and purity as well as differentparameters associated with the DASL assay, and weestablished that only those arrays with a present callvalue higher than 30% would be included in the anal-ysis. In this study, we performed WGE analyses of al-lograft biopsy samples obtained during clinicalHCVrec to evaluate distinctive gene expression pat-terns in the early postoperative period that may beassociated with progression to severe histological dis-ease. Patients were classified into groups according tothe severity of their fibrosis 36 months post-OLT. Thenumber of samples in each group was relatively small.However, this longitudinal study was performed in awell-selected group of patients with recurrent HCVinfection who underwent a large number of protocolbiopsy procedures in the absence of antiviral therapy.Moreover, the initial signatures obtained with the

Figure 3. Pairwise scatter plots for the 9 bead types included in the final LASSO model. G1 observations are plotted in red, and G3observations are plotted in blue.

832 MAS ET AL. LIVER TRANSPLANTATION, July 2011

DASL assay were validated with a second assay(QPCR) in an independent group of patients. The firststep of the analysis was the comparison of the geneexpression signatures in allograft biopsy samplesobtained during HCVrec and in normal liver tissue.The gene expression signature at the time of HCVrecwas characterized by activation of the innate immuneresponse, inflammation, and overexpression of inter-feron-inducible genes. Innate immune lymphocytesare believed to play important roles in the immediateresponse to viral infections through the production ofinterferon and/or the recognition of virus-infectedcells.38 Natural killer cells mediate the lysis of virus-infected cells via natural cytotoxicity and antibody-de-pendent cellular cytotoxicity and are controlled bypositive and negative cytolytic signals. The recruit-ment of nonspecific cells to the allograft may reduceviral replication by noncytolytic mechanisms or aug-ment viral replication and direct cytopathic injury.Moreover, the overexpression of interferon and inter-feron-inducible genes has been observed. The immuneresponse is likely insufficiently vigorous to keep viralreplication under control, but it may maintain chronicliver injury.39 An analysis at the time of HCVrecshowed different gene expression profiles in patientswho developed mild, moderate, or severe fibrosis 3years post-OLT (Fig. 2). As expected, larger differenceswere found between G1 and G3, with samples fromG2 being misclassified. These findings may haveresulted from the limited accuracy of histological eval-uations of core biopsy samples with respect to the di-agnosis of intermediate stages of fibrosis and/or fromthe overlapping of the molecular events involved inthe progression of fibrosis.

An important finding was the overexpression of ACEin allograft biopsy samples from G3 patients. Thisgene was also present in the best model using LASSO(2 bead types for the ACE gene; Fig. 3) and was vali-dated in the independent set of patients. A number ofstudies have shown that components of a localhepatic RAS are up-regulated in fibrotic liversof humans and in experimental animal models.32,40

Furthermore, the blockade of RAS by ACE inhibitorsand angiotensin type 1 receptor antagonists signifi-cantly attenuated liver fibrosis in experimental modelsof chronic liver injury.32 Moreover, because drugsblocking the action of angiotensin II have also beenassociated with the reduction of fibrosis in differenthuman and experimental models, Rimola et al.40

assessed the possible beneficial effects of these drugson graft fibrosis during HCVrec after OLT. Theauthors concluded that the administration of angio-tensin-blocking agents may be helpful in reducing thedevelopment of graft fibrosis in patients with HCVrecafter OLT. These findings open a completely new fieldin the treatment of HCVrec post-OLT.

Differential expression of genes associated with IL-28 signaling was observed between G1 and G3. IL-28B encodes a cytokine distantly related to type Iinterferons and the IL-10 family.41 The up-regulationof interferon-stimulated genes (ISGs) in the livers of

patients with chronic HCV before treatment might berelated to the poor induction of ISGs and the impairederadication of HCV during treatment.42 It has beenpostulated that the different expressions of hepaticISGs before treatment might be due to polymorphismsin IL-28B. In parallel to gene expression analysis, ge-nome-wide association studies have been used toidentify loci associated with the response to treat-ment; a genetic variation in IL-28B has been found topredict hepatitis C treatment–induced viral clear-ance.41-44 In a recent publication, Fukuhara et al.45

concluded that in patients with recurrent HCV infec-tions after OLT, combination analyses of single nucle-otide polymorphisms of IL-28B in recipient and donortissues and mutations in HCV RNA allow the predic-tion of a sustained virological response to pegylatedinterferon/ribavirin therapy. Moreover, Honda et al.46

showed in a recent publication that the IL-28Brs8099917-G allele is associated with higher expres-sion of hepatic ISGs as well as a poorer treatmentresponse, and these findings may provide a functionalexplanation for the association between the IL-28B ge-notype and the response to the interferon-a/ribavirintreatment of chronic hepatitis C.

As previously reported,47,48 IL-28B interacts with aheterodimeric class II cytokine receptor consisting ofIL-10RB and IL-28RA. This gene, IL-28A, and IL-29are 3 closely related cytokine genes that encode pro-teins known as type III interferons, are induced byviral infections, and have antiviral activity.41 It is pos-sible that IL-28B could mediate antiviral signalingthrough IL-10 signaling as well as signal transducerand activator of transcription 1 activation. In ourstudy, the IL-28 and IL-28RA genes were up-regulatedin G3 patients. The dominant expression signaling ofIL-28B by T helper 2 lymphocytes may modulatepathways in livers with HCVrec post-OLT and contri-bute to the development of fibrosis. This postulate issupported by published experimental evidence inhumans and primates showing that in general, Thelper 2–polarized responses promote fibrosis,whereas T helper 1 cytokines may be antifibrogenic.49

An analysis of distinctive gene expression profilesduring HCVrec provides a unique opportunity for iden-tifying biomarkers of fibrosis progression. In this series,in which all patients underwent protocol biopsy at fre-quent intervals post-OLT, no significant differenceswere observed between those with mild disease andthose with severe disease with respect to previouslyrecognized predictors of fibrosis progression and out-comes (eg, donor age, histology at the time of recur-rence, time to recurrence or acute rejection, and immu-nosuppression intensity).50-52 Our findings reinforcethe importance of identifying other predictors of fibrosisprogression such as gene expression early after OLT

Overall, this longitudinal study shows (1) the feasi-bility of performing gene expression analysis withFFPE liver samples and (2) the potential clinical use-fulness of a set of molecular markers for predictingearly after OLT the severity and progression of HCVrecduring long-term follow-up.

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 833

REFERENCES

1. Forman LM, Lewis JD, Berlin JA, Feldman HI, LuceyMR. The association between hepatitis C infection andsurvival after orthotopic liver transplantation. Gastroen-terology 2002;122:889-896.

2. Kim WR, Brown RS Jr, Terrault NA, El-Serag H. Burdenof liver diseases in the United States: summary of aworkshop. Hepatology 2002;36:227-242.

3. Davis GL, Albright JE, Cook SF, Rosenberg DM. Projec-ting future complications of chronic hepatitis C in theUnited States. Liver Transpl 2003;9:331-338.

4. Wiesner RH, Sorrell M, Villamil F. Report of the firstInternational Liver Transplantation Society expert panelconsensus conference on liver transplantation and hepa-titis C. Liver Transpl 2003;9:S1-S9.

5. Berenguer M, Prieto M, Rayon JM, Mora J, Pastor M,Ortiz V, et al. Natural history of clinically compensatedhepatitis C virus-related graft cirrhosis after liver trans-plantation. Hepatology 2000;32:852-858.

6. Futagawa Y, Terasaki PI, Waki K, Cai J, Gjertson DW. Noimprovement in long-term liver transplant graft survivalin the last decade: an analysis of the UNOS data. Am JTransplant 2006;6:1398-1406.

7. Yilmaz N, Shiffman ML, Stravitz RT, Sterling RK, LuketicVA, Sanyal AJ, et al. A prospective evaluation of fibrosisprogression in patients with recurrent hepatitis C virusfollowing liver transplantation. Liver Transpl 2007;13:975-983.

8. Berenguer M, Prieto M, San Juan F, Rayon JM, MartinezF, Carrasco D, et al. Contribution of donor age tothe recent decrease in patient survival among HCV-infected liver transplant recipients. Hepatology 2002;36:202-210.

9. Terrault NA, Shiffman ML, Lok AS, Saab S, Tong L,Brown RS Jr, et al. Outcomes in hepatitis C virus-infected recipients of living donor vs. deceased donorliver transplantation. Liver Transpl 2007;13:122-129.

10. Sanchez-Fueyo A, Restrepo JC, Quinto L, Bruguera M,Grande L, Sanchez-Tapias JM, et al. Impact of therecurrence of hepatitis C virus infection after liver trans-plantation on the long-term viability of the graft. Trans-plantation 2002;73:56-63.

11. Rayhill SC, Wu YM, Katz DA, Voigt MD, Labrecque DR,Kirby PA, et al. Older donor livers show early severe his-tological activity, fibrosis, and graft failure after livertransplantation for hepatitis C. Transplantation 2007;84:331-339.

12. Iacob S, Cicinnati VR, Hilgard P, Iacob RA, Gheorghe LS,Popescu I, et al. Predictors of graft and patient survivalin hepatitis C virus (HCV) recipients: model to predictHCV cirrhosis after liver transplantation. Transplanta-tion 2007;84:56-63.

13. Ciccorossi P, Maina AM, Oliveri F, Petruccelli S, LeandroG, Colombatto P, et al. Viral load 1 week after liver trans-plantation, donor age and rejections correlate with the out-come of recurrent hepatitis C. Liver Int 2007;27:612-619.

14. Berenguer M, Royuela A, Zamora J. Immunosuppressionwith calcineurin inhibitors with respect to the outcomeof HCV recurrence after liver transplantation: results of ameta-analysis. Liver Transpl 2007;13:21-29.

15. Martin P, Busuttil RW, Goldstein RM, Crippin JS, Klint-malm GB, Fitzsimmons WE, Uleman C. Impact of tacroli-mus vs. cyclosporine in hepatitis C virus-infected livertransplant recipients on recurrent hepatitis C: a prospec-tive randomized trial. Liver Transpl 2004;10:1258-1262.

16. Berenguer M, Aguilera V, San Juan F, Benlloch S, RubinA, Lopez-Andujar R, et al. Effect of calcineurin inhibitors inthe outcome of liver transplantation in hepatitis C virus-positive recipients. Transplantation 2010;90:1204-1209.

17. Ryder SD, Irving WL, Jones DA, Neal KR, UnderwoodJC, for Trent Hepatitis C Study Group. Progression ofhepatic fibrosis in patients with hepatitis C: a prospec-tive repeat liver biopsy study. Gut 2004;53:451-455.

18. Wright M, Goldin R, Fabre A, Lloyd J, Thomas H, TrepoC, et al., for HENCORE Collaboration. Measurement anddeterminants of the natural history of liver fibrosis inhepatitis C virus infection: a cross sectional and longitu-dinal study. Gut 2003;52:574-579.

19. Firpi RJ, Abdelmalek MF, Soldevila-Pico C, Cabrera R,Shuster JJ, Theriaque D, et al. One-year protocol liverbiopsy can stratify fibrosis progression in liver transplantrecipients with recurrent hepatitis C infection. LiverTranspl 2004;10:1240-1247.

20. Penland SK, Keku TO, Torrice C, He X, KrishnamurthyJ, Hoadley KA, et al. RNA expression analysis of forma-lin-fixed paraffin-embedded tumors. Lab Invest 2007;87:383-391.

21. Benchekroun M, DeGraw J, Gao J, Sun L, von Bogus-lawsky K, Leminen A. Impact of fixative on recovery ofmRNA from paraffin-embedded tissue. Diagn Mol Pathol2004;13:116-125.

22. Bibikova M, Chudin E, Arsanjani A, Zhou L, Garcia EW,Modder J, et al. Expression signatures that correlatedwith Gleason score and relapse in prostate cancer.Genomics 2007;89:666-672.

23. Bedossa P, Poynard T. An algorithm for the grading ofactivity in chronic hepatitis C. The METAVIR CooperativeStudy Group. Hepatology 1996;24:289-293.

24. Bibikova M, Talantov D, Chudin E, Yeakley JM, Chen J,Doucet D, et al. Quantitative gene expression profiling informalin-fixed, paraffin-embedded tissues using univer-sal bead arrays. Am J Pathol 2004;165:1799-1807.

25. Gui J, Li H. Penalized Cox regression analysis in thehigh-dimensional and low-sample size settings, withapplications to microarray gene expression data. Bioin-formatics 2005;21:3001-3008.

26. Ma S, Song X, Huang J. Supervised group Lasso withapplications to microarray data analysis. BMC Bioinfor-matics 2007;8:60.

27. Tibshirani R. The Lasso method for variable selection inthe Cox model. Stat Med 1997;16:385-395.

28. Archer KJ. Identifying important predictors using L1 penal-ized models and random forests. In: JSM Proceedings 2009.Alexandria, VA: American Statistical Association; 2009.

29. R Development Core Team. R: A Language and Environ-ment for Statistical Computing. Vienna, Austria: R Foun-dation for Statistical Computing; 2009.

30. Gentleman RC, Carey VJ, Bates DM, Bolstad B, DettlingM, Dudoit S, et al. Bioconductor: open software develop-ment for computational biology and bioinformatics.Genome Biol 2004;5:R80.

31. Park MY, Hastie T. L1-regularized path algorithm forgeneralized linear models. J R Stat Soc Series B StatMethodol 2007;69:659-677.

32. Warner FJ, Lubel JS, McCaughan GW, Angus PW. Liverfibrosis: a balance of ACEs? Clin Sci (Lond) 2007;113:109-118.

33. Roche B, Samuel D. Risk factors for hepatitis C recurrenceafter liver transplantation. J Viral Hepat 2007;14 (suppl 1):89-96.

34. Gane EJ. The natural history of recurrent hepatitis Cand what influences this. Liver Transpl 2008;14(suppl 2):S36-S44.

35. Ravo M, Mutarelli M, Ferraro L, Grober OM, Paris O,Tarallo R, et al. Quantitative expression profiling ofhighly degraded RNA from formalin-fixed, paraffin-em-bedded breast tumor biopsies by oligonucleotide micro-arrays. Lab Invest 2008;88:430-440.

834 MAS ET AL. LIVER TRANSPLANTATION, July 2011

36. Bibikova M, Yeakley JM, Wang-Rodriguez J, Fan JB.Quantitative expression profiling of RNA from formalin-fixed, paraffin-embedded tissues using randomlyassembled bead arrays. Methods Mol Biol 2008;439:159-177.

37. Hoshida Y, Villanueva A, Kobayashi M, Peix J, ChiangDY, Camargo A, et al. Gene expression in fixed tissuesand outcome in hepatocellular carcinoma. N Engl J Med2008;359:1995-2004.

38. Doherty DG, O’Farrelly C. Innate and adaptive lymphoidcells in the human liver. Immunol Rev 2000;174:5-20.

39. Rosen HR. Hepatitis C virus in the human liver trans-plantation model. Clin Liver Dis 2003;7:107-125.

40. Rimola A, Londono MC, Guevara G, Bruguera M, NavasaM, Forns X, et al. Beneficial effect of angiotensin-block-ing agents on graft fibrosis in hepatitis C recurrenceafter liver transplantation. Transplantation 2004;78:686-691.

41. Thio CL, Thomas DL. Interleukin-28b: a key piece of thehepatitis C virus recovery puzzle. Gastroenterology 2010;138:1240-1243.

42. Sarasin-Filipowicz M, Oakeley EJ, Duong FH, ChristenV, Terracciano L, Filipowicz W, Heim MH. Interferon sig-naling and treatment outcome in chronic hepatitis C.Proc Natl Acad Sci U S A 2008;105:7034-7039.

43. Thompson AJ, Muir AJ, Sulkowski MS, Ge D, Fellay J,Shianna KV, et al. Interleukin-28B polymorphismimproves viral kinetics and is the strongest pretreatmentpredictor of sustained virologic response in genotype 1hepatitis C virus. Gastroenterology 2010;139:120-129.

44. Rauch A, Kutalik Z, Descombes P, Cai T, Di Iulio J,Mueller T, et al. Genetic variation in IL28B is associatedwith chronic hepatitis C and treatment failure: agenome-wide association study. Gastroenterology 2010;138:1338-1345.

45. Fukuhara T, Taketomi A, Motomura T, Okano S, Nino-miya A, Abe T, et al. Variants in IL28B in liver recipientsand donors correlate with response to peg-interferon andribavirin therapy for recurrent hepatitis C. Gastroenter-ology 2010;139:1577-1585.

46. Honda M, Sakai A, Yamashita T, Nakamoto Y, MizukoshiE, Sakai Y, et al., for Hokuriku Liver Study Group.Hepatic ISG expression is associated with genetic varia-tion in interleukin 28B and the outcome of IFN therapyfor chronic hepatitis C. Gastroenterology 2010;139:499-509.

47. Sheppard P, Kindsvogel W, Xu W, Henderson K, Schluts-meyer S, Whitmore TE, et al. IL-28, IL-29, and theirclass II cytokine receptor IL-28R. Nat Immunol 2003;4:63-68.

48. Kotenko SV, Gallagher G, Baurin VV, Lewis-Antes A,Shen M, Shah NK, et al. IFN-k mediate antiviral protec-tion through a distinct class II cytokine receptor com-plex. Nat Immunol 2003;4:69-77.

49. Marra F, Aleffi S, Galastri S, Provenzano A. Mononuclearcells in liver fibrosis. Semin Immunopathol 2009;31:345-358.

50. Carrion JA, Torres F, Crespo G, Miquel R, Garcıa-Valde-casas JC, Navasa M, Forns X. Liver stiffness identifiestwo different patterns of fibrosis progression in patientswith hepatitis C virus recurrence after liver transplanta-tion. Hepatology 2010;51:23-34.

51. Carrion JA, Navasa M, Forns X. Retransplantation inpatients with hepatitis C recurrence after liver transplan-tation. J Hepatol 2010;53:962-970.

52. Blasco A, Forns X, Carrion JA, Garcıa-Pagan JC, Gila-bert R, Rimola A, et al. Hepatic venous pressure gradi-ent identifies patients at risk of severe hepatitis Crecurrence after liver transplantation. Hepatology 2006;43:492-499.

LIVER TRANSPLANTATION, Vol. 17, No. 7, 2011 MAS ET AL. 835