Journal of Hepatology 42 (2005) 860–869 www.elsevier.com/locate/jhep

Global gene repression in hepatocellular carcinoma and fetal liver, and suppression of dudulin-2 mRNA as a possible marker for the cirrhosis-to-tumor transition Ce´dric Coulouarn1, Ce´line Derambure1, Gre´gory Lefebvre1, Romain Daveau1, Martine Hiron1, Michel Scotte1,2, Arnaud Franc¸ois3, Maryvonne Daveau1, Jean-Philippe Salier1,* 1

Inserm Unite´ 519 and Institut Fe´de´ratif de Recherches Multidisciplinaires sur les Peptides, Faculte´ de Me´decine-Pharmacie, 22 Bvd Gambetta, 76183 Rouen cedex, France 2 Service de Chirurgie Ge´ne´rale et Digestive, Centre Hospitalier Universitaire, Rouen, France 3 De´partment de Pathologie, Centre Hospitalier Universitaire, Rouen, France

Background/Aims: Whether the transcriptional reprogramming process induced by hepatocellular carcinoma recapitulates that of the developing liver is at present unclear. Methods: With a complete coverage of the liver transcriptome by microarray using adult livers as controls, we searched for similarities and differences in mRNA abundances between hepatocellular carcinoma nodules and fetal livers taken before (early) or after (late) the 22–24th week of gestation. Changes in some mRNA levels were studied in further liver samples by quantitative RT-PCR. Results: Altered gene expression in hepatocellular carcinoma mostly results in down-regulated mRNAs which largely overlap with those repressed in the late fetal liver. Different frequencies of transcription factor binding sites in the down-regulated genes vs control genes as well as changes in abundance of mRNAs for relevant transcription factors point to a transcriptional repression. The down-regulated mRNAs code for proteins involved in (i) transcription and translation, (ii) specific functions of the differentiated hepatocyte or (iii) activation of proliferation and apoptosis. Conclusions: Apoptosis limitation is likely to predominate over active proliferation during liver development and hepatocellular carcinoma. Repression of the apoptosis-associated dudulin-2 mRNA points to a potential marker for the transition from a carcinoma-free to carcinoma-associated cirrhosis. q 2005 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Keywords: Cancer; Development; Dudulin-2; Fetal liver; Hepatocyte; Microarray; Protein function; Transcription factor; Transcriptome 1. Introduction Hepatocellular carcinoma (HCC) is a primary liver cancer, the main causes of which are hepatitis B or C virus infection, alcohol abuse or aflatoxin B1 intoxication. In most instances, HCC develops in the setting of chronic hepatitis or cirrhosis [1]. Numerous HCC-associated epigenomic alterations result in a dysregulated expression Received 29 November 2004; received in revised form 27 January 2005; accepted 28 January 2005; available online 11 April 2005 * Corresponding author. Tel.: C33 235 14 84 59; fax: C33 235 14 85 41. E-mail address: [email protected] (J.-P. Salier).

of genes and proteins [1]. Liver transcriptome analysis by microarray has resulted in the identification of hundreds of genes with an aberrant under- or over-expression in HCC as compared to the surrounding cirrhotic tissue [2–7]. However, when cumulated, many of these data appear to be blurred or sometimes contradictory, and an overall picture of altered gene regulation remains elusive [1,6,8]. Liver development entails the ordered activity of transcription factors which are mostly liver-enriched transcription factors (LETF) [9–11]. These LETFs orchestrate the up- or down-regulation of numerous target genes in the fetal hepatocyte, a cell which must simultaneously adapt to changes in body metabolism, escape apoptosis

0168-8278/$30.00 q 2005 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jhep.2005.01.027

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

and proliferate [9–11]. Among such target genes, the a-fetoprotein (AFP) gene is highly expressed in fetal as opposed to adult liver and its re-expression is frequently observed in HCC [12]. Other genes have been shown to exhibit a similar up-regulation in fetal liver (FL) and HCC as compared to adult liver [13–16]. In fact, it has been suspected that the transcriptional reprogramming induced in HCC could mimic that of the developing liver [16,17]. Yet, genome-wide studies during liver development have focused on a relatively small number of regulated genes [18–21] and the few transcriptome studies that have compared FL vs HCC provided limited information on a possible gene overlap [16,17]. Hence, one still lacks complete data that would point to the global similarities and differences of gene expression in FL vs HCC. With a microarray covering every gene expressed in fetal, adult or tumorous liver [22], we have now identified a general trend to gene under-expression in FL as well as HCC, as compared to normal adult liver. This appears to be controlled, at least partly, by a limited number of [LETF/ binding site] combinations. The mRNAs repressed in FL and HCC are prominantly involved in transcription/ translation, cell proliferation/apoptosis, or differentiated hepatocyte metabolism, and one of them provides a prognostic marker for the cirrhosis-to-HCC transition.

861

that obtained from other, cumulated analyses [6]. The Genesis tool was used for data analysis by clustering [24]. Data from the Gene Ontology Consortium [25] were used to retrieve information on protein functions. Quantitative Reverse Transcription PCR (q-RT-PCR) of mRNAs was done as described [22] and the primers are listed as a supplementary Table S1 on the journal web site (doi:10.1016/j.jhep.2005.01.027). Determination of dudulin-2 mRNA level was done with an Assays-on-Demand kit (ref 4331182) and a Taqman 7700 equipment from ABI. Every mRNA level were normalized with the 18S mRNA level.

2.3. Computerized search for LETF binding sites and comparison of occurrence in classes Gene promoter sequences were retrieved with the EZ-Retrieve program [26]. Control promoters from genes that appeared not to be regulated in this study were chosen on the sole basis of promoter sequence availability. For every gene, the first 5 kb of sequence upstream of the transcription start site were used for a search of potential binding sites for any transcription factor with the TRANSFAC library [27] and the vertebrate option of the TFSEARCH program (http://www.cbrc.jp/research/db/TFSEARCH.html). When a potential binding site was thus identified, this site was retained for further analysis provided it was present in at least 30% of the promoters of at least one of the two classes of regulated or control genes. Differences in occurrence of this binding site between classes were evaluated with a nonparametric Wilcoxon’s test. For the binding sites listed in Section 3, the threshold for a similarity between a consensus and an actual sequence within a promoter was set at 80% (CDP, CRE, E47, GATA, HSF1, Ik-2, XBP-1) or 90% (AP-1, C/EBP, HSF2, NF-kB).

3. Results 2. Patients and methods

3.1. Comparison of mRNA levels in FL, adult liver and HCC by microarray

2.1. Human subjects and RNA sources Liver fragments were obtained under strict anonymity from the digestive surgery unit of Charles Nicolle Hospital (Rouen, France). The clinical data are provided in Table 1. A fragment of a cancerous nodule as well as distant cirrhotic tissue were taken whenever an HCC resection was performed. HCC-free cirrhotic liver was obtained by surgical biopsy for histological diagnosis in patients who were operated on for various, nonliver-related pathologies but presented cirrhosis suspicion at surgery. Control human liver was obtained mostly in patients operated on for benign liver tumor or liver metastasis of a non-hepatic cancer, in which cases the tissue was taken away from the tumor. According to the current French rules and ethical guidelines, neither an informed consent nor an advice from an ethical committee are requested prior to analysis of RNA in resected tissues that would otherwise be disposed off. Histopathology was carried out by a trained pathologist. Tissue storage, culture of hepatoma cell lines and RNA extraction were done as described [22]. Following crude liver cell dissociation by collagenase treatment, hepatocytes and non-parenchymal cells were separated by centrifugation, as described [23]. Two sets of early or late FL RNAs (a pool of several livers covering the 15–24th week of gestation or another pool of 38 livers covering the 22–40th week of gestation, respectively) were purchased from Clontech.

2.2. Transcriptome analysis Our set of human cDNA probes dubbed Liverpool that is tailored to a complete coverage of the human liver transcriptome (ca. 10,000 genes), the associated LiverTools database, as well as the procedures from array preparation to final data handling have all been detailed [22]. In this study, every RNA sample was subjected to three rounds of hybridization and the resulting, normalized values were used for a selection of regulated mRNAs, i.e. whose abundance varies significantly (P!0.05) between samples [22]. Such triplicates result in a false discovery rate that is below 10% of the total number of regulated mRNAs (not detailed). This figure is consistent with

We analyzed liver samples with a microarray that provides a complete coverage of the liver transcriptome [22]. The developmental stage was taken into account with two samples of FL RNAs pooled from either 15–24-weekold (early FL) or 22–40-week-old fetuses (late FL). We also used RNAs from nine HCCs and four control adult livers (HCC1–HCC9 and C1–C4 in Table 1). We selected every mRNA whose level was significantly different in early FL and/or late FL and/or at least three HCC samples, as compared to the mean level found in controls. This resulted in a selection of 1436 mRNAs, as summarized in Fig. 1A. The most immediate observation is the very high fraction of down-regulated mRNAs found in HCCs (83.9% of 704 mRNAs). In fact, this feature was obtained whatever the number of HCC samples chosen for the cut-off noted above (data not shown). Moreover, a predominating downregulation of mRNAs levels was also observed in early FL (63.0% of 466 mRNAs) and late FL (60.0% of 647 mRNAs). We then used a principal component analysis (PCA) whereby the samples were gathered on the basis of similarity in mRNA abundance pattern. The PCA presented in Fig. 2 immediately pointed to a shared location of the four control livers within a narrow three-dimensional space. Moreover, the early FL was located apart from all other samples whereas the late FL and the nine HCCs shared the same broad area (of note, no particular location or clustering

862

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

Table 1 Clinical data in patients with HCC or HCC-free cirrhosis, and control patients Patient

Sex

Age

Pathology

Etiology

Tumor gradea

Number of nodules

Vascular invasion

HCC1 HCC2 HCC3 HCC4 HCC5 HCC6 HCC7 HCC8 HCC9

M M M M M M M M M

60 63 69 49 73 72 61 45 49

HCC HCC HCC HCC HCC HCC HCC HCC HCC

4 2 3 1 2 3 3 2 1

O6 1 1 1 1 1 1 1 3

Yes No No No No Yes No No No

HCC10 HCC11 HCC12 HCC13 HCC14 HCC15 HCC16 HCC17 HCC18 HCC19 HCC20 HCC21 HCC22 HCC23 HCC24 HCC25 HCC26 HCC27 CIR1 CIR2 CIR3 CIR4 CIR5 CIR6 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13

M M M M F M F M M M M M M M M M M F M F F F M F M M M F F M F F F F F F M

70 65 71 65 73 68 73 66 52 72 68 78 56 68 70 82 49 42 65 55 50 50 45 48 62 76 74 94 71 72 68 42 62 53 30 45 57

HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC HCC CIR CIR CIR CIR CIR CIR Meta. (colon)b Meta. (colon) Meta. (colon) Meta. (breast) Hemangioma Meta. (colon) Meta. (breast) HCCc CHGc Hemangioma FNHc Adenoma Hemangioma

Hemochromat HCV HCV ALC ALC ALC HCVCALC HCVCALC HBVCHCVC ALC HCV HCV HCV HCV HCV ALC ALC ALC ALC ALC ALC ALC ALC ALC ALC HBV HBVCALC None of above HCV HCV ALC ALC ALC HCV

1 2 2 3 3 1 2 2 2 2 2 3 3 3 3 2 3 2

1 1 2 2 1 2 1 1 3 3 7 1 1 1 1 1 1 1

No No No No Yes Yes No No Yes Yes Yes No Yes No No Yes Yes No

HCC, patient with hepatocellular carcinoma; CIR, patient with an HCC-free cirrhosis; C, control patient; HBV, hepatitis B virus infection; HCV, hepatitis C virus infection; ALC, alcohol abuse. a Differentiation grade [46]. b Whenever a control liver was resected for a hepatic metastasis in the follow-up of a non-hepatic cancer, the tumor origin is noted in brackets. c Histologically normal liver taken away from (i) an HCC nodule of unknown etiology developed in a non-cirrhotic liver (HCC), or (ii) a cholangiocarcinoma (CHG), or (iii) a focal nodular hyperplasia (FNH).

of any HCC sample related to etiology, grade, tumor size or vascular invasion could be found here). This suggested that the similarity of gene expression between FL and HCC was closest with late FL. We next focused on the mRNAs whose abundance was found to be different in FL and HCC samples vs control

adult livers. The data are presented in Fig. 1B. Remarkably, the very limited fraction of mRNAs with an opposite regulation in late FL vs HCCs (4.5% of 332 mRNAs) contrasted with a relatively high fraction of opposite regulations in early FL vs HCCs (24.6% of 126 mRNAs). This again indicated a global similarity of gene

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

mRNAs regulated in : late FL n=647

early FL n=466

B

HCC n=704

mRNAs regulated in : HCC & early FL

HCC & early FL & late FL n=33

n=93

HCC & late FL n=299 332

126

up-regulation

opposite regulation

down-regulation

Fig. 1. Similarities and differences in mRNA levels regulated in early, late FL and/or HCCs vs controls. The diagrams depict the numbers of mRNAs whose levels were found to be up- or down-regulated in a given condition vs controls (A) or in a similar or opposite direction between conditions (B). A given mRNA was included whenever it was found to be regulated in at least the early or late FL or at least three out of nine HCC samples as compared to its average value in four control adult livers, as determined by microarray. In A, the total number of mRNAs is only 1436, for redundancy between subsets.

down-regulation in late FL and HCC. A bi-dimensional hierarchical clustering based onto the subset of mRNAs included in Fig. 1B is presented in Fig. 3. Again, the late FL clustered with a subset of three HCCs, whereas the early FL was apart from all other samples, alike what was observed above by PCA with a much higher number of mRNAs. The data in Fig. 3 further demonstrated that all nine HCCs and late FL shared a strong down-regulation of numerous mRNA levels (Section 1, 65.6% of all mRNAs and Section 3, 24.6%) and an up-regulation of a very limited number of other mRNA levels (Section 2, 9.8% of all mRNAs), as compared to the control livers. In contrast, the early FL exhibited quite a limited number of downregulated mRNAs (Section 3). Overall, these data again demonstrated that (i) altered gene expression in HCC mostly results in a decreased abundance of the corresponding mRNAs and (ii) this under-expression is also found in the late FL, i.e. after the 22–24th week of gestation.

early FL C4 C3 C2 C1 HCC2 HCC3 late FL HCC1 HCC5 HCC7 HCC4 HCC9 HCC8 HCC6

A

863

I y

z II 6 x

8

4

4

9

1 late

Control

1 3 2

III

2 3 5 7

Fold induction

early

Fetal liver

HCC

Fig. 2. Clustering of liver sources by PCA. The 1436 mRNA levels found to be significantly up- or down-regulated in at least early or late FL (open square) or three HCCs (closed triangle 1–9) as compared to their mean level in control livers (open dot 1–4), i.e. the 1436 mRNAs in Fig. 1A, were used for PCA. The axes depict the first three variance components of the mRNAs included in the analysis, which together represent 84.9% of the total variance.

x3

1:1

x3

Fold repression

Fig. 3. Clustering of samples or mRNAs by bi-dimensional hierarchical clustering. The 425 mRNAs found to be significantly regulated in at least early FL or late FL, as well as in at least three HCC samples as compared to their mean level in controls, i.e. the 425 mRNAs in Fig. 1B, were used for bi-dimensional hierarchical clustering (average linkage option). The samples are clustered horizontally and the mRNAs are clustered vertically (the gene symbols are omitted). In every sample a change (fold) in a given mRNA level relative to its mean level in the four control samples is shown as a colored bar of variable intensity (scale at the bottom). Subsets I or II contain mRNAs that are, respectively, down- or up-regulated in late FL and HCCs. Subset III contains mRNAs that are down-regulated in early FL, late FL and HCCs.

864

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

3.2. LETF binding sites in the genes for down-regulated mRNAs

Table 2 Relative abundance of a given LETF-encoding mRNA in FLs or HCCs vs controls

We next investigated whether given LETF binding sites in the genes for some of the mRNAs above could account for the similar down-regulation seen in late FL and HCC. From the subset of mRNAs down-regulated in late FL and HCCs (i.e. most of the 332 mRNAs in Fig. 1B), we first selected every gene for which at least 5 kb of promoter sequence were available, which resulted in 21 such genes, here below designated the down-regulated genes. As a control, we screened a set of genes with an expression at least in liver [22] and without any particular regulation of the corresponding mRNA in this study (i.e. not included in Fig. 1). We randomly retained 48 such genes for which a promoter sequence was available. A computerized search for any potential binding site disclosed a significant difference (P!0.05) in the number of binding sites for 11 distinct LETFs between the two classes of 21 downregulated vs 48 control genes. This included an increased frequency of binding sites for four LETFs (AP-1, C/EBP, NF-kB, XBP) along with a decreased frequency of binding sites for seven other LETFs (CDP, CRE, E47, GATA, HSF1, HSF2, Ik-2) in the down-regulated genes. These data indicate that the global mRNA down-regulation seen in late FL and HCC is, at least partly, accounted for by an increased or decreased probability of LETF binding to the corresponding gene promoters. This mechanism in turn implies a change in abundance of the cognate LETFs during the events under study. Therefore, we also examined whether the abundances of the corresponding LETF-encoding mRNAs were modified in our FL and/or HCC samples vs adult livers. Although many differences in abundance of such mRNAs between these groups did not reach statistical significance, the corresponding hierarchical clustering, shown as a supplementary Fig. S1, pointed to changes in mRNA abundances, mostly for some members of the AP-1, ATF, C/EBP, GATA and XBP families. Again, a similarity of the late FL and HCCs was observed whereas the early FL was located apart. Moreover, Table 2 presents significant changes in the abundance of given LETF mRNAs. When a significant change was observed in either HCC or late FL, a similar trend was concomitantly found in the other condition. Remarkably, we noted opposite regulations for related LETF mRNAs, as examplified with (i) C/EBPa and -g at the C/EBP sites, and (ii) c-Jun, JunB, ATF-3 and ATF-7 at the AP-1 sites. This dynamic is consistent with a change of activity of the target promoters [28,29]. Collectively, our findings indicate that the condition-associated abundance of given LETFs along with the specific frequency of their binding sites onto some liver-expressed genes can explain, at least partly, the similar down-regulation of some hepatic mRNAs observed in late FL and HCC.

LETF binding site a

AP-1 (C)

C/EBP (C)

NF-kB (C) XBP (C) CRE (K)

E47 (K) GATA (K) Ik-2 (K)

LETF

LETF mRNA Early FL/C

c-Jun JunB JunD c-Fos C/EBPa C/EBPb C/EBPg C/EBPd p105 XBP-1 ATF-1 ATF-2 ATF-3 ATF-4 ATF-5 ATF-6 ATF-7 E47 GATA-4 GATA-6 Ikaros

b

0.54 0.70 0.97 0.89 1.00 0.82 0.87 0.72 0.90 0.98 1.04 0.85 1.12 1.38 0.80 0.69 2.530.70 1.13 0.65 1.42

Late FL/C

HCC/C

0.47 2.00 1.29 0.84 0.89 1.28 2.05 0.95 1.45 2.15 0.89 0.85 2.451.18 0.88 1.29 0.75 0.410.83 1.82 0.41-

0.301.461.19 0.86 0.641.13 1.421.02 1.24 1.660.77 1.03 1.651.19 1.03 0.73 0.290.84 1.13 1.24 0.36-

a A sequence with similarity to this binding site exhibits a significantly high (C) or low frequency (K) in the promoters of genes whose mRNAs are down-regulated in both late FL and HCC vs control promoters (see Section 3). Accession numbers for binding sites in the TRANSFAC data library: AP-1, M00173; C/EBP, M00116; CRE, M00040; E47, M00071; GATA, M00075; Ik-2, M00087; NF-kB, M00052; XBP, M00251. b Value presented as a ratio: normalized abundance of mRNA in the indicated sample/average of normalized abundances in four control adult livers (C). In the HCC/C column, the mean ratio for nine samples is given. A star indicates a significant difference (P!0.05) between FL vs controls (estimated as in 22) or between HCCs vs controls (one-way analysis of variance).

3.3. Functionally defined subsets of down-regulated mRNAs found in late FL and HCC We searched whether functionally defined mRNAs that were similarly down-regulated in late FL and HCC could help identify some cellular events that are shared in these two conditions. Within the subset of 332 regulated mRNAs in Fig. 1B, 158 mRNAs corresponded to proteins with defined functions, as exemplified in Table 3 and detailed in our supplementary Table S2 on line. Remarkably, we identified three prominent functional subsets in which most or all mRNAs were down-regulated. They cover transcription and translation, or cell proliferation and apoptosis, or they are hallmarks of the differentiated hepatocyte, as follows. Actors of the transcriptional machinery included, for instance: (i) member 2 of SWI/SNF subfamily c, a modifier of chromatin structure, (ii) Fos-like antigen-1 and -2 that participate in AP-1 formation and the MKP-1-like protein tyrosine phosphatase that controls AP-1 activation, (iii) the LETF STAT-3 and its inhibitor

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

865

Table 3 Functional classification of mRNAs with a regulated abundance in HCCs and late FL vs control liversa

Transcription/translation Transcription (up-regulatedb) FOS-like antigen-1 MKP-1 like protein Tyr phosphatase Transcription (down-regulatedb) FOS-like antigen 2 SWI/SNF subfamily c, member 2 Cell death-regulatory protein GRIM19 STAT3 Translation (down-regulatedb) Ribosomal protein L29 Ribosomal protein S3A Ribosomal protein S18 Eukaryotic transl. elongation factor 1a1 Ribosomal protein L13a Deoxyhypusine synthase DEAD box polypeptide 6 Ribosomal protein S26 Cell proliferation/apoptosis Cell proliferation (down-regulatedb) Interleukin 6 receptor Peroxiredoxin 1 DEAD box polypeptide 6 Apoptosis (down-regulatedb) Cullin 4A Death associated transcription factor 1 Caspase 4 Bruton agammaglobulinemia tyr kinase Cell death-regulatory protein GRIM19 Dudulin 2 Functions of differenciated hepatocyte Detoxication (down-regulatedb) Cytochrome P450, IIA, polyp. 6 Cytochrome P450, IVF, polyp. 12 Cytochrome P450, IIIA, polyp. 7 Formyltetrahydrofolate dehydrogenase Cytochrome P450, IIE, polyp. 1 Cytochrome P450, IIIA, polyp. 4 Metabolism (down-regulatedb) Glycogen synthase 2 (liver) Glucokinase regulatory protein Phosphoenolpyruvate carboxykinase 1 Aldo-keto reductase family 1B1 Plasma protein (up-regulatedb) Fibrinogen, a polypeptide Plasma protein (down-regulatedb) Fibrinogen, b polypeptide

IMAGE

Late FL

HCC

110,503c 309,800

2.59d 2.41

2.49(3)e 2.37(3)

309,748 111,704 366,332 308,551

0.29 0.42 0.26 0.30

0.60(6) 0.57(4) 0.42(5) 0.36(5)

198,542 757,511 112,363 308,473 130,029 83,125 198,622 310,610

0.30 0.22 0.38 0.27 0.44 0.22 0.29 0.31

0.62(4) 0.53(5) 0.49(5) 0.43(5) 0.42(6) 0.41(5) 0.39(5) 0.34(6)

120,306 112,471 198,622

0.38 0.28 0.29

0.54(5) 0.44(5) 0.39(5)

310,431 113,561 126,322 246,748 366,332 321,275

0.48 0.38 0.39 0.41 0.26 0.25

0.64(3) 0.57(4) 0.54(4) 0.53(3) 0.42(5) 0.15(9)

77,451 127,203 121,305 128,680 77,826 83,240

0.39 0.31 0.41 0.34 0.23 0.32

0.65(3) 0.55(4) 0.55(4) 0.47(5) 0.40(6) 0.36(7)

113,358 126,617 742,082 28,882

0.36 0.45 0.32 0.27

0.74(6) 0.62(4) 0.55(5) 0.45(5)

429,555

0.40

2.05(6)

112,334

0.32

0.49(5)

a

The complete data are provided as a supplementary Table S3 on line. In most instances, a similar regulation was found in late FL and HCC. c IMAGE clone number used as a unique identifier for every mRNA. d Ratio of the abundance in late FL vs the mean abundance in four control livers. e Ratio of the mean abundance in nine HCC samples vs the mean abundance in four control livers (number of HCC samples with a significantly different abundance vs controls). b

GRIM-19. Down-regulated mRNAs for proteins of the translational machinery were for instance: (i) deoxyhypusine synthase that controls the formation of the translation initiation factor-5A, (ii) the ribosomal proteins L13a, L29, S3A, S18 and S26, and (iii) the a1

subunit of translation elongation factor-1 which delivers aminoacyl tRNAs to ribosomes. Down-regulated mRNAs for proteins involved in cell proliferation were, for instance, the DEAD box polypeptide6 believed to be involved in embryogenesis and cell growth

C. Coulouarn et al. / Journal of Hepatology 42 (2005) 860–869

and division, peroxiredoxin-1 whose proliferative effect may be of relevance in cancer, and the IL-6 receptor that mediates the IL-6-induced cell proliferation. Strikingly, the mRNAs for apoptosis activators such as cullin 4A, the death-associated transcription factor 1, caspase 4, the Bruton’s agammaglobulinemia tyrosine kinase, GRIM19 and dudulin-2 were all down-regulated. Taken together, these data suggest that limitation of apoptosis predominates over active proliferation during liver development and HCC. mRNAs for proteins that are typical of detoxication, metabolism and plasma protein production in the differentiated hepatocyte were down-regulated. Detoxication proteins included various cytochromes P450, and formyltetrahydrofolate dehydrogenase whose down-regulation is proposed to enhance tumor cell proliferation. Metabolic enzymes were, for instance, those participating in glucose/glycogen metabolism, such as glycogen synthase2, the glucokinase-regulatory protein, phosphoenolpyruvate carboxykinase-1, and aldo–keto reductase 1B1. Among the plasma proteins, an opposite regulation of the mRNAs for fibrinogen-a and -b chains was observed in HCC, which is possibly related to the importance of some fibrinogen peptides in tumor cell angiogenesis [30]. All other plasma protein mRNAs were down-regulated. 3.4. Microarray data confirmed by q-RT-PCR As a control, we selected the six mRNAs with the most marked and frequent difference in abundance between our HCCs vs control samples above. The AFP mRNA was also studied, given the use of blood AFP in the clinical follow-up of HCC [31]. The relative abundance of these mRNAs was next determined by q-RT-PCR in a novel set of HCCs and controls (HCC10–HCC28 and C5–C13 in Table 1) as well as in our early or late FL. The results shown as a supplementary Fig. S2 perfectly fit our analysis made by microarray, including the strong or borderline up-regulation of the KIAA0789 or AFP mRNA, respectively, as well as the down-regulation of mRNAs coding for plasma proteins (haptoglobin, apolipoprotein C3, orosomucoid, albumin) or other proteins (dudulin-2). 3.5. Dudulin-2 mRNA as a marker of the cirrhosis-to-HCC transition As the abundance of dudulin-2 mRNA was strikingly different in HCCs vs controls, it was further measured in cirrhotic tissues and found to be down-regulated not only in the HCC nodules but also in the paired cirrhotic samples. This feature was found regardless of the tumor differentiation grade (Fig. 4), etiology or vascular invasion (data not shown). On the contrary, the mean dudulin-2 mRNA level in livers from HCC-free cirrhotic patients was slightly above the mean level in controls and no overlap between the levels in HCC-free vs HCC-associated cirrhotic livers was found.

*

100

$

*

$

$

$

75 mRNA abundance (%)

866

50

ns

p