Re-Expression of Transcription Factor ATF5 in Hepatocellular

Transcription factors represent an important class of genes that play key roles in controlling cellular proliferation, cell cycle modulation, and attractive targets for cancer therapy. Here, we report on the novel finding of common ATF5 down-regulations in hepatocellular carcinoma (HCC), a highly malignant tumor with a dismal clinical course. Array-based mapping in HCC highlighted a high and consistent incidence of transcription factor ATF5 repressions on regional chr.19q13. By quantitative reverse transcription-PCR, profound down-regulations of ATF5 were further suggested in 78% of HCC tumors (60 of 77 cases) compared to their adjacent nontumoral liver (P = 0.0004). Restoration of ATF5 expression in 3 nonexpressing HCC cell lines demonstrated a consistent growth inhibitory effect (P < 0.029) but minimal induction on cellular apoptosis. Subsequent flow cytometric investigations revealed a G(2)-M cell cycle arrest in HCC cells that were ectopically transfected with ATF5 (P < 0.002). The differential expressed genes from the functional effects of ATF5 were examined by array profiling. Over a hundred genes were identified, among which ID1 contains the ATF/CREB target binding sequences within its promoter region. An inverse relationship between ATF5 expressions with ID1 transcriptions was verified in HCC (P = 0.019), and a direct interaction of ATF5 on the promoter of ID1 was further demonstrated from electromobility shift assay. Examination of causal events underlying the silencing of ATF5 in HCC suggested copy number losses, promoter hypermethylation, histone deacetylation, and DNA mutations to be the likely inactivating mechanisms. In conclusion, our finding supports a tumor suppressive role for ATF5 in HCC, and highlighted ID1 as a potential downstream target.


Introduction
Hepatocellular carcinoma (HCC) is the 5th most common cancer worldwide and the 3rd most cause of cancer mortality (1).The overall dismal outcome of patients diagnosed with HCC is largely attributed to the tumor not being diagnosed in time for curative surgery.By the time of clinical presentation, most patients are at advanced inoperable stages, often with intrahepatic and extrahepatic metastases, which render surgical resection applicable to only a minority of patients (2).In addition, the high incidence of tumor recurrences, possibly from micrometastasis of tumor cells prior to curative surgery, further reduces patients' 5-year survival (3).Although the causative links of etiologic factors, including chronic hepatitis infections, heavy alcohol intake, dietary aflatoxin, and the male gender, in the development of HCC have been firmly established based on epidemiologic grounds (1), the molecular pathways by which HCC develops and progresses remain largely elusive.The recognition of reliable predictive biomarkers and a better understanding of biological pathway(s) by which HCC arises and progresses would undoubtedly provide directions for more efficacious therapies and improvements on overall patient prognosis.
Studies on the molecular pathogenesis of human cancers including breast, colon, prostate, and lung have guided the development of gene-based biomarkers and molecular-targeted therapies (4).In HCC, despite molecular investigations have indicated recurrent sites of genomic aberrations including regional amplicons, loss of heterozygosity, and cytogenetic translocations in the hepatic transformation from putative premalignant lesion of liver cirrhosis (5)(6)(7), few tumor suppressors and oncogenes have been defined within these causative loci.In this regard, recent advances in array-based transcriptional mapping offers the feasibility to correlate genomic aberrations with gene expression data.Such technological approach has allowed the unprecedented identification of candidate genes within casual sites of colorectal and invasive ductal breast carcinomas (8,9).
By Spectral Karyotyping, we have previously reported on the novel presence of frequent nonreciprocal rearrangements of chr.19 in HCC, where abnormalities on 19q could be identified in as much as f50% of primary tumors (7).Common chr.19q aberrations have also been described in a variety of human malignancies, including oligodendroglioma (10), astrocytoma (11), malignant gliomas (12), neuroblastoma (13), and ovarian carcinoma (14), where adverse prognostic associations have been further suggested in astrocytoma (11) and neuroblastoma (13).In prostate cancer, genome-wide linkage analysis has further suggested a susceptibility locus for a more aggressive tumor phenotype on chr.19q (12).Microsatellite analysis has defined the common deleted region to chr.19q13, where the presence of putative tumor suppressor gene(s) has long been speculated (11)(12)(13).Here, we describe an investigative study on the transcriptional activities of candidate genes on 19q in HCC.Using a high-resolution cDNA array, positional mapping indicated a high and consistent incidence of Activating Transcription Factor 5 (ATF5; located on chr.19q13.3)down-regulations in HCC tumors and cell lines.The ATF5 gene is a member of the ATF/CREB [cyclic AMP (cAMP)-dependent activating transcriptional factor/cAMP response element binding protein] family of transcription factors.The mammalian ATF/CREB family represents large groups of basic leucine zipper (bZIP) transcription factors with diverse physiologic functions that range from metabolite homeostasis to cellular differentiations and regulation of the cell cycle (13).Despite that the ATF/CREB family emerges as important cell survival factors, few studies have performed to characterize the ATF5 gene in cancer.In this study, the functional influence of ATF5 on HCC cell growth and its potential mechanistic actions were investigated.Our study highlighted a tumor suppressive role for ATF5 in HCC and suggested ID1, an oncogenic transcription factor (15), as one of its transcriptionally regulated target.

Materials and Methods
Patients and cell lines.Tumorous liver tissues were collected from 87 patients who underwent curative surgery for HCC at Prince of Wales Hospital, Hong Kong.The corresponding adjacent nontumoral liver tissue was available for 77 patients.Informed consent was obtained from each patient recruited, and the study protocol was approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong.Among these 87 cases, patients' age ranged from 30 to 78 years (median age, 55 years) and a male to female ratio of 4.44 (71 males and 16 females).Patients were predominantly chronic HBV carriers (93%) with identifiable cirrhosis in the nontumorous liver indicated in 65 cases (75%) and histologic signs of chronic hepatitis suggested in 22 patients.The disease stage of tumors collected was classified according to the American Joint Committee on Cancer staging criteria (16), which graded 4 cases as stage I, 49 cases as stage II, 27 as stage III, and 7 as stage IV.
Transcriptional mapping on Chr.19q.The expression array experiments were carried out as previously described (20).Briefly, differentially labeled Cy3 or Cy5 cDNA from cell line or primary HCC and normal liver pool were hybridized onto Human cDNA Arrays (Ontario Cancer Institute) that contained clones on chr.19q at a resolution of f100 Kb.Hybridized images acquired were analyzed by GenePix Pro3.0 (Axon Instruments).Duplicate array hybridizations were performed for each tested sample.Custom Normalization Suite4 was employed to integrate normalized signal ratios with physical map location of each cDNA clone.
qRT-PCR.Total RNA from cell lines, primary HCC, and adjacent nontumoral liver was subjected to first-strand cDNA synthesis using random hexanucleotide primer.Amplification of ATF5 was performed using SYBR Green PCR (Applied Biosystems) with ACTB as internal reference gene.Primers for ATF5 were 5 ¶-GAGTGGCGACAGGATAGAGC-3 ¶ (sense) and 5 ¶-CCTCCCTCCCCTTAGCGTAGTGT-3 ¶ (antisense), and the primers for ACTB were 5 ¶-CGCGAGAAGATGACCCA GAT-3 ¶ (sense) and 5 ¶-GTACGGC-CAGAGGCGTACAG-3 ¶ (antisense).Emission intensity from PCR assays was detected by the iCycler (BioRad Laboratories).Threshold cycles were averaged and expressed as a ratio relative to the mean value determined from seven normal human livers.Three adult human normal liver mRNAs were acquired from commercial companies (Ambion, Strategene, and CloneTech), and the remaining four normal livers were obtained from the colorectal cancer liver metastases, of which the liver at the surgical margin was histologically confirmed to be absent of underlying hepatic disease or cirrhosis.A ratio of <0.5 was considered down-regulation.
Colony formation assay.pcDNA3.1-ATF5and pcDNA3.1-vector-transfectedHep3B, HKCI-9, and HKCI-C1 cells were seeded at 2 to 4 Â 10 4 per well in 6-well culture plate.After 2 to 4 wk of incubation in selective medium containing 1.5 mg/mL G418, colonies were fixed in 70% ethanol and stained by crystal violet.Visible colonies containing >50 cells were scored, and results from triplicate assays were expressed as a mean from 3 independent experiments.
Flow cytometry.HKCI-9 and Hep3B cell lines were stably transfected with either pcDNA3.1-ATF5or pcDNA3.1-vector.Approximately 1 Â 106 cells were cultured for 18 h in 6-well culture plate.Both floating and attached cells were collected, washed with PBS, and fixed in 70% cold ethanol for 1 h at À20jC.Fixed cells were incubated with RNase A and propidium iodide prior to flow cytometric analysis (BD FACSCalibur; Becton Dickinson).
Expression profiling.Hep3B cells transiently transfected with pcDNA3.1-ATF5and geneticin-selected clone at 4th week were studied.Total RNA from transfectants and vector control was subjected to cRNA synthesis and hybridized onto CodeLink Human 20K-Bioarray (Amersham Biosciences).Signals captured by GenePix laser scanner (Axon) were analyzed by CodeLink Expression Analysis software.A differential gene expression was considered when a fold change was >2.0 or <0.5.Hierarchical clustering was performed by Eisen clustering and graphic presentation by Treeview. 5Potential molecular pathways involved were examined by OntoExpress, 6 which utilizes information from the KEGG pathway database.
Electrophoretic mobility shift assay.Hep3B cells transfected with either pEGFP-c2-ATF5 or pEGFP-c2 vector were extracted for nuclear proteins.Equal amounts of nuclear extracts were subjected to ATF5 protein/DNA binding using the LightShift chemiluminescent EMSA (Pierce).A biotin end-labeled DNA oligo containing the CRE consensus sequence of the ID1 promoter (5 ¶-TTTATGAATGGGTGA CGTCACAGGCCTGGCG-3 ¶) was incubated with the nuclear extracts or in combination with anti-green fluorescent protein (GFP) antibody.The DNA-protein and DNA-proteinantibody complexes were resolved on a 6% native PAGE and transferred to a Hybond-N+ membrane.Biotin end-labeled DNA was crosslinked onto the membrane, and detected using a streptavidin-horseradish peroxidase conjugate antibody and chemiluminescent substrate (GE Healthcare Bio-Sciences).
DNA sequencing.PCR amplifications of 5 ¶untranslated region, exon 1 and exon 2, of ATF5 gene were performed using primers designed at the introns exon boundary (Supplementary Table S1).The PCR product was subjected to BigDye Terminator Cycle Sequencing (Applied Biosystems) and analyzed on ABI PRISM3100.Sequence alignment was performed with the support of ''Multiple Sequence Alignment with Hierarchical Clustering.'' 7he data obtained from 23 HCC tumors and 18 cell lines was compared to 10 normal human livers.
Fluorescence in situ hybridization.Dual-color fluorescence in situ hybridization (FISH) was performed on 18 HCC cell lines as described (21).BAC RP11-510I16 containing ATF5 and reference RP11-110A23 on chr.1p31.3 that seldom displays imbalances in HCC (5) were directly labeled with Spectrum Green-dUTP and Spectrum Orange-dUTP, respectively (Vysis).Counterstained in DAPI, each cell line was enumerated for hybridized signals in at least 50 interphase and metaphase cells.
5Aza-dC and TSA treatments.Six HCC cell lines, Hep3B, HKCI-2, HKCI-8, HKCI-9, HKCI-10, and HKCI-C1 were exposed to 5-Aza-2'-deoxycytidine (5Aza-dC) or TrichostatinA (TSA; Sigma-Aldrich) or a combination of both.For 5Aza-dC treatment, medium with 1 Amol/L 5Aza-dC was changed every 24 h for up to 72 h.For TSA treatment, cells were exposed to 300 nmol/L TSA for 24 h.In combinatory treatment, cells were treated with 1 Amol/L 5Aza-dC for 72 h followed by 300 nmol/L TSA for an additional 24 h.In parallel, untreated cells were cultured along with all conditions and served as control experiments.Cells were harvested for total RNA and examined for ATF5 expression by qRT-PCR.
Bisulfitemodified DNA sequencing.Genomic DNA from HCC cell lines was chemically modified by bisulfite treatment using the CpGenome DNA Modification (Intergen).Approximately 1 Ag DNA was denatured in 3 mol/L NaOH and processed according to manufacturer's instructions.Two CpG dense sites within 1,000 bp upstream from exon 1 were indicated in the promoter region of ATF5.The methylation pattern of these two CpG islands (island 1, À973 to À681; island 2, À132 to +219) relative to 3 normal livers was examined by direct sequencing of bisulfite modified DNA.Primer sequences for CpG island 1 were 5 ¶-TGATAGTAGGTGTTGGATAGTTTAA-3 ¶ (sense) and 5 ¶-ACAAACCTTATACACAAAACTCAACTC-3 ¶ (antisense), and the primer sequences for CpG island 2 were 5 ¶-TTAGGA GGAGGAAATTA-GATTT-3 ¶ (sense) and 5 ¶-AAACTACAAAACCATAACC-3 ¶ (antisense).Bisulfite-modified DNA at 100 ng was amplified with specific primers at annealing temperatures of 56jC for CpG island 1 and 54jC for CpG island 2. Freshly prepared PCR products were subcloned into pCR2.1-TOPOvector (Invitrogen).Ten colonies were selected from each sample.Nucleotide sequences of isolated plasmid DNA were determined on ABI PRISM3100 Genetic Analyzer after BigDye Terminator Cycle reaction.
Statistical analysis.Differences in ATF5 expressions between HCC tumors and adjacent nontumoral liver were analyzed by the paired Student's t test.The Mann-Whitney U test was performed to evaluate the ATF5 expression in the early and advanced stage tumors, between genders, and in specimens arising from chronic hepatitis and liver cirrhosis.Functional effects of ATF5 in cell lines studied were assessed by Student's t test.Correlative analysis between ATF5 and ID1 expressions was performed by Pearson correlation.All data were expressed as mean F SEM. Analysis was performed with SPSS and GraphPad Prism.A P value of <0.05 was considered significant.

Results
ATF5 expressions in HCC.Transcriptional mapping along chr.19q indicated two distinct down-regulated genes, ATF5 and KLK5, both located on chr.19q13.33 (Fig. 1A).On further examination, reduced expressions of KLK5 were observed only in cell lines, whereas down-regulations of ATF5 were common in both cell lines and primary tumors (Fig. 1B).Validative analysis by qRT-PCR on the same array-analyzed specimens confirmed a concordant ATF5 down-regulation, except for one case (Fig. 1C).The prevalence of ATF5 repressions was further investigated in a cohort of HCC tumors and paired adjacent nontumoral liver.By qRT-PCR, a median expression of ATF5 at a 0.12 and quartiles from 0.03 to 0.41 was suggested in 87 HCC compared to normal livers.When compared to their adjacent nontumoral liver (median, 1.22; quartiles, 0.39-3.05),ATF5 appeared to be down-regulated in 78% of HCC (60 of 77 cases) by a median factor of 18-fold to >100-fold (P = 0.0004; Fig. 2B).Expressions of ATF5 in the adjacent nontumoral livers were similar between chronic hepatitis and liver cirrhosis (P = 0.88; Fig. 2A).In HCC tumors, levels of ATF5 expression did not show significant differences between early and advanced stage tumors, and between genders (P > 0.09; Fig. 2C  and D).
Functional effects of ATF5.The functional influence of ATF5 on cell growth was examined in three HCC cell lines, Hep3B, HKCI-9, and HKCI-C1, with negligible ATF5 expressions.Ectopic expression of ATF5 induced a consistent growth inhibitory effect in the three cell lines as assessed by colony formation and MTT assays (Fig. 3A and B).The growth attenuation by 30% to 50% was, however, not accompanied by a corresponding increase in apoptotic cells (Fig. 3C).The ATF5-induced growth suppression in Hep3B and HKCI-9 was further studied by flow cytometric analysis, which revealed a cell cycle arrest at the G 2 -M phase in both cell lines (P < 0.002; Fig. 3D).
Potential regulatory pathways and modulated target ID1.The effect of ATF5 on the transcriptional changes of Hep3B was investigated by gene expression profiling.In response to ATF5 ectopic expression, transient and geneticin-selected stable transfectants showed a concordant up-regulation of 57 and downregulation of 55 transcripts.Hierarchical clustering analysis of these 112 candidates depicted distinct cluster dendrograms of ATF5 transfectants and vector controls (Fig. 4A).Bioinformatic analysis of 78 known genes from the 122 differential transcripts suggested few biological pathways including cell cycle, regulation of actin cytoskeleton, mitogen-activated protein kinase (MAPK) signaling, and focal adhesion that were likely modulated by ATF5 (Table 1).In an effort to elucidate potential downstream targets of ATF5, the promoter region of 78 known genes was also examined for the presence of ATF/cAMP-responsive element binding protein  A, ATF5 expressions in the adjacent nontumoral livers were similar between chronic hepatitis and liver cirrhosis (P = 0.88).Nevertheless, profound down-regulations of ATF5 were found in the HCC tumors irrespective of a cirrhotic or chronic hepatitis background (Mann-Whitney U test; ***, P < 0.0001).B, qRT-PCR also indicated significant ATF5 down-regulations in HCC tumors compared to their corresponding adjacent nontumoral liver (n = 77; paired t test; ***, P = 0.0004).C and D, differences in the ATF5 expression were not suggested between gender (P = 0.16) nor between tumors of early and advanced stages (P = 0.09).NS, not significant.

Cancer Research
Cancer Res 2008; 68: (16).August 15, 2008 DUSP1, and YWHAH contain the consensus sequence in the proximity of their promoter region.The ATF/CREB consensus site in ID1 and RRAS was found to locate before the transcription start site at À826bp and À1.6 kb, respectively, whereas in YWHAH and DUSP1, consensus sequence was located after the transcription start site at +71 bp and +46 bp, respectively.Because ID1 harbors the ATF/CREB consensus site in the closest proximity and before the transcription start site, the possible interaction between ATF5 and ID1 was further investigated.Down-regulation of ID1 mRNA and protein in response to ATF5 expression was confirmed in both transient and stable transfectants of Hep3B (Fig. 4B).An inverse relationship of ATF5 mRNA transcribed with ID1 expression levels was further established in HCC tumors and cell lines (n = 35; r = À0.399;P = 0.019; Fig. 4C).The potential direct interaction of ATF5 with the ATF/CREB consensus binding site of ID1 promoter was also examined by EMSA.Due to the lack of an effective antibody against ATF5, recombinant GFPtagged vector pEGFP-c2-ATF5 was constructed and transfected into Hep3B cells with corresponding pEGFP-c2 vector as control.The GFP conjugated ATF5 protein was found to localize predominantly in the nucleus, whereas vector control was contained within the cytoplasm (Fig. 4D).The nuclear extracts from the transfectants were subsequently subjected to EMSA analysis.As shown in Fig. 5F, a supershifted band suggestive of ATF5/DNA binding was detected in the pEGFP-c2-ATF5 transfectant with anti-GFP antibody.This band was found to disappear in the presence of unlabeled competitor oligo.
Inactivation mechanisms of ATF5.To investigate the possible underlying mechanisms in the inactivation of ATF5, mutational analysis, gene copy enumerations, and investigation of epigenetic changes had been carried out.Mutational analysis was conducted on 23 HCC tumors and 18 cell lines with varying expressions of ATF5.Despite a silent mutation that was suggested in the Hep3B cell line (Exon1 Leu30Leu), 3 somatic mutations were found in the primary tumors (7%, 3 of 41 cases).These mutations included Exon2 Leu141Phe, Exon2 Val257Met, and Exon2 Arg275Trp (Supplementary Table S1).According to qRT-PCR analysis, 2 mutated cases showed a down-regulation of ATF5 by 2.6-fold and 8.1-fold as compared to normal livers, although the remaining case did not display aberrant expression.Enumeration of ATF5 copy number by FISH analysis, on the other hand, indicated one or more copy loss in 11 of 18 HCC cell lines (61%), which included HKCI-1, HKCI-3, HKCI-8, HKCI-9, HKCI-C2, SNU387, SNU398, SNU475, PLC/PRF/5, Huh7, and HepG2 (Fig. 5B).
The possible involvement of epigenetic inactivation was investigated in 6 HCC cell lines (Hep3B, HKCI-C1, HKCI-2, HKCI-8, HKCI-9, and HKCI-10).Treatment with demethylating single agent 5Aza-dC indicated a re-expression of ATF5 in 3 of 6 cell Common ATF5 Repressions in HCC www.aacrjournals.orglines (Hep3B, HKCI-C1, and HKCI-8; Fig. 5A).Although the use of histone deacetylase inhibitor TSA alone could not induce ATF5 expressions in any cell lines, the combinatory application with 5Aza-dC was able to stimulate ATF5 re-expressions in the three cell lines (HKCI-2, HKCI-9, and HKCI-10) that were unresponsive to single agent 5Aza-dC.Moreover, a synergistic action of both drugs led to an increased ATF5 re-expression in HCKI-C1 and Hep3B by f40-fold and f5-fold, respectively, as indicated from qRT-PCR.Bisulfite sequencing of the 2 CpG dinucleotides regions of the ATF5 promoter was conducted on Hep3B, HKCI-C1, HKCI-2, HKCI-8, HKCI-9, and HKCI-10.An increased methylation index between À794 to À688 of CpG island 1 was suggested in Hep3B, HKCI-C1, and HKCI-C1-8 compared to reference normal livers (Fig. 5C).Hypermethylation of CpG island 1 in the remaining three cell lines, however, was not suggested.According to the prediction of TESS,8 a number of transcription factor binding sites including NF-E2, E-12, SP1, GC box, and hemoglobin was suggested on regional À794 to À688.For CpG island 2, despite its location being in closer proximity to the transcriptional start site, the methylation content was similar to normal liver controls in 6 of 6 cell lines examined.

Discussion
Transcription factors play essential roles in the control of normal cell physiology, and their alterations can lead to abnormalities in cell proliferation, differentiation, and survival.In many childhood acute leukemias, transcription factors have been altered through chromosomal translocations, resulting in their functional properties being either repressed or inappropriately activated.In this study, we have defined a novel transcription factor ATF5 within the common aberrant region chr.19q13 of HCC.Frequent down-regulations of ATF5 were found in HCC tumors (78%), and significant repressions from their putative premalignant lesion of liver cirrhosis and chronic hepatitis were also indicated (P = 0.0004).Although differences in the down-regulated levels with tumor progression was not suggested because ATF5 is normally transcribed in the liver (23), our finding may have inference for ATF5 down-regulations in the early transcriptional dysregulation contributory to the development of HCC.A, hierarchical clustering of 112 differential-expressed transcripts following ATF5 transfections in Hep3B indicated 2 distinct dendrograms that coincided with transient and geneticin-selected pcDNA3.1-ATF5-transfectedcells relative to vector controls.Genes containing the CRE consensus sequence in the proximity of their promoter region are highlighted (*).B, a reduced ID1 protein translation following ectopic expression of ATF5 was verified in Hep3B cells.C, an inverse relationship between ATF5 expressions and ID1 transcriptions was further established in an independent series of primary HCC tumors and cell lines.D, immunofluorescence analysis indicated predominant nuclear localization of GFP-tagged ATF5 fusion protein, whereas pEGFP-c2 vector remained localized in the cytosol.E, translation of GFP-tagged ATF5 fusion protein (60 kDa) in pEGFP-c2/ATF5-transfected Hep3B cells was confirmed.GFP protein (33 kDa) translated in pEGFP-c2 vector transfection is also shown.F, EMSA was performed using the nuclear extract containing GFP-ATF5 fusion protein and DNA oligos containing ATF consensus sequence.A supershifted band was detected with anti-GFP antibody (Ab) as shown in lane 5, suggesting that ATF5 could directly interact with the CRE consensus binding site of ID1 promoter.
The ATF5 gene belongs to the ATF/CREB family of bZIP proteins that control gene transcription through binding to the CRE consensus DNA sequence on the cellular promoter (20).Despite their diverse activities, ATF/CREB family members share a common ability in mediating cell signaling pathways and maintenance of cellular homeostasis.Furthermore, several ATF family members have been shown to be directly or indirectly involved in cell survival mechanisms.For example, ATF2 have been reported to inhibit growth and metastasis (24).Ectopic expression of B-ATF has been shown to reduce the growth rate of murine cells (25).On the other hand, ATF3 and ATF4 appear to function as positive or negative regulators of cell viability depending on cellular context (26,27).Although ATF5 has been shown to play an essential role in the oligodendrocyte cell survival and neuronal progenitor cell differentiation (28,29), the role(s) of ATF5 in cancer pathogenesis remained largely undefined.
In this study, we demonstrated that restoration of ATF5 expression in HCC cells resulted in growth inhibition through cell cycle arrest at the G 2 -M phase (Fig. 3).Similar to other members of the ATF/CREB family, it would seem that ATF5 also functions in the modulation of transcription networks involved in the cell cycle progression.To further explore the biological pathways and downstream target genes under the transcriptional control of ATF5, we performed gene expression profiling on ATF5transfected HCC cells.Ectopic expression of ATF5 induced >100 differential-expressed transcripts (Fig. 4).Informatic analysis of these differential genes suggested four regulatory pathways including cell cycle, actin cytoskeletal, MAPK signaling, and focal adhesion to be likely affected by ATF5 (Table 1).In support of our finding, we found deregulated cell cycle genes, where the suppressed activities of YWHAH, an inhibitor of DNA damageinduced cell cycle arrest (30) and Cyclin D3, an interacting partner of a G 2 -M-specific protein kinase, p58 (PITSLRE; ref. 31) may well have predisposed to the growth arrest observed.Although it is well-established that disruption of actin microtubule organization can induce G 2 -M arrest (32), cooperative signaling between growth factor receptor tyrosine kinases, integrins, and the actin cytoskeleton is also fundamental to cell cycle progression (33).Under the transcriptional influence of ATF5, we found censored actin cytoskeleton regulation and focal adhesion pathway, where decreased expressions of major components such as RRAS and EGFR were found.Despite that RRAS has a high degree of sequence homology to RAS, unlike RAS, RRAS does not activate the MAPK pathway but rather elicits important integrin-dependent cellular behaviors such as adhesion and actin filamentous organizations  34).EGFR , on the other hand, can promote skin tumorigenesis through mitigating G 2 -M arrest in UV-exposed keratinocytes (35).Thus, repressed genes expression and, thereof, the suppressed pathways resultant from ATF5 restoration may have interfered with the transcriptional signaling network that normally regulates the cell cycle.Studies on neural progenitor cells and oligodendrocyte precursor differentiations have implicated ATF5 as a transcriptional repressor, which competes for binding to the CRE region (28,29).In line with these findings, we found expressions of four CRE site-containing genes, namely ID1, RRAS, YWHAH, and DUSP1, to be repressed in response to ectopic ATF5 expression.A reduced transcription of ID1 mRNA and protein translated after ATF5 transfection was verified in vitro, and an inverse relationship between ATF5 and ID1 expressions was further suggested in HCC tumors (r = À0.399;P = 0.019; Fig. 4C).Result from EMSA demonstrated the likelihood that ATF5 could directly interact with ID1 through the ATF/CRE consensus binding.
ID1 belongs to the helix-loop-helix family of transcription factors, and plays an important role as a signaling modulator of pathways involved in cellular developments, cell cycle progression, and neoplastic transformation (36).The level of ID1 protein correlates significantly with both cancer progression and overall prognosis of a number of human malignancies (15).ID1 is commonly overexpressed in HCC, where a functional role in the regulation of HCC cell growth and hepatic fibrogenesis has been suggested (37,38).Increased expressions of ID1 have also been implicated with increased risk of HCC development in patients with liver cirrhosis (38).Inhibition of ID1 expression, on the other hand, could result in a suppressed cell proliferation possibly through the induction of cellular senescence and G 2 -M cell cycle arrest (39).In prostate cancer, ID1 is thought to exert its stimulatory effects through the activation of MAPK signaling pathway (40).Indeed, the MAPK pathway is also known to play an integral role in coordinating growth and survival signaling in the development and progression of HCC (41).Moreover, overexpression of intermediate proteins of MAPK pathway has been suggested in f70% of HCC tumors (42).It is therefore plausible that repressed ATF5 expression in HCC could result in the upregulation of ID1 and subsequent uncontrolled transcriptional activities, such as augmentation of MAPK signaling and inactivation of cell cycle checkpoint.
Studies on the possible inactivating mechanisms of ATF5 suggested copy loss, promoter hypermethylation, histone modification, and DNA mutation in its common down-regulation in HCC.Of interest, a synergistic action of 5Aza-dC and TSA in the re-expression of ATF5 was suggested in 5 of 6 HCC cell lines examined.This finding might hold therapeutic potentials for HCC patients, as clinical analogues of 5Aza-dC and TSA are currently available.The development of therapies that specifically target transcriptional factor abnormalities holds promise for improving outcome of diseases such as HCC that remains challenging to treat.Moreover, therapeutic strategies based on downstream target ID1 might be effective given the clear relationship of ID1 expression with cancer cell behaviors, and clinical outcomes in patient subgroups (31).Taken together, our present study highlighted a novel transcription factor ATF5 in HCC.It is evident that much effort has still to be made to identify and characterize molecular partners of ATF5 protein, and to dissect the mechanism of ATF5-induced oncogenesis.A better understanding of ATF5 partner proteins and associated pathways may potentially provide new options for therapeutic intervention in patients with HCC.

Figure 1 .
Figure 1.A transcript Map of Chr.19q in HCC.A, normalized gene expression levels of HCC tumors (n = 10) and cell lines (n = 10) have been plotted against physical map locations of cDNA clones in sequential order of megabases.Inverted lines, relative expression levels indicated substantial reduced expressions of clone R08306 representing the ATF5 gene and clone W68361 representing KLK5.B, a consistent down-regulation of ATF5 but not KLK5 was suggested in both HCC tumors and cell lines.C, validative analysis by qRT-PCR on array-analyzed specimens suggested a concordant down-regulation of ATF5 transcriptions in HCC tumors and cell lines, except for one tumor H6.

Figure 2 .
Figure 2. ATF5 expression in HCC tumors and adjacent nontumoral liver.A, ATF5 expressions in the adjacent nontumoral livers were similar between chronic hepatitis and liver cirrhosis (P = 0.88).Nevertheless, profound down-regulations of ATF5 were found in the HCC tumors irrespective of a cirrhotic or chronic hepatitis background (Mann-Whitney U test; ***, P < 0.0001).B, qRT-PCR also indicated significant ATF5 down-regulations in HCC tumors compared to their corresponding adjacent nontumoral liver (n = 77; paired t test; ***, P = 0.0004).C and D, differences in the ATF5 expression were not suggested between gender (P = 0.16) nor between tumors of early and advanced stages (P = 0.09).NS, not significant.

Figure 4 .
Figure 4. Identification of ID1 as potential downstream target of ATF5.A, hierarchical clustering of 112 differential-expressed transcripts following ATF5 transfections in Hep3B indicated 2 distinct dendrograms that coincided with transient and geneticin-selected pcDNA3.1-ATF5-transfectedcells relative to vector controls.Genes containing the CRE consensus sequence in the proximity of their promoter region are highlighted (*).B, a reduced ID1 protein translation following ectopic expression of ATF5 was verified in Hep3B cells.C, an inverse relationship between ATF5 expressions and ID1 transcriptions was further established in an independent series of primary HCC tumors and cell lines.D, immunofluorescence analysis indicated predominant nuclear localization of GFP-tagged ATF5 fusion protein, whereas pEGFP-c2 vector remained localized in the cytosol.E, translation of GFP-tagged ATF5 fusion protein (60 kDa) in pEGFP-c2/ATF5-transfected Hep3B cells was confirmed.GFP protein (33 kDa) translated in pEGFP-c2 vector transfection is also shown.F, EMSA was performed using the nuclear extract containing GFP-ATF5 fusion protein and DNA oligos containing ATF consensus sequence.A supershifted band was detected with anti-GFP antibody (Ab) as shown in lane 5, suggesting that ATF5 could directly interact with the CRE consensus binding site of ID1 promoter.

Figure 5 .
Figure 5. Inactivating mechanisms of ATF5.A, treatment with 5Aza-dC could induced re-expression of ATF5 in HCC cell lines.A synergistic action with TSA was further suggested in HKCI-C1 and HKCI-8 as shown.B, dual-color FISH was performed with ATF5 probe (green ) and internal reference probe (orange ).Bi and Bii, copy loss of AFT5 in HKCI-8 and HKCI-9.C, bisulfite sequencing analysis on CpG islands 1 and 2 of ATF5 promoter showed an increased methylation content within À794 to À688 (in bracket ) of Hep3B, HKCI-C1, and HKCI-8 compared to normal livers (NL ).Each circle represents methylation frequency from the analysis of 10 single clones.