Main

Sustained basic, clinical, and epidemiological studies have previously been performed to address the major problems caused by the fourth most malignant tumor in the world: hepatocellular carcinoma1 (http://www-dep.iarc.fr/). In many previous studies, this has led to the detection of various mutations of genes (eg p53, IGF2R, p16, PTEN) typically associated with the development of hepatocellular carcinoma.2 Besides these specific mutations, the carcinogenesis of hepatocellular carcinoma is known to be correlated with telomere shortening3 and chromosomal instability.4

Chromosomal instability is defined by microscopically detectable partial or complete gain or loss of chromosomes typically occurring in hepatocellular carcinoma and also in many other solid tumors.5, 6 The typical set of chromosomal imbalances found in comparative genomic hybridization studies includes gains of 1q, 6p, 8q, 17q, and 20q, and losses of 1p, 4q, 6q, 8p, 13q, 16q, and 17p detected in >90% of 780 cases analyzed up to now (see references 7, 8, 9, 10 for review:11). These aberrations have been correlated with etiological factors such as viral (hepatitis B, hepatitis C) infections.12, 13, 14 Clinical impact has been reported for losses of 8p and gain of 20q with respect to stage and prognosis.15 On the basis of genomic alterations detected by aCGH, hepatocellular carcinoma and its benign counterpart, hepatocellular adenoma, could be differentiated.16 It was also noted that dedifferentiation of hepatocellular carcinoma was closely correlated to losses of 4q and 13q.11 By fluorescence in situ hybridization, a stepwise increase in genomic instability, as well as a highly significant correlation of histological grading and grade of polyploidization, was seen.17 This is of clinical interest because morphological grading has been shown to correlate with survival.18, 19 Elucidation of the processes causing dedifferentiation of hepatocellular carcinoma is required because hepatocellular carcinoma dedifferentiated in more than half of patients within 7–34 months.20, 21

Investigations into the genomic and mRNA expression level have been attempted to provide insight into these processes, but a combined analysis has not yet been carried out. Using newly developed software tools for the analysis of genomic array data, we therefore aimed to identify genes deregulated during dedifferentiation of hepatocellular carcinoma, particularly those located in chromosomal regions known to be altered in poorly differentiated hepatocellular carcinoma.

Materials and methods

Clinical, serological, and pathomorphological data of 23 patients with hepatocellular carcinoma treated at Hannover Medical School between 1998 and 2002 are given in Table 1. Tumor specimens from these patients were analyzed with regard to mRNA expression in hepatocellular adenoma in an earlier study.22 Tumor material was obtained as surgical specimens taken for diagnostic purposes as permitted by the Ethics Committee of Hannover Medical School. Unfixed tissue samples of 0.5 cm in diameter were snap-frozen within 30 min after resection and stored at −80°C. A part of the tissue was cut out for RNA extraction; another piece was fixed in formalin and embedded in paraffin for histological examination. This ensured the closest proximity and homogeneity of the tissues used for different experiments.

Table 1 Clinical, morphological, and serological data of the 23 patients analyzed
Full size table

Histological classification according to WHO criteria,23 was carried out after H&E and reticulin staining by two experienced pathologists (L.W., P.F.). A total of 4 well, 11 moderately, and 8 poorly differentiated hepatocellular carcinoma were detected with histological signs of partial or complete dedifferentiation (Figure 1).

Figure 1
figure 1

Histological differentiation of hepatocellular carcinoma showing cytological and architectural features. (a) Well-differentiated hepatocellular carcinoma is defined by nuclei shape, which is very similar to normal hepatocytes. Additional staining (eg silver staining) may be required to ensure correct diagnosis. (b) Moderately differentiated carcinoma reveals a heterogeneous pattern with respect to cytology of nuclei and architectural pattern. (c) In moderate-to-poorly differentiated hepatocellular carcinoma, prominent nucleoli are seen in the nuclei; the shape of the nuclei and membrane is irregular. The trabecular pattern is also disturbed in a prominent way, making recognition of trabecular structures difficult. (d) In dedifferentiated hepatocellular carcinoma, nuclei have a bizarre and highly irregular shape, and lose their architectural pattern, making it difficult to separate these carcinomas from dedifferentiated carcinomas of other histogenetic origin.

Full size image

For 22/23 cases of hepatocellular carcinoma, cytogenetic characterization was already performed using aCGH; detailed information is given in an earlier study.16 The findings described according to the international system of cytogenetic nomenclature ISCN24 are given in Supplementary Table 1. In brief, the most commonly gained region was 1q22.1–23.1 in 20 cases (91%). Gains of 8q were observed in 14 cases (64%), losses of 8p in 11 cases (50%), gains of 20q in 9 cases (41%), losses of 13q in 9 cases (41%), losses of 4q in 8 cases (36%), losses of 6q in 8 cases (36%), gains of 6p in 7 cases (32%), and losses of 17p in 7 cases (32%).

Methods

cDNA microarray analysis

cDNA microarray experiments were carried out as previously described.22 In brief, total RNA was isolated using Trizol (Invitrogen, Karlsruhe, Germany) followed by RNeasy Mini Kit (Qiagen, Hilden, Germany) and linearly amplified (MessageAmp aRNA Kit; Ambion, Huntingdon, UK). Aliquots of 1.5 μg of amplified RNA from the tumors and the human reference were reverse transcriptase labeled with Cy3 (reference) and Cy5 (tumor) (Amersham Biosciences, Little Chalfont, UK). Labeled RNA of tumors and references were comparatively hybridized overnight onto a cDNA chip containing the ‘Resgen clone set’ with >43 000 spots covering the entire human genome (Stanford Functional Genomics Facility, Stanford, CA, USA) according to the posted protocol (http://brownlab.Stanford.edu/protocols.html). Arrays were imaged using a GenePix 4000B scanner and dependent software (GenePix Pro 4.1; Axon Instruments). Data were exported to the Stanford Microarray Database (http://smd.stanford.edu). Fluorescence ratios were normalized for each array. Data were excluded if the regression correlation was <0.6, or the mean spot intensity/median background intensity was <1.5. The measured genes were subsequently ‘mean-centered’ (ie reported for each gene relative to the mean ratio across all samples, rather than relative to the human reference).

The hierarchical clustering algorithm based on Pearson correlation (centered) coefficients grouped the genes (as well as the samples) on the basis of similarity in their expression pattern. Normalized and log-transformed expression data served as input for significance analysis of microarrays25 with standard parameters. False discovery rates varied between 0.0 and 0.54.

RT–PCR

Quantitative RT-PCR was performed on the selected genes MT1F, MT1G, and TOP2A, to confirm the array results.

In a total volume of 120 μl, 1 μg of total RNA was reverse transcribed by M-MuLV reverse transcriptase using random hexamer primers (RevertAid First Strand cDNA Synthesis Kit; Fermentas, St-Leon-Rot, Germany). TOP2A cDNA samples (2 μl) were amplified on an iCycler iQ real-time detection system (Bio-Rad Laboratories, München, Germany) using a QuantiTect SYBR Green PCR kit (Qiagen, Hilden, Germany). TBP was used as the internal control gene. Aliquots of 5 μl of each cDNA sample were used for the analysis of MT1F and MT1G expression. This expression was performed using an iCycler iQ real-time detection system (Bio-Rad Laboratories). Primers and probes for MT1F and MT1G were synthesized by TibMolBiol (Berlin, Germany). Detailed PCR conditions and primer sequences for all genes used are shown in Table 2. β-Actin was used as the internal control gene. For calibration, the human reference (Stratagene, Amsterdam, Netherlands) was used. Relative quantification of gene expression was determined according to the ΔΔ-CT method using an iCycler iQ real-time detection system software version 3.1 (Bio-Rad Laboratories). All samples were analyzed in triplicate.

Table 2 Sequences of primers, probes, and cycler conditions used for verification of array results by PCR
Full size table

Results

cDNA Microarray Analysis

Gene expression analysis led to successful evaluation of all 23 cases; 7977 cDNA clones with the most varying expression among the tumor and reference samples were selected for further bioinformatical analysis.

Hierarchical Clustering

Unsupervised hierarchical clustering separated hepatocellular carcinoma into two groups of 11 and 12 cases. All well-differentiated carcinoma were seen in the first group. In the second group, all low-differentiated carcinoma occurred, with the exception of case N91. In the former group, six moderately differentiated carcinoma were seen, whereas five moderately differentiated carcinoma occurred in the latter (Figure 2).

Figure 2
figure 2

Unsupervised hierarchical clustering of global gene expression analysis. Upregulated genes are marked in red, downregulated genes in green. Two groups of hepatocellular carcinoma were obtained using this approach, with all well-differentiated carcinoma clustering in group A and all low-differentiated carcinoma clustering in group B with only one exception (N91). Moderately differentiated carcinomas were found in both clusters. Carcinomas of these two groups were analyzed in regard to histological grading and 13q−.

Full size image

Statistical analysis was further performed by correlation with histological grading. For this parameter, well and moderately differentiated carcinoma were taken together (15 cases) and compared to low-differentiated carcinoma (8 cases). A highly significant correlation (P<0.001, χ2-test) was detectable in regard to the histological grading. For example, four metallothioneins (MT1E, MT1F, MT1M, MT2A) had the lowest fold change (0.09–0.18) in poorly differentiated hepatocellular carcinoma. Supervised cluster analyses for these genes branched off all dedifferentiated tumors from well and moderately differentiated samples (Figure 3).

Figure 3
figure 3

Supervised cluster analyses for members of the metallothionein family branched off all dedifferentiated carcinoma from well and moderately differentiated tumors. Well-differentiated carcinomas are grouped together in one group, whereas low-differentiated carcinoma are in the other group. Upregulated genes are marked in red, downregulated genes in green.

Full size image

Analysis of Gene Expression by Statistical Analysis of Microarrays

Using this approach for ‘histological grading’, the number of significantly dysregulated genes with a false discovery rate of <5% was 860. A total of 827 genes were upregulated in the group of dedifferentiated carcinoma; 33 genes were downregulated (Supplementary Tables 2 and 3). For loss of 13q, the number of dysregulated genes was 531.

For further analysis of these findings, we focused on 20 genes seen with the highest or lowest fold changes (Table 3). For histological grading, the fold change of the most upregulated 20 genes was 2.58–7.47. For the 20 downregulated genes, this change was 0.09–0.47. For the parameter loss of 13q, the top 20 upregulated genes had fold changes of 2.62–4.00. Only two genes (MDFIC and GALE) were downregulated, with fold changes of 0.38 and 0.40, respectively.

Table 3 SAM in regard to histological grading and loss of 13q
Full size table

Six genes were found in both sets of upregulated genes, one from dedifferentiated carcinoma and one from carcinoma with 13q− (BIC, CPNE1, RBPMS, RFC4, RPSA, TOP2A). All of these genes mentioned are involved in cell-cycle control and proliferation. Supervised cluster analyses for these genes, branched off all dedifferentiated carcinoma from well and the moderately differentiated carcinoma. Furthermore, these analyses branch off the most carcinoma with 13q− from carcinoma without 13q− (Figure 4).

Figure 4
figure 4

Supervised clustering including six genes (BIC, CPNE1, RBPMS, RFC4, RPSA, TOP2A) branched off all dedifferentiated hepatocellular carcinoma from well and moderately differentiated carcinoma, with only one moderately differentiated hepatocellular carcinoma sample (N1) occurring in the cluster of dedifferentiated carcinoma. Vice versa, no dedifferentiated carcinoma was seen in the groups of well or moderately differentiated carcinoma. Furthermore, these analyses branched off most of carcinoma with loss of 13q without 13q−. Upregulated genes are marked in red, downregulated genes in green.

Full size image

In the sets of genes downregulated in poorly differentiated carcinoma, four metallothioneins (MT1E, MT1F, MT1M, MT2A) had the lowest fold change (0.09–0.18). All metallothioneins are allocated to chromosome band 16q13. A loss of 16q13 may have caused the downregulation in three samples (N1, N37, N83) according to the results of aCGH.

Real-Time RT-PCR for Relative Quantification of TOP2A, MT1F, and MT1G

Results obtained for mRNA expression analysis were proven exemplarily for genes TOP2A, MT1G, and MT1F. TOP2A was found upregulated in both sets, MT1G and MT1F were downregulated in the group of dedifferentiated hepatocellular carcinoma. RT–PCR confirmed the results of array analysis revealing a systematic increase of mRNA expression of TOP2A and a decrease for MT1G and MT1F as also found by statistical analysis of microarrays, summarized graphically in Figure 5.

Figure 5
figure 5

Graphical illustration of RT–PCR. Array-based mRNA expression analysis was confirmed by these experiments. The decreased gene expression on the mRNA level for MT1F and MT1G was also detectable by quantitative RT-PCR as demonstrated by significantly lower ΔΔ-CT values in low-differentiated carcinoma compared to well and moderately differentiated carcinoma. Vice versa, TOP2A was seen to be upregulated in low-differentiated carcinoma, as expected after array-based expression analysis.

Full size image

Comparison of Cytogenetic Data with mRNA Expression

Because no up- or downregulated gene was located on 13q, and looking for additional factors that could be influenced by deletion of 13q, an Internet-based search using the Sanger Data Bank (www.microrna.sanger.ac.uk) was performed and identified miRNAs mir-621, mir-16-1, and mir-15a within the deleted region of 13q. As reported earlier, miRNAs can alter gene expression in a direct or nondirect way (for review see reference 25), and thus are possible candidates modulating gene expression.

Discussion

Array-based gene expression analyses on hepatocellular carcinoma have been performed to identify genes in regard to grading, staging, and etiological factors associated with clinical outcome.26, 27, 28, 29, 30, 31

As 50–75% of patients develop dedifferentiation within 7–34 months,20, 21 this has a major clinical impact due to the extremely aggressive clinical behavior of dedifferentiated hepatocellular carcinoma.18, 19 We aimed to identify genes involved in the dedifferentiation of hepatocellular carcinoma. We previously showed that chromosomal instability increases parallel to dedifferentiation of hepatocellular carcinoma.17 Moreover, the number of copy number alterations increases parallel to dedifferentiation of hepatocellular carcinoma. Hepatocellular adenoma have a significantly lower number of copy number alterations than carcinoma. Hepatocellular adenoma can also be distinguished from carinoma based on global gene expression profiling.22 Comparing aCGH and gene expression data, it was possible to define a set of seven upregulated genes localized on chromosome band 1q22. As histological dedifferentiation is accompanied by loss of 13q,11, 16 we focused on gene expression in different grades of hepatocellular carcinoma bearing 13q or not. We therefore aimed in this study to analyze whether well and poorly differentiated carcinoma can be distinguished based on gene expression profiling, in order to identify genes involved in the dedifferentiation of hepatocellular carcinoma.

Overall statistical analysis of microarrays analysis revealed 860 genes significantly dysregulated in dedifferentiated hepatocellular carcinoma and 533 genes related to loss of 13q. When comparing 20 genes most upregulated both during dedifferentiation and due to loss of 13q, 6 genes (BIC, CPNE1, RBPMS, RFC4, RPSA, TOP2A) were found. B-cell receptor-inducible gene (BIC) was originally identified as a retroviral integration site in avian leukosis virus,32 and its product, miRNA 155, accumulates in various B-cell neoplasias.33 Upregulation of BIC is important in lymphomagenesis due to interaction with C-MYC,34 which is a gene that also plays a fundamental role in hepatocarcinogenesis. CPNE1 encodes a calcium-dependent membrane-binding protein. It has a broad tissue distribution, and is thought to be involved in membrane trafficking. RBPMS is a member of the RRM family of RNA-binding proteins. The protein encoded by this gene has a single, putative RRM domain, and its function is probably binding to RNA.

RPSA belongs to a family of extracellular matrix glycoproteins called laminins. These proteins have been implicated in cell adhesion, differentiation, migration, signaling, and metastasis, processes frequently altered in cancer cells, determining their invasive and metastatic potential. RFC4, together with PCNA, is necessary for elongation of primed DNA templates by DNA polymerase-δ and DNA polymerase-ɛ. Topoisomerase 2α encodes a protein that controls and alters the topologic states of DNA during transcription. It is found in the nucleus and involved in chromosome condensation, chromatid separation, and the relief of torsional stress during the transcription and replication of DNA. It catalyzes the transient breaking and rejoining of two strands of duplex DNA, which allows the strands to pass through one another, thus altering the topology of DNA. Moreover, high Topoisomerase 2α expression appears to be linked with cellular dedifferentiation and potentially aggressive tumor phenotype in invasive breast cancer.35

For the genes mentioned above, more detailed information can be obtained by accessing the URLs in Supplementary Table 2.

Significantly downregulated genes were seen to be associated with dedifferentiation. Four metallothionein genes had the lowest fold change. Metallothioneins are a group of cysteine-rich intracellular proteins with high affinity for metal ions. Their enzymatic activity is not known. Metallothionein genes are located in chromosome band 16q13. In hepatocellular carcinoma, a decrease in the intensity of immunohistochemical staining has been reported for metallothioneins.36, 37 Metallothioneins were also found to be downregulated in hepatocellular carcinoma in a global mRNA expression analysis by Okabe et al,30 concordant with our results. In patients suffering from hepatocellular carcinoma, metallothioneins were measured in serum, and the concentration correlated inversely with the grade of differentiation of hepatocellular carcinoma.38, 39 In oxidative stress, eg due to infection with hepatitis-B virus or hepatitis-C virus leading to chronic inflammation, metallothioneins are reported to be cell-protective in mice.40 Conversely, MT-1 and MT-2 knockout mice are markedly more sensitive to chemically induced carcinogenesis.41, 42 Very recently, it has been shown that expression of metallothioneins is suppressed in hepatocellular carcinoma mediated through inactivation of C/EBPα by the PI3K/AKT cascade.43 Elements of this pathway (eg ErbB2, mTOR) can be modulated by various drugs44 that are probably candidates for targeted treatment of hepatocellular carcinoma.

Interestingly, none of the downregulated genes were identified within the deleted region of 13q. However, three known miRNAs (mir-621, mir-16-1, mir-15a) are located in the deleted chromosomal region of 13q and therefore may be lost. This could lead to a less effective inhibition of targets of these miRNA and to an increase of gene expression of putative oncogenes, particularly on the protein level. It may also be speculated that partial loss of miRNA positively upregulates critical cell-cycle control genes such as those listed above.

In conclusion, dedifferentiation of hepatocellular carcinoma is associated with upregulation of genes involved in cell-cycle control and proliferation. Notably, a significant portion of these genes is also upregulated in hepatocellular carcinoma with deletion of 13q. It is unclear how genes located in chromosome regions other than 13q are upregulated because downregulated genes were not identified within the deleted region of 13q. As three known miRNAs (mir-621, mir-16-1, mir-15a) are located in the deleted chromosomal region of 13q, we speculate that these miRNAs are involved in dedifferentiation and in the induction of chromosomal instability in hepatocellular carcinoma and should be investigated in further studies.