Hepatocellular carcinoma (HCC) is frequently associated with infiltrating mononuclear inflammatory cells. We performed laser capture microdissection of HCC-infiltrating and noncancerous liver-infiltrating mononuclear inflammatory cells in patients with chronic hepatitis C (CH-C) and examined gene expression profiles. HCC-infiltrating mononuclear inflammatory cells had an expression profile distinct from noncancerous liver-infiltrating mononuclear inflammatory cells; they differed with regard to genes involved in biological processes, such as antigen presentation, ubiquitin-proteasomal proteolysis, and responses to hypoxia and oxidative stress. Immunohistochemical analysis and gene expression databases suggested that the up-regulated genes involved macrophages and Th1 and Th2 CD4 cells. We next examined the gene expression profile of peripheral blood mononuclear cells (PBMC) obtained from CH-C patients with or without HCC. The expression profiles of PBMCs from patients with HCC differed significantly from those of patients without HCC (P < 0.0005). Many of the up-regulated genes in HCC-infiltrating mononuclear inflammatory cells were also differentially expressed by PBMCs of HCC patients. Analysis of the commonly up-regulated or down-regulated genes in HCC-infiltrating mononuclear inflammatory cells and PBMCs of HCC patients showed networks of nucleophosmin, SMAD3, and proliferating cell nuclear antigen that are involved with redox status, the cell cycle, and the proteasome system, along with immunologic genes, suggesting regulation of anticancer immunity. Thus, exploring the gene expression profile of PBMCs may be a surrogate approach for the assessment of local HCC-infiltrating mononuclear inflammatory cells. [Cancer Res 2008;68(24):10267–79]
- peripheral blood mononuclear cells
- hepatocellular carcinoma
- laser capture microdissection
- infiltrating mononuclear inflammatory cells
- gene expression profile
Hepatocellular carcinoma (HCC) is one of the most frequent malignancies worldwide ( 1). It commonly develops from chronic liver diseases, such as viral hepatitis ( 2) and chronic hepatitis, resulting from hepatitis C virus (HCV) infection, is a major risk factor. Indeed, 7% of patients with liver cirrhosis (LC) caused by persistent HCV (LC-C) infection develop HCC annually ( 3).
Cancer tissues are often associated with infiltrating inflammatory cells, such as tumor-associated macrophages ( 4), T lymphocytes ( 5), and antigen-presenting cells ( 6). These tumor-infiltrating mononuclear inflammatory cells are thought to be important modulators of HCC ( 7). However, their actual role remains controversial. Increased numbers in HCC have been correlated with a fair prognosis ( 8), but tumor-infiltrating mononuclear inflammatory cells in HCC tissues have also been found to involve more FOXP3+ regulatory T cells ( 9) and provide a cancer-favorable environment that leads to resistance to therapy. Characterization of tumor-infiltrating mononuclear inflammatory cells may be valuable in understanding tumor immunology and, possibly, in predicting the prognosis of HCC patients ( 7).
Peripheral blood mononuclear cells (PBMCs) consist of immune cells, such as monocytes and lymphocytes, and are essential players in the host immune defense system, which responds to various abnormal conditions in the host ( 10). PBMCs and tumor-infiltrating mononuclear inflammatory cells contain CTLs, specifically cytocidal to cancer tissues ( 11) and regulatory T cells that can suppress the host immune response against cancer ( 9). Thus, PBMCs may potentially reflect host immune status. However, there are limited assays for assessing the immune status of PBMCs, such as a proliferation assay, measurements of cytokine production, and the assessment of cytocidal potential.
The advent of cDNA microarray technology for the analysis of gene expression profiles has been useful in comprehensively disclosing underlying molecular features and has provided considerable information for basic science and clinical medicine. We have analyzed gene expression in liver diseases ( 12, 13) and believe it may become a useful diagnostic tool using liver tissue biopsy samples ( 14). We have also reported that gene expression profiling of PBMCs predicted the effect of IFN for the eradication of HCV ( 15) and can provide biomarkers not only for the control of blood sugar but also possibly for predisposing diabetic factors ( 16). Gene expression profiling of PBMCs from patients with renal cell carcinoma can be used to predict their response to systemic chemotherapy ( 17). Thus, gene expression information from the cellular components of peripheral blood may be useful in interpreting the internal condition of the patient.
In this study, we used DNA microarray technology to examine differences in gene expression profiles between HCC-infiltrating and noncancerous liver-infiltrating mononuclear inflammatory cells, which were selectively microdissected ( 12), and the gene expression profiles of PBMCs from LC-C patients with or without HCC. We observed distinct transcriptional features of HCC-infiltrating mononuclear inflammatory cells, reflecting the immune status of the local environment. Intriguingly, the transcriptional features of the HCC-infiltrating mononuclear inflammatory cells were shared with PBMCs from HCC patients. Thus, we suggest the possibility that the gene expression profile of PBMCs may be useful as a clinical surrogate biomarker for the assessment of the internal environment of HCC patients with chronic hepatitis C (CH-C) infection.
Materials and Methods
Study subjects. All patients participating in this study had advanced chronic liver disease, cirrhosis, or persistent HCV infection. Twelve patients who developed HCC as a consequence of advanced chronic liver disease related to hepatitis C and who underwent surgical treatment were enrolled (Supplementary Table S1). HCC and noncancerous liver tissues were obtained and frozen. For analysis of gene expression profiles in PBMCs, 32 LC patients without HCC and 30 LC patients with HCC (Supplementary Table S2) were included. Development of HCC was diagnosed by computed tomography (CT) or magnetic resonance imaging with contrast reagents and abdominal angiography with CT imaging in arterial and portal flow phases ( 18). The pathologic tumor node metastasis classification system of the Liver Cancer Study Group of Japan was used for the staging of HCC. LC was diagnosed by pathologic findings in biopsy specimens where available; otherwise, radiological imaging, platelet counts, serum hyaluronic acid levels, and indocyanine green retention rates were considered for the diagnosis of cirrhosis. The study has been approved by the institutional review board, and informed consent was obtained from all patients enrolled in the study.
Isolation of PBMCs. PBMCs were isolated from heparinized blood samples by Ficoll-Hipaque density gradient centrifugation, as reported previously ( 15).
Laser capture microdissection. HCC and noncancerous liver tissues obtained during surgery were frozen in optimum cutting temperature compound (Sakura Finetech; ref. 13). All HCC tissues were nodular and clearly separated by noncancerous tissues macroscopically. Cells infiltrating HCC tissues were visualized under a microscope and precisely excised by laser capture microdissection (LCM) using a CRI-337 (Cell Robotics, Inc.), as previously performed (Supplementary Fig. S1A; ref. 12). Cells infiltrating noncancerous tissues of CH-C patients were visualized and excised similarly.
RNA isolation and amplification. Total RNA was isolated from PBMCs or tissue samples using a microRNA isolation kit (Stratagene) in accordance with the supplied protocol with slight modifications. Isolated RNA was then amplified twice using antisense RNA and an Amino Allyl MessageAmp aRNA kit (Ambion), as described previously ( 13). The reference RNA sample was isolated from the PBMCs of a 29-yr-old healthy male volunteer and was amplified in the same manner. Amplified RNAs from the PBMCs of patients and the healthy volunteer were labeled with Cy5 and Cy3 (Amersham), respectively. Equal amounts of amplified RNAs were hybridized to an oligo-DNA chip (AceGene Human Oligo Chip 30K, Hitachi Software Engineering Co., Ltd.) overnight and were then washed for image scanning.
DNA microarray image analysis. The fluorescence intensity of each spot on the oligo-DNA chip was determined using a DNA Microarray Scan Array G (PerkinElmer). The images obtained were quantified using a DNASIS array (v2.6, Hitachi Software Engineering Co., Ltd). For normalization, the intensity of each spot without oligo-DNA was subtracted from that with oligo-DNA in the same block. A validated spot was determined when the intensity of the spot was within the intensity ±2 SDs for each block. By calibrating the median to base quantity, the intensities of all spots were adjusted for normalization between Cy5 and Cy3.
Quantitative real-time detection PCR. Real-time detection PCR (RTD-PCR) was performed as previously described ( 15). Briefly, template cDNA was synthesized from 1 μg of total RNA using SuperScript II RT (Invitrogen). Primer pairs for chemokine (C-C motif) receptor 1 (Ccr1), histone acetyltransferase 1 (Hat1), mitogen-activated protein kinase kinase 1 interacting protein 1 (Map2k1ip1), phosphatidylinositol glycan anchor biosynthesis, class B (PigB), toll-like receptor 2 (Tlr2), superoxide dismutase 2 (Sod2), cytokeratin 8 (Krt8), Krt18, Krt19, and glyceraldehydes-3-phosphate dehydrogenase, as an internal control of expression, were purchased from the TaqMan assay reagents library (Applied Biosystems). Synthesized cDNA was mixed with the TaqMan Universal Master Mix (Applied Biosystems), as well as each primer pair and reaction was performed using ABI PRISM 7900HT. Relative expression level of each gene was calculated compared with that of internal control in each sample. Results are expressed as means ± SE.
Flow cytometry analysis. Flow cytometry analysis was performed as described previously ( 19). Briefly, isolated PBMCs were incubated in PBS supplemented with 2% bovine serum albumin (Sigma-Aldrich JAPAN K.K.) with antihuman CCR1 and CCR2 antibodies labeled with Alexa Fluor 647 (Becton Dickinson Pharmingen). The fluorescence intensity of the cells was measured using a FACSort (Becton Dickinson).
Immunohistochemistry. Surgically obtained HCC and noncancerous liver tissues were fixed with neutral buffered formalin, embedded in paraffin, cut into 4-μm sections, and mounted on microscope slides. The fixed slides were deparaffinized and subjected to heat-induced epitope retrieval 98°C for 40 min. After blocking endogenous peroxidase activity in the tissue specimen using 3% hydrogen peroxide, the slides were incubated with appropriately diluted primary antibodies, antihuman CD4 or antihuman CD14 mouse monoclonal antibodies (Visionbiosystems Novocastra). The reaction was visualized by the REAL EnVision Detection System (DAKO) followed by counterstaining with hematoxylin.
Statistical analysis. Hierarchical clustering and principal component analysis of gene expression was performed using BRB-ArrayTools. 1 Fisher's exact test was used to examine the significance of hierarchical clustering in the dendrogram. A class prediction was performed by three nearest neighbors, incorporating genes that were differentially expressed at the P = 0.002 significance level, as assessed by the random variance t test (BRB-ArrayTools). For genes to analyze in a pathway, we used a P value of <0.05 with 2,000 permutations to avoid underestimating the presence of meaningful signaling pathways that were coordinately up-regulated or down-regulated with subtle differences ( 13). The cross-validated misclassification rate was computed, and at least 2,000 permutations were performed for a valid permutation P value. The univariate t values for comparing the classes were used as weights. Student's t-test was performed for RTD-PCR data, and P values of <0.05 were deemed to be statistically significant. The population of CCR1-positive or CCR2-positive cells in PBMCs by flow cytometry analysis was tested for differences (with P < 0.05) by the Mann-Whitney U-test, using SPSS software (SPSS Japan, Inc.).
Analysis of expression data for biological processes and networks. As for genes significantly up-regulated or down-regulated in HCC-infiltrating mononuclear inflammatory cells compared with noncancerous liver-infiltrating mononuclear inflammatory cells or in PBMCs in LC without HCC compared with LC with HCC at P < 0.05, we have performed analysis of the biological processes using the MetaCore software suite (GeneGo), as described previously ( 13). Possible networks were created according to the list of the differentially expressed genes using the MetaCore database, a unique curated database of human protein-protein and protein-DNA interactions, transcription factors, and signaling, metabolic, and bioactive molecules. The P value was calculated as described previously ( 13).
Gene expression data of major leukocyte types and analysis of DNA microarray expression data. Gene expression data for leukocytes were retrieved through publicly accessible databases. 2 The gene set database GDS1775, which includes gene expression data for major leukocyte types, was obtained and subjected to one-way clustering analysis using BRB-Array Tools with genes that were up-regulated in HCC-infiltrating mononuclear inflammatory cells for the enrolled cases above.
Gene expression in mononuclear inflammatory cells infiltrating into HCC tissue. HCC is frequently associated with infiltrating mononuclear inflammatory cells ( 20), and various attempts have been made to understand their biological significance ( 8, 9, 21). We selectively obtained HCC-infiltrating mononuclear inflammatory cells by LCM and compared their gene expression profiles with those of noncancerous liver-infiltrating mononuclear inflammatory cells obtained in the same way (Supplementary Fig. S1A; Supplementary Table S1). The gene expression profiles of HCC-infiltrating mononuclear inflammatory cells showed that 115, 206, and 773 genes were up-regulated and 52, 114, and 750 genes were down-regulated compared with those of noncancerous liver-infiltrating mononuclear inflammatory cells at P levels of <0.005, <0.01, and <0.05, respectively (Geo accession no.3 GSE 10461; Supplementary Fig. S1B).
Genes at the P < 0.05 level were analyzed with regard to their role in biological processes in HCC-infiltrating mononuclear inflammatory cells compared with noncancerous liver-infiltrating mononuclear inflammatory cells using the MetaCore pathway analysis software. The significant processes, in which the up-regulated genes in HCC-infiltrating mononuclear inflammatory cells were involved, included antigen presentation, an immunologically important process in antigen-presenting cells, such as monocyte/macrophages and dendritic cells ( Table 1 ; ref. 22). The genes involved in this process were the genes for the CD1d molecule and C-type lectin domain family 4 for glycolipid antigen recognition ( 23, 24) and CD86, an accessory molecule indispensable for provoking an immune response ( 25), suggesting an activated immune reaction in these cells. The up-regulated genes in HCC-infiltrating mononuclear inflammatory cells were also involved in the ubiquitin-proteasomal proteolysis process, with significant genes, such as those encoding ubiquitin-conjugating enzymes and proteasome subunits. This process is required to eradicate unnecessary proteins, which are ubiquitinated, and then degraded in proteasomes ( 26). Processes related to the steps of gene expression, such as transcription by RNA polymerase II, mRNA processing, and the process of the cell cycle were also represented in the genes up-regulated in HCC-infiltrating mononuclear inflammatory cells, indicating enhanced cellular activity. Genes involved in the process of double-strand breaks, such as topoisomerase II α4 ( 27), and proliferating cell nuclear antigen (PCNA; ref. 28) genes involved in responses to hypoxia and oxidative stress, such as thioredoxin, peroxiredoxin, and antioxidant protein, were also up-regulated, suggesting that HCC-infiltrating mononuclear inflammatory cells were in an activated inflammatory status and under hypoxic or oxidative stress, presumably caused by the HCC. Thus, the profile of up-regulated genes in HCC-infiltrating mononuclear inflammatory cells suggested an inflammatory status, possibly triggered by antigenic stimulation of HCC tissues.
Fewer processes were identified for the down-regulated genes. One intriguing process identified was that of integrin-mediated cell matrix adhesion, suggesting that HCC-infiltrating mononuclear inflammatory cells may be less adhesive in the local tissues where they were found (Supplementary Table S3).
Subpopulation analysis of HCC-infiltrating mononuclear inflammatory cells using immunohistochemistry and transcriptional analysis. Tumor-infiltrating mononuclear inflammatory cells consist of a mixed cell population, including macrophages, effector T cells, and regulatory T cells, which have been considered to be both cancer-favorable or cancer-unfavorable ( 8, 21). HCC-infiltrating and noncancerous liver-infiltrating mononuclear inflammatory cells were immunohistochemically evaluated to examine the characteristics of the subpopulations. CD14-positive monocytes/macrophages were prominent in HCC-infiltrating mononuclear inflammatory cells, whereas they were rarely observed in noncancerous liver-infiltrating mononuclear inflammatory cells ( Fig. 1A ). CD4-positive helper T cells were observed in both HCC tissues and noncancerous liver tissues, although in noncancerous liver tissues, these cells tended to accumulate within the aggregates of mononuclear inflammatory cells, whereas they seemed to be scattered in HCC-infiltrating mononuclear inflammatory cells ( Fig. 1A).
Next, we examined the genes that were significantly up-regulated in HCC-infiltrating mononuclear inflammatory cells compared with noncancerous liver-infiltrating mononuclear inflammatory cells, relative to subpopulations of leukocytes, and explored how they may be relevant to leukocyte subpopulations, using the database of the human immune cell transcriptome in the Gene Expression Omnibus3 (Geo accession no. GDS1775), which covers 26 immune regulatory cells, such as T cells, B cells, natural killer cells, macrophages, dendritic cells, basophils, and eosinophils. Among the 206 extracted, up-regulated genes in HCC-infiltrating mononuclear inflammatory cells (at the P < 0.01 level), 97 annotated genes were used for one-way hierarchical clusters ( Fig. 1B). Most genes among 97 annotated up-regulated genes in HCC-infiltrating mononuclear inflammatory cells were shown to be expressed with higher magnitude in lipopolysaccharide-stimulated or lipopolysaccharide-unstimulated macrophages than in other types of major leukocytes. The next subpopulations, including the second most number of genes for relatively high magnitude of expression, were Th1 and Th2 CD4 cells under conditions supplemented with interleukin-12 (IL-12) and IL-4, respectively (Geo accession no.3 GSM90858), secreting Th1 and Th2 cytokine profiles, respectively, suggesting that featured genes expressed in HCC-infiltrating mononuclear inflammatory cells were indicative of CD4 helper T cells, secreting a variety of cytokines.
Thus, this expression analysis showed that, in HCC lesions with tumor antigens, there was an accumulation of antigen-presenting cells, monocyte/macrophages, and CD4 helper T cells, which were in a cytokine-secreting condition, with enhanced cellular biological activities, including ubiquitin-proteasomal proteolysis, presumably under a hypoxic and oxidative stress environment caused by the HCC. The overall inflammatory status represented by HCC-infiltrating mononuclear inflammatory cells was not determined in terms of an anticancer effect, because no obvious shift of CD4 helper T cells to the Th1 or Th2 condition was indicated.
Distinct gene expression profile of PBMCs obtained from patients with cirrhotic liver disease complicated with HCC. The HCC-infiltrating mononuclear inflammatory cells were distinct in terms of expressed genes. The putative biological processes involving these up-regulated genes in tumor-infiltrating mononuclear inflammatory cells suggested a general influence of the HCC on the local environment of the host, represented by stress-response genes. We, thus, examined whether PBMCs in the systemic circulation of the patient might also be influenced by the development of HCC. PBMCs were obtained from 30 patients with LC associated with HCC and from 32 patients with LC not associated with HCC, and the gene expression profiles were compared (Geo accession no.3 GSE10459).
Unsupervised hierarchical clustering analysis using 17,903 filtered genes, the expression values of which were not missing in >50% of the cases, identified two major clusters of patients, with and without HCC (data not shown). To examine the reproducibility and the reliability of the clustering, we excluded unchanged genes in all samples (genes with less than a 1.8-fold difference in >85% of samples) to remove noise. This hierarchical clustering analysis using 1,917 filtered genes confirmed two clear clusters in patients with or without HCC ( Fig. 2A ). In one major cluster, including the most LC cases, there was a subcluster, LC/HCC, which included more of the HCC patients located next to the cluster of patients with HCC (LC/HCC; Fig. 2A). The reproducibility of the clustering (proportion, averaged over replications and over all pairs of samples in the same cluster, BRB-ArrayTools) was 93%. Sensitivity and specificity to HCC in this cluster analysis is 88% and 76%, respectively. These cirrhotic patients without HCC were followed for at least a further 12 months to detect HCC; none of those in the LC group developed HCC over this time. The principal component analysis was performed with the filtered 1,917 genes and the two major groups; classifying LC and HCC were similarly observed ( Fig. 2B).
To further confirm that gene expression in the PBMCs of patients with HCC was distinct from that in patients without HCC, analysis of PBMC gene expression was performed by a supervised learning method using categories of LC-C or HCC, age, gender, serum alanine aminotransferase (ALT), and α-fetoprotein (AFP). It showed that patients with or without HCC were significant classifiers (P < 0.0005), assigned with 1,430 predictor genes (P < 0.002; Table 2 ). Of 32 patients with LC, eight (25%) were misclassified as having HCC, and 2 of 30 patients with HCC (6.7%) were misclassified as not having HCC, indicating that the overall accuracy of the prediction of a patient with or without HCC was 84% ( Table 2). Other clinical variables supposed to be related to HCC occurrence, such as age (ref. 29; >68 or ≤ 68 years old), gender ( 30), and ALT(ref. 31; >50 or ≤50 IU/L), could not differentiate gene expression in PBMCs. AFP (>20 or ≤20 ng/mL) was actually significant but was a much less powerful classifier (P < 0.02, assigned with 301 classifier genes). The prediction accuracy for categories of LC-C versus HCC and the AFP value >20 versus ≤20 ng/mL is not significantly affected whenever the number of predictor genes is reduced to below 62 (Supplementary Fig. S2). Taken together, these results by unsupervised and supervised analysis methods indicate that HCC development in LC-C patients significantly affects the gene expression profile in PBMCs.
Features of biological processes for which gene expression was significantly altered in PBMCs in HCC patients. We next examined the biological processes possibly affected by HCC development, given the expression profiles in PBMCs from patients with HCC. Statistical analysis showed that 867 genes were up-regulated and 989 genes were down-regulated in PBMCs from patients with HCC, compared with those without HCC (P < 0.005). Six representative genes, Ccr1, Hat, Map2k1ip1, PigB, Tlr2, and Sod2, were randomly selected from genes which were biologically important and differentially expressed between LC and HCC groups, and their expression was confirmed by RTD-PCR (Supplementary Fig. S3A). To exclude the possibility of circulating cancer cells, we have also examined the expression of Afp, Krt8, Krt18, and Krt19. No expression was detected for Afp (data not shown), and no statistically significant difference was found for expression of Krt8, Krt18, and Krt19 between patients with HCC and without HCC (Supplementary Fig. S3A). The expression data were also confirmed by flow cytometric analysis. We evaluated how many cells in blood expressed CCR1 and CCR2 and confirmed that populations expressing CCR1 and CCR2 were significantly higher in PBMCs from patients with HCC than those without (Supplementary Fig. S3B). To understand the biological processes in PBMCs for which up-regulated or down-regulated genes were observed, we used MetaCore. The up-regulated genes in PBMCs from patients with HCC were involved in processes such as ubiquitin-proteasomal proteolysis (e.g., heat shock 70 kDa protein 4, ubiquitin conjugating enzymes), mRNA processing (e.g., heterogeneous nuclear ribonucleoproteins, RNA methyltransferase), antigen presentation (e.g., MHC class I polypeptide-related sequence A, B), cell cycle (e.g., HAT1, PCNA), and the response to hypoxia and oxidative stress (e.g., glutaredoxin 2, SOD2, thioredoxin; Table 3 ). These differentially up-regulated biological processes were also up-regulated processes in HCC-infiltrating inflammatory cells ( Table 1). Thus, PBMCs from HCC patients present antigens in conditions of hypoxia and oxidative stress. Additionally, genes involved in other processes, such as apoptosis (e.g., apoptotic peptidase activating factor 1, caspase 9) and T-cell receptor (TCR) signaling (e.g., CCR1, CCR2, TCR α-chain), were also up-regulated in PBMCs from patients with HCC, suggesting vulnerabilities of PBMCs and activated T-cell signaling, respectively, in HCC development.
Biological processes involving the down-regulated genes in PBMCs from patients with HCC included skeletal muscle development, the estrogen receptor 1 (ESR1) nuclear pathway, NOTCH signaling, feeding, and neurohormones signaling, neurogenesis, leptin signaling, and IL-12, IL-15, and IL-18 signaling (Supplementary Table S4), showing no obvious connection compared with the down-regulated genes in HCC-infiltrating mononuclear inflammatory cells (Supplementary Table S3). These results indicate that HCC development in cirrhotic liver can influence PBMCs, providing distinct transcriptional features of up-regulated genes even during the operable stage of HCCs.
Networks of genes commonly up-regulated or down-regulated in both PBMCs and HCC-infiltrating mononuclear inflammatory cells. Analysis of the gene expression profiles of HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients showed that the development of HCC altered the gene expression of local infiltrating mononuclear inflammatory cells and systemically circulating PBMCs; interestingly, the affected biological processes were largely the same. To further explore these presumed local and systemic influences resulting from HCC development, we examined how individual genes were affected by constructing a network.
We found 773 up-regulated and 750 down-regulated significant genes in HCC-infiltrating mononuclear inflammatory cells compared with noncancerous liver-infiltrating mononuclear inflammatory cells at the P < 0.05 level. In PBMC gene expression, we observed 2,111 up-regulated and 2,027 down-regulated genes in the PBMCs of HCC patients, compared with LC patients at the P < 0.05 level. Among these genes, 378 were significant in both HCC-infiltrating mononuclear inflammatory cells and PBMCs from patients with HCC ( Fig. 3A ). For these 378 genes commonly altered genes, 70% of them were up-regulated or down-regulated in both HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients, whereas expression of the remaining 30% of them was discordant.
We used MetaCore software to perform network construction for 172 up-regulated and 93 down-regulated genes in both HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients. The signal pathway network revealed three central genes, PCNA ( 32), SMAD3 ( 33), and nucleophosmin ( 34), which were all up-regulated in HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients ( Fig. 3B). PCNA had interactions with proteasome subunit genes, PSMC2, PSMC6, PSMD12, and thioredoxin and DNA polymerase iota genes. SMAD3 was linked with cyclin-dependent kinase 7 and cyclin G2 with various genes related to the cell cycle. Nucleophosmin was connected to ubiquitin-conjugating enzyme e2e3 and glutaredoxins. Notably, FOXP3, a marker of regulatory T cells, and Janus-activated kinase 3 (JAK3), related to interleukin signaling ( 35), were up-regulated and down-regulated, respectively, in HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients in the constructed gene network.
The network constructed for individual genes whose expression was commonly altered in HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients also supported a condition of HCC-related stress. The network also indicated that immune reactions in patients with HCC are complex, because down-regulated JAK3, an interleukin signaling molecule, and up-regulated FOXP3 and SMAD3, known molecules of anticancer immunity, are involved in this network. Biological processes in HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients also included the antigen-presentation process.
In this study, we explored gene expression in local infiltrating mononuclear inflammatory cells in HCC and noncancerous liver tissues and in PBMCs obtained from patients with hepatitis C–related LC, with or without HCC. Gene expression profiles of HCC-infiltrating mononuclear inflammatory cells were quite distinct from those of noncancerous liver-infiltrating mononuclear inflammatory cells, showing their differing roles in anticancer immunity. We also investigated gene expression in systemically circulating PBMCs from LC-C patients with or without HCC and found that PBMC gene expression profiles from patients with or without HCC were significantly different. Intriguingly, many biological processes involving the up-regulated genes were shared between HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients, suggesting that the local inflammatory effect evoked by HCC development is systemically projected in the host.
Tumor-infiltrating mononuclear inflammatory cells have been investigated to examine their roles in local cancer tissues. We have selectively obtained aggregates of infiltrating mononuclear inflammatory cells in HCC and noncancerous liver tissues by LCM without contamination of carcinoma or parenchymal cells. We have shown that the process of antigen-presentation ( 36) is a distinguishing feature for up-regulated genes in HCC-infiltrating mononuclear inflammatory cells compared with noncancerous liver-infiltrating mononuclear inflammatory cells. Consistently, immunohistochemical staining of HCC and noncancerous liver tissues revealed that the HCC-infiltrating mononuclear inflammatory cells are primarily monocytes/macrophages, a lineage of phagocytes and antigen-presenting cells ( 37). Helper CD4 T cells were also found but seemed to be scattered in the HCC-infiltrating mononuclear inflammatory cells, compared with their intensive accumulation in infiltrating mononuclear inflammatory cells in noncancerous liver tissues. Correspondingly, analysis using a publicly available gene expression database of major leukocytes showed that up-regulated genes in HCC-infiltrating mononuclear inflammatory cells were primarily featured for macrophages and Th1 and Th2 CD4 cells, preconditioned with IL-12 and IL-4, respectively. These findings could be interpreted in that HCC expresses tumor-antigens ( 38) different from the surrounding noncancerous liver tissues; consequently, phagocytes gather in HCC tissues, take up antigens expressed by HCC tissues, and interact with CD4 cells ( 39). The scattered distribution and transcriptional features of both the Th1 and Th2 predisposed status of CD4 helper T cells in HCC-infiltrating mononuclear inflammatory cells suggests their versatile inflammatory status in cancer immunity, although there was no obvious shift of the Th1/Th2 balance, which is considered to be important in cancer immunity ( 40).
Other characteristic biological processes involving the up-regulated genes in HCC-infiltrating mononuclear inflammatory cells included the response to hypoxia and oxidative stress ( 41), the ubiquitin-proteasome system, cell cycle, mRNA processing, ER, and cytoplasm. The ubiquitin-proteasome system is unique to eukaryotic cells and important in maintaining the normal biological activity of cells, with pleiotropic effects in higher animals ( 42). The cell cycle requires precise regulation of cyclin-dependent kinase under strict control by ubiquitination and subsequent protein degradation ( 32). Taken together, these processes involving the up-regulated genes may reflect a protective local response of the host, corresponding to the stress environment of HCC. In this sense, the double-strand break repair gene up-regulation may be interpreted as the cells responding to maintain normal cellular activities although they are exposed to a harmful environment by the HCC ( 43).
The biological processes involving the up-regulated genes in PBMCs from HCC patients, compared with those from LC-C patients without HCC, were, to a substantial degree, the same, involving the up-regulated genes in HCC-infiltrating mononuclear inflammatory cells, such as ubiquitin-proteasomal proteolysis, ER, and cytoplasm, mRNA processing, antigen presentation, the cell cycle, and the response to hypoxia and oxidative stress. The reflection of these transcriptional features of HCC-infiltrating mononuclear inflammatory cells by PBMCs from HCC patients suggests a systemically projected influence of local HCC development, which is presumably the result of the stress environment caused by HCC and the host's reaction even when the size of the tumor is relatively small. In addition to exploring these biological processes, we also constructed networks of individual genes, the expression of which was similarly up-regulated or down-regulated, to depict commonly affected biological processes in tumor-infiltrating mononuclear inflammatory cells and PBMCs under HCC development in more detail. The networks highlighted three central genes, nucelophosmin, PCNA, and SMAD3, as up-regulated genes. They are connected to individual genes involved in ubiquitin, proteasomes, the cell cycle, and oxidative stress ( Fig. 3B). Interestingly, the immunologically important molecules, FOXP3 and JAK3, are in the network as up-regulated and down-regulated genes, respectively. FOXP3 is a transcriptional marker for regulatory T cells ( 44), and SMAD3 is also believed to be important in maintaining regulatory T cells ( 45). JAK3, which is associated with the interleukin receptor common γ chain ( 35) and is important in lymphoid development ( 46), was also involved in the network, suggesting that HCC influences the host immune system, which can be observed not only in HCC-infiltrating mononuclear inflammatory cells but also in the PBMCs of HCC patients. Thus, the network features of individual genes, commonly affected in HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients, further imply that the anticancer immunity of the host in response to HCC development involves the antigen presentation process to initiate the immune reaction.
The mechanism by which PBMCs from HCC patients reflect the transcriptional features of HCC-infiltrating mononuclear inflammatory cells requires further study. We observed that the population of CCR1-expressing and CCR2-expressing cells in PBMCs from HCC patients was higher than in those from LC-C patients. However, HCC-infiltrating mononuclear inflammatory cells did not show up-regulation of these genes. The meaning of the up-regulated CCR1 and CCR2 should be further investigated because chemokines are key molecules for the recruitment of inflammatory cells, regulating cellular adhesion and transendothelial migration, and the activation of inflammatory cells ( 47). The biological process of integrin-mediated cell matrix adhesion, genes involved in which were down-regulated in HCC-infiltrating mononuclear inflammatory cells, may suggest that these cells were able to remigrate into the microcirculation with the enriched blood flow in HCC tissues. The process of integrin-mediated cell matrix adhesion in HCC-infiltrating inflammatory cells may imply weaker adhesion of infiltrating mononuclear inflammatory cells to cancer tissues compared with noncancerous liver tissues ( 48). PBMCs are also presumed to be affected by humoral factors from HCC tissues ( 49). Another possibility is the presence of hematogenous spreading and circulating HCC cells because mRNA for AFP was detected in circulation ( 50). Because two-thirds of HCC patients enrolled for gene expression analysis of PBMCs showed serum AFP value <100, the presence of circulating HCC cells would not be evaluated by the detection of Afp gene expression alone. Therefore, we have examined expression of Krt8, Krt18, and Krt19, as well as Afp. Despite of the possibility of circulating cancer cells, we neither detected expression of Afp nor found significantly different expression of Krt8, Krt18, and Krt19 between HCC and LC-C patients without HCC. Furthermore, genes up-regulated in HCC tissues compared with noncancerous liver tissues 3 did not correlate to up-regulated genes in PBMCs of HCC patients, indicating that different signature of gene expression in PBMCs between HCC and LC-C patients is not the reflection of the possible migrating cells from HCC tissues. In addition, all HCC cases, except for a case in gene expression analysis of PBMCs, were radiologically free of tumor thrombus in the vessel, which was indicative of microscopic invasion free or concomitant with invasion in the periphery of third or lower branch of vessels, suggesting that contribution of circulating cancer cells were presumed to be sufficiently small for the distinct difference of gene expression signature of PBMCs.
Although the number of enrolled HCC patients for analysis with local inflammatory cells was relatively small compared with the number of patients for analysis of PBMCs, our study has shown shared features of gene expression profiles of HCC-infiltrating mononuclear inflammatory cells and PBMCs from HCC patients, showing a complex immune status of the host in anticancer immunity. This finding suggests the possibility that readily accessible PBMCs can be used as a surrogate tissue to assess the local inflammatory environment surrounding cancers through examination of gene expression profiles.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Nakamura for her invaluable contribution to this study.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
↵3 Unpublished data.
- Received March 10, 2008.
- Revision received September 5, 2008.
- Accepted September 25, 2008.
- ©2008 American Association for Cancer Research.