Nasopharyngeal carcinoma (NPC) is a common and malignant tumor that is prevalent in Southeast Asia and Southern China. NPC could be divided into the well-differentiated, the moderately differentiated, the poorly differentiated and the undifferentiated of NPC according with the histopathology examination in China at present. It is a complex polygenic and multifactorial disease just as many other solid tumors.1 The tumorigenesis of the NPC may be caused through the accumulation of various genetic alterations so that the genome-wide investigation will be required to address crucial problems in its biology.2–4In addition, pathogenesis of NPC is multifactorial and multistage towards advanced NPC, although genetic alterations in each stage have not been fully understood. Primary assessment of an NPC is currently based on the microscopic examination of cells and tissues, and therapeutic decisions are based on a limited set of clinical, histological, and biological characteristics. Such classification system has allowed important advances in cancer treatment, but is not always accurate. Among the rapidly emerging technologies, microarrays allow quantitative and simultaneous measurement of the mRNA expression levels of thousands of genes in a biological sample.5,6Gene expression profiles will permit tumor classification in more homogeneous diagnostic and prognostic groups, as well as the identification of new clinically and biologically relevant tumor subclasses at the molecular level. Transcriptome analysis using microarray is expected to identify novel molecular targets for diagnostics, classification, progression in each stage in carcinogenesis of NPC as the expression levels of such genes are often dys-regulated in NPC tissues.7–9
Seldom studies had hitherto combined microarray and microdissection techniques to study the carcinogenesis mechanism of NPC. However, the cell line related to the NPC as the study material were always been used in many studies,10–15 while the whole NPC biopsy tissues were been used in the other studies.16,20 As NPC typically presents as a solid mass characterized by a highly desmoplastic stromal with usually a low percentage of tumor cells (<40%). In some tissue samples of NPC the lymphocyte comprise more than 95% of total population of cells making up this tumour. This greatly hampers comparative gene expression studies and necessitates selective procurement of both normal and tumor homogeneous cells from within complex tissues.21–23In addition, the bulk of the tissue sample from a single NPC patient was always too less to harvest the enough total RNA for the microarray study, so some studies used the NPC tissue pools by mixing many samples from different patients together as the material of study. This pool included heterogeneous cells so it surely had negative effect on the study.5,9,24
In our study, we took the homogeneous carcinoma cells microdissected from NPC tissue samples and the pure epithelium pillar cells from non-NPC tissue samples of nasopharynx to construct the whole human genome expression profiles.25 We aimed firstly, to show that sufficient aRNA could be obtained from the “in vitro transcription” (IVT) by amplified the total RNA; secondly, to identify different groups of genes that could be used to illuminate the carcinogenesis, the classification and the development of NPC.
The tissue samples were collected from the nasopharynx of the patients hospitalized in the Department of Otolaryngology Head and Neck Surgery, Peking University Third Hospital. Informed consent was obtained from all of the patients. Among 18 nasopharyngeal tissue samples (14 men and 4 women), 13 samples as the case group were from NPC samples and the rest samples were from nasopharyngeal normal or inflammatory mucous as the control group. All of the patients were confirmed by histopathology examination. The age of the NPC patients was from 18–69 (42.9±11.3) years, while the age of patients in the control group was from 21–54 (45.9±19.7) years. As 6 samples were from non-keratinizing squamous-cell carcinoma and the other 6 samples were from undifferentiated carcinoma in the case group, we might further assign these 12 tumor samples into two subgroups according to their histopathology and the criteria of WHO (WHO II: n=6, WHO III: n=6). The other NPC sample was from the keratinizing squamous-cell carcinoma so it should be as the WHO type I. If we used the category of TNM which represent the clinical progression of each NPC sample, we also could classified the NPC samples as the primary subgroup (stages 1 and 2, n=5) and the advanced subgroup (stages 3 and 4, n=7). Each fresh tissue sample from all of the patients was grossly dissected into two parts, the one of that was fixed in formalin for diagnostic analysis and possible immunohistochemical analysis, the other part of tissue sample was washed with 0.9% NaCl solution and immediately submerged them into RNAlater RNA Stabilization Reagent (Qiagen, USA) within 10 minutes. The volume of the RNAlater RNA Stabilization Reagent used to preserve the tissue sample was at least 10 times bigger than that of the tissue block. Then we preserved them at -20°C in refrigerator for microarray analysis.
Microdissection, RNA extraction and amplification
Modifying the ONO-121 Ergonomic Joystick Micromanipulator (Olympus, Japan) as our microdissected device to harvest homogeneous tumor cells from 22 NPC tissue samples and pure nasoparyngeal pillar epidermal cells from the other 10 non-NPC tissue samples which preserved in RNAlater Stabilization Reagent. Then total RNA was isolated from these microdissected homogeneous tissue cells of each samples using Absolutely RNA Microprep Kit (Stratagene, USA) in accordance with the protocols recommended. The quality and approximate yield of RNA was analyzed on an Agilent 2100 Bioanalyzer (Agilent, USA) using about 1 μl of the RNA sample. The stock total RNA was distributed into aliquots and stored at -70°C. Taking 11 μl of this total RNA for the first cycle IVT using the MessageAmp™ aRNA kit (Ambion, USA) following the protocol provided by the corporation. This amplification procedure consists of reverse transcription with an oligo (dT) primer bearing a T7 promoter and IVT of the resulting DNA with T7 RNA polymerase to generate hundreds to thousands of antisense RNA (aRNA) copies from each mRNA. These aRNA were used to access the second IVT for array hybridization and did reverse transcription in accordance with protocols recommended by Affymetrix using commercially available buffers and proteins (Invitrogen, USA). Biotin labelling and about 250-fold linear amplification followed phenol-chloroform clean up of the reverse transcription reaction product and was done by the second IVT over a reaction time of 8 hours.
Hybridization to the GeneChip HG-U133Plus.2.0
We hybridized 15 μg of labelled cRNA onto the GeneChip Human Genome-U133 Plus.2.0 Array (HG-U133.Plus.2.0) using recommended procedures of prehybridisation, hybridisation, washing, and staining with streptavidin-phycoerythrin (SA-PE) for each sample. Antibody amplification was done with a biotin-linked antibody to streptavidin (Vector, USA) with a goat-IgG blocking antibody (Sigma, USA). A second application of the SA-PE dye was used after additional wash steps had been done. After automated staining and wash protocols, the arrays were scanned by the Affymetrix GeneChip Scanner 3000 (Agilent, USA) and quantitated with GeneChip Operating Software (GCOS, Affymetrix, USA). The GeneChip HG-U133.Plus.2.0 consists of about 54 000 probe sets, each probe set containing about 16 perfect match and corresponding mismatch 25mer oligonucleotide probes representing sequences (or genes), most of which have been characterized in terms of function or disease association. These probe sets representing over 38 500 well-characterized human genes that can be used to explore human biology and disease processes. We could access the whole human genome for a global view of the changes in gene expression that occur during NPC progression by using this microarray. The raw, unnormalized probe level data were then analyzed by GCOS for final normalisation and modelling. Median intensity and the perfect match/mismatch modelling (PM/MM) algorithm was used for the normalization of the 18 arrays.
Semi-quantitative reverse transcription polymerase chain reaction (sqRT-PCR)
We did sqRT-PCR measurement of gene expression levels using the same amplified aRNA samples hybridized to the genechips. E2F6 (sequence ID in affymetrix chip was 203957-at) and TSPAN-1 (sequence ID in affymetrix chip was 209114-at ) genes were selected for analysis on the basis of their variation in expression. Primers were designed for these loci with the gene sequences freely available from the NetAffx (www.affymetrix.com) and the Primer3 algorithm for primer design. The primer of E2F6 were GCAAGGATAGAAATGCCATCA (left primer) and TGTGAAAATGTGGGAGCAGA (right primer). The primer of TSPAN-1 gene were ACAATGGCTGAGCA-CTTCCT (left primer) and TTTTTGGTCGTGAGCC-TTTT (right primer). Product sizes were kept short (E2F6 was 209 bp and TSPAN-1 was 255 bp) to allow the maximum ability to work under varying conditions relative to aRNA quality. The samples were taken at alternating cycle numbers between 28 and 32 to ensure that the sqRT-PCR reaction products were in a linear range of accumulation. These samples were then arranged in ascending order, diluted with 10 μl loading buffer, and 10 μl of each sample was loaded onto 3% denaturing acrylamide gels. Electrophoresis at 80 W was done for 1.5 hours, or until sufficient separation of the xylene cyanol and bromophenol blue dyes was achieved. Gels were then fixed, removed from the rear plate, transferred to filter paper, and dried. We first assessed these dry gels using autoradiography (about 6 hours exposure, no intensification), and analysable gels were then exposed to phosphorimaging screens. Primers that failed to produce a single clear band were used again with different annealing temperatures until a single band was produced. The primers chosen proved suitable to use and gave clean, at least a single bands for analysis. Although high-cycle samples inevitably achieved pixel-saturation, care was taken to keep exposure times to a minimum, so as to keep intensity within the informative range on most cycle-totals within each set. To determine the linear range of the primers, we analysed their absolute intensities using Microsoft Excel graphing functions. We then did phosphorimager quantification analysis (Bio-Rad Laboratories, Hercules, CA) and RT-PCR product band intensities were quantitatively compared with normalized, model-based estimates of expression from the data of each GeneChip.
After scanning and low-level quantification using GCOS, we adjusted arrays to a common baseline and estimated expression using PM-MM model.26 We eliminated genes that were not present in at least 1 sample in any group, and exported expression data for the remaining genes to GeneSpring 6.1 for more filtering and analysis. After transforming all data by taking logarithms, we ranked genes by variability over all 18 samples, and we retained all the genes that were significantly more variable than the median variance. We selected differentially expressed genes from the filtered gene list using the two-sample t test, and then with the Gene Ontology (GO) from NetAffx Links to describe the biological process, molecular function, and cellular component for the genes. The Gene Ontology Mining Tool maped GeneChip probe sets to the hierarchical vocabularies. In addition to providing the GO terms for annotated genes, it provided graphical, interactive views of probe set representation within the biological process, molecular function, or cellular component hierarchies. These graphs allowed us to visualize input probe lists in the context of biological information and to visually determine the relationships among probe sets based on their locations in the GO graph. As the graph is hierarchical, we could view their results within a context of broad or detailed GO categories. To find the top differentiation expression marker genes between different groups, a global permutation test as an overall, multiple comparison-free test of whether the number of differentially expressed genes exceeded that which might arise by chance with using the software of GenePattern. The signal-to-noise gene selection method looked at the difference of the means in each of the classes (or samples), but scaled by the sum of the standard deviations: (m1-m2) / (s1+s2) (m1 was the mean of class 1 and s1 was the standard deviation of class 1). If the denominator were (s12+s22)/2 and m1 was the mean then this would be the statistic used in the t test. In this test, the observed number of significantly differentially expressed genes was compared with the distribution of numbers of differentially expressed genes generated by repeatedly permutating the labels of the samples and recalculating the t test at the specified level of significance. A independent-samples t test or one-way ANOVA was used to analyze the difference rate of E2F6/GAPDH and TSPAN-1/GAPDH expression between the case group and the control group respectively. All of the result of sqRT-PCR in the study analysis were performed with SPSS 12.0. AP value less than 0.05 was considered statistically significant.
Nasopharyngeal carcinogenesis genes
Eighteen patients were included in the study, and their clinical characteristics were shown in Table 1. Before doing gene expression analysis, all of the 18 samples were divided into two groups (NPC groups and non-NPC groups) according with the histopathology examination, and those NPC samples could be further classified by the criteria of WHO into three categories: WHO type I, WHO type II and WHO type III.
To select discriminatory genes, firstly, with two-way hierarchical clustering analysis (Pearson correlation), we compared expression data in the NPC samples (NPC, n=13) and that in the non-NPC samples (normal, n=5). We selected a subset of candidate genes by filtering on signal intensity to eliminate genes with uniformly low expression or genes whose expression did not vary significantly across the samples. After log transformation, a t test was used to select discriminatory genes. One hundred and twenty-seven genes were selected by t tests with nominal P values of 0.001 for which expression differed in NPC and non-NPC groups, or differentially expressed. The probability that these numbers of genes would be selected by chance and the differentially expressed fold between the case groups and the control groups. One hundred and twenty seven genes would be selected by a P value less than 0.001 with the 2-fold differentiated expression between the two groups (Figure 1).
Secondly, we compared gene expression data in the GeneChip. Eight differentiated expression genes were selected from the filtered 127 carcinogenesis related genes by t tests with P values of 0.01 (Figure 2).
Functional classification of discriminatory genes
The 127 genes classed as the most significantly “differentially expressed gene” between the NPC-group and the non-NPC group at P=0.001 with at least 2-fold differentially expressed were listed in Figure 1. These genes showed 2.00–23.18 fold decrease or 2.00–6.55 fold increase in expression in NPC samples compared with that in non-NPC samples. Using the NetAffx Gene Ontology (GO) Mining Tool classed of these differentially expressed genes included cellular component 61 genes, molecular function 71 genes and biological process 79 genes. Function classes of these differentially expressed genes involved 89 genes in apoptosis, cell adhesion or cytoskeleton, protein transport, signal transduction, RNA transcription, RNA splicing or transport, cell cycle, and protein translation; the remainder 36 genes have unknown functions. The distance and the value of permutation testing of each genes were listed. The 8 top most up-regulated and down-regulated genes between the two groups were hierarchical clustered in Figure 2.
In this cross-validation analysis, we began with all 127 filtered genes to avoid selection bias. Every observation sample in turn was left out and the remaining samples were used to select differentially expressed genes; we then constructed a compound covariate predictor to classify the left-out sample. Twelve of 13 NPC samples and 4 of 5 non-NPC samples were correctly classified, for an overall accuracy of 88.9%. Results of permutation testing showed that such a high cross-validated classification accuracy is significant (P <0.01).
Confirmation of expression measurements
To confirm measurement of RNA concentrations, expression values derived from adjusted affymetrix chip data were correlated with values from sqRT-PCR for 2 variably expressed genes (Table 2). Pearson correlations were significantly positive for all of the 2 genes (Table 3).
To investigate crucial genetic alterations in the tumorigenesis of the NPC, we applied high density oligonucleotide arrays containing up to 48 000 genes to monitor genome-wide gene expression in the 13 microdissected NPC tissue samples and that in the 5 microdissected non-NPC tissue samples from mucous membrane in nasopharynx. Correlation of expression profile data was investigated. Furthermore, we could divided the 13 NPC samples into three groups according with the histopathology by the criteria of WHO (WHO type I was a keratinizing squamous-cell carcinoma similar to carcinomas that arise from other sites of the head and neck. WHO type II was a non-keratinizing epidermoid carcinoma while the WHO type III represented the undifferentiated carcinoma, they were also refered to as lymphoepithelioma or Shminke tumors. The WHO type II was the common form of the disease in China), there were 6 samples belong to WHO type II, 6 samples to WHO type III, and only one sample to WHO type I in all of the 13 NPC samples which were used to hybridized to the GeneChip in our study. We could also classified all of the NPC samples into the clinical stages 1 and 2 (or the primary stage of NPC) and the clinical stages 3 and 4 (or the advanced stage of NPC) according with the progression of the disease in each patient using the criteria of TNM.
We obtained high quality RNA from each microdissected homogeneous tissue sample of NPC and non-NPC biopsy tissue, and then the RNA was used in the line amplification for driving the sufficient aRNA. From hybridizating them on the GeneChip to assess patterns of gene expression in individual tumors and identify molecular profiles using gene expression patterns of human NPC to accurately predict the carcinogenesis of NPC and to find the oncogenes or the tumor suppressor genes of NPC, to find classified genes between different subgroups or to find the genes for predicting the progress of NPC.
Gene expression patterns associated with the carcinogenesis of NPC are highly complex. In the past, investigators using some microarray which held seldom genes on it to filtering single gene biomarkers to assess the carcinogenesis of NPC have seldom produced conclusive results. There is little information about the usefulness of gene expression arrays in human NPC.9,11,16–19,27 For example, the researchers did not care many correlation between commonly measured the carcinogenesis of NPC and predictive and prognostic markers in the NPC progression. So we chose oligonucleotide microarray which could provide us with the whole human genome expression profile of over 48 000 genes. And our results support the idea that assessment of expression of a few individual genes together will be powerful enough to untangle the heterogeneity of clinical NPCs. The patterns of expression of many genes could be successful in distinguishing between the NPC samples and the non-NPC samples, between the samples of WHO type II and the samples of WHO type III or across in the different progression subgroups in the clinical of NPC. A key point of our study was to focus on genes that could be reliably measured and to exclude those that were unlikely to be expressed in any sample. As a result, our analysis will have excluded some differential genes with low expression, some of which might be biologically interesting.
Although NPC were highly heterogeneous, the differentially expressed gene list gave some clues to the mechanisms of NPC tumorigenesis. In general, we selected 8 top most up-regulated and down-regulated genes (included the gene of E2F6, CTPS, TSPAN-1, DSG3 and so on) between NPC and non-NPC groups. Overexpressed genes in NPC associated with tRNA lyase activity, protein-nucleus import, intracellular protein transport, protein folding, cell cycle, cell proliferation, apoptosis, nucleic acid metabolism, nucleotide biosynthesis, immune response, pyrimidine metabolism, nucleotide-sugar metabolism, CTP synthase activity, catalytic activity, co-chaperonin activity, receptor activity, DNA replication, DNA binding, ATP binding, electron transport, and regulation of transcription functions, on cellular component. Description these genes constructed as the component of cytoplasm, nucleoplasm, nulelus, integral to membrane, cytosol microtubule, and cytoplasm respectively in the NPC samples, whereas downexpressed genes in NPC involved in cell adhesion, cell motility, cell proliferation, proteolysis and peptidolysis, phagocytosis, signal transduction, calcium ion binding, protein binding, chemotaxis inflammatory response, iron ion transport, chemokine activity, and serine-type endopeptidase inhibitor activity, integral to membrane, plasma membrane or action for extracellular space soluble fraction.
However, a challenge was whether it will be quantitative enough to detect expression discrimination across the different groups although we had shown that expression values obtained by oligonucleotide arrays were correlated fairly well. To detect such genome-wide alterations in the RNA copy number, sqRT-PCR has been applied to all of the 18 samples. The expression difference of 2 genes was confirmed by the result of sqRT-PCR compared it with the GeneChip. Although the validation set was very small, it does lend support to the suggestion that top marker genes expression could be used to predict or diagnose the NPC. This study also showed that expression value of some genes (such as the genes included in Figure 2) could effectively classify tumours according to the type of molecular profiling implicated in defining the new clinical categories for an individual patient. Surely, our results should be validated in a study with a large independent cohort of patients. Further patient recruitment and analysis would refine the top marker gene list by which to diagnose or classify NPC.
We thank the other colleagues in the Cancer Institute of Central South University for their help during the procession of experiment.
1. Vokes EE, Liebowitz DN, Weichselbaum RR. Nasopharyngeal carcinoma
. Lancet 1997; 350: 1087-1091.
2. Zhang S, Wu Y, Zeng Y, Zech L, Klein G. Cytogenetic studies on an epithelial cell line derived from nasopharyngeal carcinoma
. Hereditas 1982; 97: 23-28.
3. Sizhong Z, Xiukung G, Yi Z. Cytogenetic studies on an epithelial cell line derived from poorly differentiated nasopharyngeal carcinoma
. Int J Cancer 1983; 31: 587-590.
4. Chan SH. Aetiology of nasopharyngeal carcinoma
. Ann Acad Med Singapore 1990; 19: 201-207.
5. Lee JY, Dong SM, Kim SY, Yoo NJ, Lee SH, Park WS. A simple, precise and economical microdissection technique for analysis of genomic DNA from archival tissue sections. Virchows Arch 1998; 433: 305-309.
6. Xie XJ. On multiple testing, validation of gene expression profiling
, and translational research. Chin Med J 2008; 121: 1247-1248.
7. William D. Custom cDNA microarrays; technologies and application. Toxicology 2002; 181: 551-554.
8. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286: 531-537.
9. Xie L, Xu L, He Z, Zhou W, Wang L, Zhang L, et al. Identification of differentially expressed genes in nasopharyngeal carcinoma
by means of the Atlas human cancer cDNA expression array. J Cancer Res Clin Oncol 2000; 126: 400-406.
10. Lin CT, Chan WY, Chen W, Huang HM, Wu HC, Hsu MM, et al. Characterization of seven newly established nasopharyngeal carcinoma
cell lines. Lab Invest 1993; 68: 716-727.
11. Zhang B, Nie X, Xiao B, Xiang J, Shen S, Gong J, et al. Identification of tissue-specific genes in nasopharyngeal epithelial tissue and differentially expressed genes in nasopharyngeal carcinoma
by suppression subtractive hybridization and cDNA microarray. Genes Chromosomes Cancer 2003; 38: 80-90.
12. Li H, Han WN, Feng XL, Zhou W, Yao KT. Analysis of characteristics of gene expression in pericancerous stromal cells of nasopharyngeal carcinoma
by cDNA array. Chin J Cancer (Chin) 2003; 22: 235-238.
13. Zhang L, Song L, Ma Y, Huang B, Liang Q, Zeng Y. Differentially expressed gene in nasopharyngeal carcinoma
cell lines with various metastatic potentialities. Chin J Oncol (Chin) 2002; 24: 430-434.
14. Ma J, Li J, Zhou J, Li XL, Tang K, Zhou M, et al. Profiling genes differentially expressed in NGX6 overexpressed nasopharyngeal carcinoma
cells by cDNA array. J Cancer Res Clin Oncol 2002; 128: 683-690.
15. Guo X, Lui WO, Qian CN, Chen JD, Gray SG, Rhodes D, et al. Identifying cancer-related genes in nasopharyngeal carcinoma
cell lines using DNA and mRNA expression profiling
analyses. Int J Oncol 2002; 21: 1197-1204.
16. Shao JY, Zeng WF, Zeng YX. Molecular genetic progression on nasopharyngeal carcinoma
. Chin J Cancer (Chin) 2002; 21: 1-10.
17. Zhang LJ, Fang Y, Ma YH, Yan J, Zhang XS, Huang BJ, et al. Gene expression profiling
in nasopharyngeal carcinoma
determined by high density cDNA microarray. Chin J Cancer (Chin) 2002; 21: 588-592.
18. Hong MH, Xiao XB, Mai HQ, Cao SM, Min HQ. Analysis of gene expression patterns of periphery lymphocytes in patients with nasopharyngeal carcinoma
. Chin J Cancer (Chin) 2002; 21:21-24.
19. Han W, Li H, Xie L, Xu L, Zhang L, Yao K. Plerosis of cDNA array of normal human nasopharyngeal tissue and nasopharyngeal carcinoma
. Chin J Oncol (Chin) 2002; 24: 114-117.
20. Li H, Han W, Zhang L. cDNA expression array in the differential expression profiles of p53 regulated genes in nasopharyngeal carcinoma
and the human normal nasopharynx. Chin J Oncol (Chin) 2001; 23: 448-450.
21. Shibata D, Hawes D, Li ZH, Hernandez AM, Spruck CH, Nichols PW. Specific genetic analysis of microscopic tissue after selective ultraviolet radiation fractionation and the polymerase chain reaction. Am J Pathol 1992; 141: 539-543.
22. Shibata D. Selective ultraviolet radiation fractionation and polymerase chain reaction analysis of genetic alterations. Am J Pathol 1993; 143: 1523-1526.
23. Lee JY, Dong SM, Kim SY, Yoo NJ, Lee SH, Park WS. A simple,precise and economical microdissection technique for analysis of genomic DNA from archival tissue sections. Virchows Arch 1998; 433: 305-309.
24. Shibata D, Hawes D, Li ZH, Hernandez AM, Spruck CH, Nichols PW. Specific genetic analysis of microscopic tissue after selective ultraviolet radiation fractionation and the polymerase chain reaction. Am J Pathol 1992; 141: 539-543.
25. Liu ZQ, Tian YQ, Peng C, Hu YF, Zhou M, Ouyang J, et al. Expression of NASG gene and its role in human nasopharyngeal homogenous tissue cells. Chin Med J 2005; 118: 1076-1080.
26. Li C, Hung WW. Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol 2001; 2: 1-11.
27. Zeng ZY, Zhou YH, Zhang W, Fan SQ, Li XL, Luo XM, et al. Gene expression profiling
of nasopharyngeal carcinoma
reveals the abnormally regulated Wnt signaling pathway. Hum Pathol 2007; 389: 120-133.
Keywords:© 2009 Chinese Medical Association
nasopharyngeal carcinoma; expression profiling; molecular marker