Cancer Research CTRC-AACR San Antonio Breast Cancer Symposium Joint Metastasis Research Society-AACR Conference on Metastasis
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
Cancer Research 67, 474-481, January 15, 2007. doi: 10.1158/0008-5472.CAN-06-1882
© 2007 American Association for Cancer Research
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pegtel, D. M.
Right arrow Articles by Thorley-Lawson, D. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pegtel, D. M.
Right arrow Articles by Thorley-Lawson, D. A.

Molecular Biology, Pathobiology, and Genetics

IFN-{alpha}–Stimulated Genes and Epstein-Barr Virus Gene Expression Distinguish WHO Type II and III Nasopharyngeal Carcinomas

D. Michiel Pegtel1, Aravind Subramanian2, David Meritt1, Ching-Hwa Tsai3, Tzung-Shiahn Sheen4, Todd R. Golub2 and David A. Thorley-Lawson1

1 Department of Pathology, Tufts University School of Medicine, Boston, Massachusetts; 2 The Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, Massachusetts; and 3 Graduate Institute of Microbiology and 4 Department of Otolaryngology, National Taiwan University Hospital, College of Medicine, National Taiwan University, Taipei, Taiwan

Requests for reprints: David A. Thorley-Lawson, Department of Pathology, Tufts University School of Medicine, Jaharis Building, 150 Harrison Avenue, Boston, MA 02111. Phone: 617-636-2726; Fax: 617-636-2990; E-mail: david.thorley-lawson{at}tufts.edu.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Nonkeratinizing nasopharyngeal carcinoma (NPC) is 100% associated with Epstein-Barr Virus (EBV) and divided into two subtypes (WHO types II and III) based on histology. We tested whether these subtypes can be distinguished at the molecular genetic level using an algorithm that analyzes sets of related genes (gene set enrichment analysis). We found that a class of IFN-stimulated genes (ISG), frequently associated with the antiviral response, was significantly activated in type III versus type II NPC. Consistent with this, replication of the endogenous EBV was suppressed in type III. A strong association was also seen with a subset of ISGs previously identified in systemic lupus erythematosus, another disease in which ‘normal’ EBV biology is deregulated, suggesting that this pattern of ISG expression may be linked to the increased EBV activity in both diseases. In contrast, unsupervised hierarchical clustering of the complete expression profiles failed to distinguish the two subsets. These results suggest that type II and III NPC have not originated from obviously distinct epithelial precursors; rather, the histologic differences may be a consequence of a differential antiviral response, involving IFNs, to chronic EBV infection. [Cancer Res 2007;67(2):474–81]


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Nasopharyngeal carcinomas (NPC) are notoriously difficult to diagnose. This is due to several factors, including the inaccessibility of the nasopharynx, nonspecificity of symptoms when the disease is in its early stages, and the tendency of the tumor to spread submucosally, and thus remains clinically unapparent even to direct inspection. NPCs are usually subclassified according to histologic criteria formulated by the WHO into undifferentiated (type III), ‘poorly’ differentiated or nonkeratinizing (type II), and keratinizing (type I) subtypes (1). Poorly differentiated NPCs have a pavimented arrangement and well-defined cell margins, whereas undifferentiated tumors exhibit a syncytial appearance and indistinct cell margins (2). Electron microscopic features indicate that NPCs are derived from the basal layers of pseudoepithelia and/or stratified epithelia (3). The current WHO classification suggests that type II and III tumors may have originated from distinct epithelial precursors.

Concern has been raised that the classification of nonkeratinizing NPCs into type II and III is subjective and that the nomenclature is confusing. The histologic differences might only reflect the naturally occurring heterogeneity of NPC tumors in general rather than in a different etiology. In fact, undifferentiated NPCs are heterogeneous tumors, which sometimes contain small foci of cells with evidence of squamous differentiation within a largely undifferentiated tumor mass (4). Due to concerns about WHO classification, alternate classification systems have been proposed. For instance the ‘Cologne classification’ divides NPC into two categories: squamous cell carcinoma (WHO type I) and undifferentiated carcinoma of nasopharyngeal type consisting of WHO types II and III (5). This classification suggests that the latter is derived from similar epitheloid precursors and has a common etiology.

The most striking etiologic factor that nonkeratinizing (types II and III) NPCs have in common is the 100% association with Epstein-Barr Virus (EBV) infection (4), whereas the EBV association with keratinizing NPCs is less consistent. Nonkeratinizing NPCs can be subdivided based on expression of the viral oncogene latent membrane protein (LMP) 1, whereas all express LMP2A (6), which has prosurvival functions, at least in B cells, and may promote NPC metastasis (7–9). Despite these advances, the origins of NPC remain relatively unclear. It is unknown, for example, if one specific epithelial cell type gives rise to all nonkeratinizing NPCs or if they derive from distinct precursors. Similarly, the exact role of EBV infection is not yet identified. For instance, it is unknown if type II and III NPCs have similar or different EBV etiologies. An important step forward was the discovery of EBV in healthy nasopharyngeal epithelial cells (10, 11), indicating that EBV may be present at the initial stages of NPC development. However, NPCs originate from translational and crypt epithelial from the lateral wall of the nasopharynx (fossa of Rosenmuller). It is unknown if healthy cells from this region also harbor EBV.

Some nonkeratinizing NPCs (mostly type III) are also recognized for their prominent infiltrating stroma consisting predominantly of lymphocytes, which may be important for tumor growth (12). In fact, historically, NPCs were referred to as ‘lymphoepitheliomas’ (13). The exact relationship between tumor subtypes and the origin of the infiltrating lymphoid stroma has not yet been elucidated, although some have speculated that the consistent presence of EBV might play a role (14).

Here, we have tested whether NPC subtypes can be distinguished by their molecular gene expression profiles. The prediction was that if nonkeratinizing NPC subtypes have a distinct etiology, this should be apparent at the molecular genetic level (15).


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Biopsy Material/Chip Hybridization
Total RNA was isolated from 20 fresh nonkeratinizing NPC biopsies from Taiwan obtained with punch forceps using endoscopic guidance. Tumor stage was designated according to the International Union Against Cancer and American Joint Committee on Cancer staging manuals (16). Histopathology was based on the WHO international histological classification (1, 17). NPCs were mostly of the advanced stage (IVA-C), and all samples in this study were EBV positive by EBV-encoded small RNA (EBER) expression. Biopsy RNA that met the quality criteria for hybridization (n = 16) was hybridized to Affymetrix microarray chips (HG-U95A). There was only sufficient material for one hybridization per biopsy; therefore, duplicates were not done. All scans were done on Affymetrix scanners, and the expression value for each gene was calculated using Affymetrix GeneChip MAS5 software.5 Briefly, the raw expression data as obtained from the Affymetrix GeneChips were scaled to account for differences in chip intensities. We calculated the mean expression level (Ex) for all genes on each array. All scans within an experiment were scaled to the array with the median Ex value (all expression values are multiplied by Exmedian/Ex). In total, 14 individual arrays of 8 WHO type III and 6 WHO type II met quality criteria and were used for the computational analysis. The array expression values were validated by quantitative reverse transcription-PCR (RT-PCR) for several genes, including the cytoskeleton gene keratin 6 and the transcription factor c-fos, as described previously (9).

Computational Analysis
Gene set enrichment analysis. Gene set enrichment analysis (GSEA) calculates an enrichment score (ES) that reflects the degree to which a set of genes S is overrepresented at the extremes (top or bottom) of the entire ranked list L. The score is calculated by walking down the list L, increasing a running sum statistic when a gene in S is encountered, and decreasing it when a gene not in S is encountered. The magnitude of the increment depends on the correlation of the gene with the phenotype. The ES is the maximum deviation from zero encountered in the random walk; it corresponds to a weighted "Kolmogorov-Smirnov-like" statistic. The significance of an observed ES is obtained by permutation testing: reshuffling the phenotype labels and resorting the gene list to determine how often an observed ES (G) occurs by chance. Statistical significance [nominal P value (NP)] is computed by comparing the observed ES (G) with a histogram of ES (G, {pi}) values corresponding to the enrichment of the same gene set G but with reshuffled data according to a set of permulations {pi} = (1,...,{Pi}). When many gene sets are tested, the value ES is normalized to adjust for variation in gene set size—the normalized ES (NES).6

Marker selection. Marker selection is described by Golub et al. (18). The signal-to-noise (SNR) statistic was used to rank genes and select marker sets. For each gene, j, the SNR statistic is computed as Formula where µ+ (j), µ–(j), {sigma}+(j), and {sigma}–(j) are the mean and SD of the two classes labeled ±1, respectively. Genes with the most positive values are correlated with class +1 and those with the most negative value are most correlated with class –1. A rank-ordered gene list is constructed by ranking the SNR statistic for each gene. A permutation test based marker analysis was used to select for candidate genes whose expression significantly correlated with WHO type II and III histology. Briefly, SNR scores were generated for all genes using class labels and then sorted. The best match (k = 1) is the gene "closer" or more correlated to the class +1 using the SNR as a correlation function. Random permutations (1,000) of the class labels were then generated, and for each case of these, we generated SNR scores and sorted genes. A histogram of SNR scores was then built for each value of k. For example, build one for the 1,000 top markers (k = 1), another one for the 1,000 second best (k = 2), etc. These histograms represent a reference statistic for the best match, second best, etc., where many genes contribute to a given value of k. For each value of k, determine different percentiles (1%, 5%, 10%, 50%, etc.) of the corresponding histogram. Compare the actual SNR scores with the different significance levels obtained for the histograms of permuted class labels for each value of k. This procedure is implemented in the statistical software package GeneCluster2, which was used for the analysis.

Hierarchical clustering. Average linking hierarchical clustering was done with the program Cluster (19) on the ~4,851 genes that survived normalization and filtering criteria (thresholded to 20 and removal of genes with a variation under 3-fold).

Quantitative RT-PCR
cDNA, in a final volume of 200 µL, was prepared from 7 µL of isolated cell line RNA and 10 µg of tumor biopsy RNA as described previously (9). cDNA from 16 NPC samples (7 type II and 9 type III) was used for quantitative RT-PCR studies. The EBV-positive Burkitt lymphoma cell line Akata [EBV positive; American Type Culture Collection (ATCC) (Manassas, VA)], B-cell lymphoma cell line BJAB (EBV-negative control; ATCC), marmoset cell line B9-58 (EBV positive), lymphoblastoid (LCL) cell line IB4 (EBV positive; gifts of E. Kieff, Brigham and Women's Hospital, Boston, MA), and an LCL from primary B cells were used as standards for quantitative RT-PCR. Quantitative RT-PCR was done with either the Syber Green or Taqman (PE Applied Biosystems, Foster City, CA) method. For Syber Green, PCRs were done in 20 µL on the ABI Prism 5700 sequence detection system (Perkin-Elmer, Wellesley, MA). For each, 10 µL of Universal Syber Green MasterMix were prepared (PE Applied Biosystems) to which 5 µL of NPC biopsy cDNA sample (or standard cDNA) were added. PCRs were designed to avoid nonspecific product and primer dimer formation by analyzing the dissociation curve using GeneAmp 5700 software version 1.3. Cycle conditions, MgCl2 (1–3 mmol/L), and primer concentrations (100–450 nmol/L) were optimized for each individual PCR. Final products were analyzed by gel electrophoresis to confirm the correct product size. Typically, the melting temperature of the specific PCR product was between 70°C and 85°C. Taqman PCRs (IFN-ß, IFN-{gamma}, and ß-actin) were done with Assays-on-Demand (PE Applied Biosystems) using 10 µL primer probe MasterMix and 10 µL cDNA. The cycle run was 5 min at 95°C and 10 min at 50°C followed by 40 cycles of 60°C for 1 min and 95°C for 15 s. EBER transcripts were analyzed by a custom-designed Taqman procedure described in detail previously (20). The Syber Green method was used for the lytic genes Z, Ead, VCA, and IFN-{alpha}. The amplimers for the EBV genes BZLF1 (Z; ref. 21), BHRF1 (Ead; ref. 22), BcLF1 (VCA; ref. 23), LMP1, EBNA1 (Q-K), and EBER (9) were described previously. PCR conditions for EBV lytic genes were 95°C for 15 s, 59°C for 30 s, and 72°C for 1 min repeated for 40 cycles. PCR was initiated by a first step of 10 min at 50°C followed by 5 min at 95°C. IFN-{alpha} PCR was done with universal amplimers (200 nmol/L), which detect all isoforms (24) 5'-TCCATGAGATGATCCAGCAG-3' and 5'-ATTTCTGCTCTGACAACCTCCC-3' (the product size is 274 bp). Cycle conditions were 95°C for 30 s, 58°C for 1 min, 72°C for 1 min, and 79°C for 10 s repeated for 40 cycles. The last step was included to avoid signal detection of nonspecific products.

Sensitivity and quantification of the PCR assays were estimated by comparison with serially diluted cDNA generated from the standard cell lines. We used IB4 or LCL cells as standard for EBV-encoded latent genes. Replication-induced Akata and B9-58 were used as positive controls and standard for EBV lytic cycle genes. To induce lytic gene expression, Akata cells were treated for 18 h with antihuman immunoglobulin G (IgG; 100 µg/mL; Sigma, St. Louis, MO). IgG-treated Akata cells stained positive for Ead, Z, and VCA protein were analyzed by indirect immunofluorescence staining (data not shown). All PCR values were normalized to corresponding ß-actin levels. Typically, we were able to detect specific EBV gene products at the single-cell (average transcript expression of EBV-encoded genes in single IB4 or Akata cells) level with <40 amplification cycles.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
WHO classification is associated with differential IFN-stimulated gene expression. Although WHO type I tumors are distinct in that they show clear evidence of keratinization, the molecular differences, if any, between type II and III are unclear. We tested if nonkeratinizing NPCs can be classified molecularly. To this end, we determined the expression profiles of 14 nonkeratinizing NPCs (type II, n = 6; type III, n = 8) with Affymetrix microarrays containing ~12,600 genes. We then did unsupervised hierarchical clustering to see if partitioning of the samples would transpire based on WHO types II and III. Hierarchical clustering is a computational algorithm that produces a relatedness dendrogram of the samples based on the similarity or dissimilarity of their gene expression profiles. Figure 1 shows that types II (Fig. 1, 2) and III (Fig. 1, 3) are randomly distributed throughout the dendrogram, indicating that no dominant gene structure was detected for either subtype.


Figure 1
View larger version (5K):
[in this window]
[in a new window]

  		Figure 1. Hierarchical clustering analysis with WHO type II and III NPCs. Dendrogram result from unsupervised average linking hierarchical clustering using the program ‘Cluster’ on 4,851 expressed genes that survived filtering and normalization criteria (see Materials and Methods) in 14 nonkeratinizing NPC biopsies, WHO type II (2) and type III (3). The two main branches of the dendrogram contain NPC tumors of both subtypes, suggesting that no dominant gene structure is associated with either subclass.

 
In an attempt to identify NPC subtype markers that distinguish type II and III tumors, we looked for differentially expressed genes with a commonly used marker analysis algorithm that uses the SNR statistic to rank genes and then 1,000 permutations of the data set to assign significance (19). Consistent with the outcome of the hierarchical clustering analysis, we found very few genetic markers that correlated significantly with WHO histology and no single genes were statistically significant at the 1% level. However, we noticed a large number of biologically related genes, IFN-stimulated genes (ISG), which were differentially expressed between type II and III tumors. ISGs are recognized as cellular response genes to IFNs often expressed on a viral infection (25). Among the top 100 markers overexpressed in type III, NPCs were >20 ISGs (>20%), including IFIT4, GBP1, OASL, BST2, IFIT2, DDX48, and IP-30. In all cases, the differences in expression level were not statistically significant after permutation testing at the single gene level (see Fig. 3 for example), and surprisingly, we detected no difference in the expression level of the IFN genes (IFN-{alpha}, IFN-ß, or IFN-{gamma}) themselves.


Figure 3
View larger version (19K):
[in this window]
[in a new window]

  		Figure 3. EBV-inducible and antiviral ISGs are systematically overexpressed in NPC type III compared with type II. Expression results of the microarray analysis of individual ISG genes in 14 NPC biopsies. A, histogram representing average relative expression for individual antiviral ISG genes in type III (black columns) and type II (gray columns). B, histogram representing average relative expression for individual ISG genes in type III (black columns) and type II (gray columns) previously reported to be associated with EBV infection (37). For comparison purposes, two housekeeping genes are also plotted, which, as expected, are equally expressed in both subtypes. Bars, SD. Note the relatively high expression levels of the EBV-associated ISGs, such as IFITM3 and IP-30 (two separate probes). Note that, in a control experiment (data not shown) on untreated HONE-1 cells (an EBV-negative epithelial cell line of NPC origin), the levels of expression of these ISGs were at least 10-fold lower than either type II or type III NPC when normalized to actin. Thus, ISG expression is elevated in both but more so in type III. Although no single gene achieves statistical significance, when considered together, a pair wise t test yields a P value of <0.01 for the data in (A) and (B).

 
GSEA confirms the association of ISGs with type III versus type II NPC. Although not significant on their own, coordinate changes in the expression level of multiple genes with a shared biology can be highly significant and potentially reveal a key component of the underlying biology (26). GSEA provides a way to exploit this (27). GSEA is a general statistical method to test for the enrichment of sets of genes in expression data and has been particularly useful in identifying molecular pathways at play in complex gene expression signatures. GSEA considers a priori–defined gene sets [e.g., genes in the same functional class (in this case ISGs) or members of a pathway]. It then provides a method to determine whether the members of these sets are overrepresented at the top (or bottom) of a gene list of markers, which have been ordered by their correlation with a specific phenotype (in this case expression level in type II versus type III NPC), and produces a gene set-gene list specific ES. The significance of this score (NP) is obtained by permutation testing. This is particularly useful for data sets with relatively small sample sizes where high dimensions may lead to stochastic overrepresentation of particular genes "large N small P paradigm" (28). We used GSEA to test the likelihood that differential ISG expression between the subtypes of NPC occurred by chance. We used an independently created ISG set with ~100 genes known to be responsive to type I and II IFNs (29). Sixty-four of these genes were represented on the Affymetrix arrays used in our study. We found that this ISG set was enriched in type III NPC. Permutation analysis indicated that such a distribution is unlikely to occur by chance (ES, 0.61; NP, 0.01; Fig. 2A ) and therefore has biological significance. In addition, a subset of ISGs with known antiviral properties (GBP, OAS, and MX genes) is strongly associated with tumor class. In conclusion, we found that differential expression of genes belonging to the ISG family provides a molecular signature that distinguishes WHO type II and III NPC.


Figure 2
View larger version (31K):
[in this window]
[in a new window]

  		Figure 2. GSEA shows differential expression of ISGs between NPC types II and III. A, GSEA (27) was done with a list of ISGs compiled by Silverman et al. (29). Of this set, 64 were represented on the Affymetrix (HG-U95Av2) chips used in this study. After filtering and normalization, 5,000 genes (L) are ranked on the X axis (‘rank in ordered data set’) of the graph from left to right. Genes on the left are up-regulated in WHO type III (‘labels’), whereas the genes ranked to the right are up-regulated in WHO type II. Green line, increasing (Y axis) with a maximum ES at ~0.6; black vertical lines, position of individual genes (n = 64) from the ISG gene set in the ordered list of ~5,000 genes. To derive significance, the data set was permuted 1,000 times and random ES were calculated. None of the random ES exceeded the observed ES; thus, the P value was <0.001 for this subset of ISGs and highly significant, indicating that the observed distribution is unlikely due to chance. B, as (A) but using a gene set of 28 ISGs associated with the disease SLE derived from Bennet et al. (31). Most genes (22 of 28) of the SLE/ISG gene set (black lines) appear relatively early (left) in the ordered data set and therefore contribute strongly to the ES (ES, ~0.75). This indicates that the genes in this particular ISG gene set are highly associated (P < 0.001) with the tumor subtype.

 
The Silverman ISG data set (29) was identified by comparing gene expression in fibrosarcoma cells before and after addition of recombinant IFNs in vitro. It is possible, however, that some of these ISGs are cell type specific (30). Recently, a different ISG expression signature (up-regulation) associated with peripheral blood mononuclear cells from systemic lupus erythematosus (SLE) patients has been described (31). EBV infection is disrupted in SLE, and SLE, like NPC, has frequently been linked with EBV infection (32, 33). Unlike NPC, the role of EBV in SLE, if any, remains unclear. Nevertheless, we felt that a set of ISGs associated with human disease and EBV might better reflect the in vivo situation with NPC than the general gene set identified in ref. 29. GSEA analysis with 28 of the SLE-associated ISGs confirmed this idea. Not only were they associated with type III tumors but the association was highly significant (ES, 0.76; NP, <0.001; Fig. 2B). Thus, in addition to a general list of ISGs, the expression levels of a subset of human disease-associated ISGs markedly distinguish between type II and III tumors.

For the analysis described above, we intentionally picked ISGs based on a subjective assessment of the Affymetrix data. A completely independent verification of our findings requires that we ask how well correlated with type III NPC are ISG sets compared with other well-defined gene sets. The Molecular Signature Database (MSigDB) contains ~1,500 such gene sets7 associated with specific signals, including several ISG sets. To objectively evaluate our findings with the ISGs, we tested our data against the MSigDB collection of gene sets. Of the ~1,500 gene sets, ~600 showed some degree of up-regulation in type III versus type II NPC with ~200 being significant at the 0.05 level, including all 15 of the ISG sets. Most striking, however, was that the most highly correlated gene set was an ISG and 5 of the top 10 were ISGs of which 4 were induced by IFN-{alpha} (Table 1 ; Supplementary Data), thus validating the outcome of the original GSEA analysis.


View this table:
[in this window]
[in a new window]

  		Table 1. The top 10 gene sets from the MSigDB that most highly correlate with type III NPC

 
IFNs are not differentially expressed at the mRNA level. Several biological mechanisms could explain the expression of ISGs in NPCs. The most obvious is that they responded to IFNs produced by the tumor cells and/or the infiltrating stroma possibly in response to EBV infection. The microarray analysis indicated that the level of IFN gene transcripts at the tumor site was very low (near the detection limit; see Supplementary Data). This suggests that neither the tumor cells nor the infiltrating stroma is a likely source for the IFNs that triggered ISG expression. However, not all isoforms of IFNs are represented on the gene arrays. In particular, the IFN-{alpha} gene has 26 isoforms (34) of which only 8 are arrayed on the U95 GeneChips. In addition, the microarray chips might not be sensitive enough to detect small differences at low expression levels.

For these reasons, we analyzed the expression of all three IFNs and their isoforms individually by sensitive quantitative RT-PCR. We could detect IFN-{gamma} transcripts but only weak expression of IFN-{alpha} (and isotypes) and IFN-ß in the biopsies. In addition, we found no significant association of IFN-{alpha}, IFN-ß, or IFN-{gamma} with WHO type (P = 0.36, 0.24, and 0.18, respectively), confirming the microarray data (Table 2 ). These results suggest that differential ISG expression in the tumors may not be the result of IFN production by either the tumor cells or the infiltrating stroma. This conclusion carries the caveat that it was only shown at the RNA level because we did not have sufficient material available to confirm it at the protein level. Alternative explanations are that local production of IFNs in close proximity to the tumors might have resulted in ISG expression or cells in the tumor responded to IFNs produced early during a viral infection or to IFN-independent mechanisms (35).


View this table:
[in this window]
[in a new window]

  		Table 2. RNA expression levels of IFN genes in type II and III NPC

 
Expression levels of the EBV-encoded early antigen (Ead) correlate with WHO classification. Many ISGs are induced by viral infection of mammalian cells to prevent viral replication and further viral spread, collectively called the ‘host antiviral response’ (25). The list of ISGs that are differentially expressed in type II versus III NPC tumors includes many (Fig. 3A ), which have well-known antiviral properties (reviewed in refs. 30, 36), implying a role for a viral agent in NPC. The obvious candidate is EBV because nonkeratinizing NPCs all contain EBV, and the list of up-regulated ISGs includes several that are reported to be induced in mammalian cells by EBV (Fig. 3B; refs. 37–39). Because all type II and III NPC tumors contain EBV, we hypothesized that differential expression levels of ISGs between the two NPC subtypes might reflect different levels of viral gene expression.

To test this hypothesis, we sought to identify latent and replicative viral genes that are consistently expressed in NPC biopsies and assess if their expression levels correlated with tumor type. We designed quantitative RT-PCRs for three genes chosen to represent the different phases of the replicative cycle [the immediate early gene Zebra (Z)], the early gene diffuse early antigen (Ead), and the late gene viral capsid antigen (VCA); three latent genes (EBNA1Q-K, LMP1, and LMP2A); and the small untranslated EBER gene, which is usually highly expressed in EBV-positive cells]. cDNA was generated from 10 µg of total biopsy RNA (DNase treated), which equals approximately 2 x 106 tumor cells, and cDNA equivalent to approximately 5 x 104 tumor cells was used in each PCR. Weak signals were detected for Z mRNA in approximately one third of the NPC samples (average estimated signal ≤1 Akata cell; data not shown), consistent with previous failures to detect the protein immunohistochemically (40). Only one sample expressed detectable VCA transcripts (data not shown), suggesting that the lytic cycle in the NPC tumor cells may be nonproductive (41). By comparison, we readily detected Ead expression in virtually all (≥90%) of the samples (Table 3 ), making Ead the best lytic gene for comparative purpose. Overall, the level of Ead transcripts in the biopsies was low, indicating that cells replicating EBV were infrequent in NPC tumors as has been reported previously (12). Ead was detectable in all of the type II tumors with levels ranging from 1.6 to 15 (average, 5.6) cell equivalent per 5 x 104 biopsy cells, whereas levels in the type III tumors were disappearingly low, ranging from 0 to 1.4 (average, 0.6) cell equivalent per 5 x 104 biopsy cells.


View this table:
[in this window]
[in a new window]

  		Table 3. RNA expression levels of EBV-encoded genes in type II and III NPC

 
EBERs were detected in all of the biopsies at a level of 103 to 104 cell equivalents per 5 x 104 biopsy cells (Table 3), indicating that EBER RNA is probably expressed in the tumors at levels comparable with a B lymphoblastoid cell line. By comparison, EBNA1 transcription was routinely detected at 10 to 102 cell equivalents per 5 x 104 biopsy cells (Table 3). Similarly, low levels of transcription were found for LMP1 and LMP2A (data not shown), indicating that protein encoding latent gene transcription is markedly lower on a cell for cell basis in the tumors compared with lymphoblastoid cell lines. EBERs (3-fold), EBNA1 (1.5-fold), and LMP2A (3-fold not shown) were expressed slightly higher in type II than in type III tumors. Whether these relatively small differences in latent gene transcription between the tumor types are biologically significant is unclear. LMP1 was not informative because it was only detected in one third of the samples. These data indicate that WHO classification correlates with differential expression levels of the EBV-encoded Ead gene, a marker for viral replication in NPC.


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The molecular characterization of NPC described in this article yielded several insights into NPC classification and the complex association with EBV infection. Notable was that, based on unsupervised hierarchical clustering, we could detect no molecular genetic basis to justify the partitioning of nonkeratinizing NPC into the undifferentiated type II and poorly differentiated type III. These results provide no support for the hypothesis that type II and III NPCs have distinct molecular genetic origins. We identified a clear distinction based on elevated expression in type III NPC of a class of biologically related genes, the ISGs, as determined by marker analysis, GSEA, and by mining a gene set database. Conversely, expression of the EBV-encoded gene early lytic antigen (Ead) was inversely correlated. The biological function of ISG expression in mammalian cells is the establishment of an ‘antiviral state’. This host response is the first line of defense against viral infection as it counteracts viral replication and spread (25). This raises the possibility that a heightened IFN response in patients with type III NPC may be responsible for up-regulating ISGs and thereby suppressing EBV replication in that tumor type and may also be responsible for the histologically perceived differences between type II and III tumors. Alternatively, EBV replication may cause down-regulation of ISG expression as has been shown for other large DNA viruses, such as herpes simplex virus-1 (35). However, it is difficult to conceive how viral replication in such a small fraction of tumor cells could cause a global down-regulation of ISGs in the tumor unless it was mediated by a soluble/diffusible factor.

Our results are particularly intriguing in light of a recent report describing unsupervised partitioning based on ISG expression of several tumors with a suspected viral etiology (42). Coordinated ISG expression has also recently been implicated in tumor progression (43). We independently verified the correlation between tumor type, ISG expression, and viral infection using the MSigDB. The available data sets that best correlated with NPC tumor type were either IFN-{alpha} treatment related or cytomegalovirus infection associated or both.

We did not find any direct evidence for differential IFN expression within the tumors by sensitive quantitative RT-PCR. There is precedence for this in studies with varicella zoster that showed IFN production not in the infected cells per se but in the adjacent uninfected cells, thereby creating an antiviral state around the site of viral replication that functions to impede the spread of the virus to adjacent tissue (44). The up-regulation of ISGs at the tumor site could also be caused by IFNs produced in response to ongoing viral replication in tissue adjacent to the tumor (45). This would be consistent with serologic studies, which indicate that increasing levels of IgA antibodies to replicative antigens are diagnostic of NPC disease progression and relapse (46), although there is little viral replication in the tumor itself. The implication is that increased viral replication, at a site adjacent to the tumor, is occurring in parallel with tumor development. Alternatively, the ISG response could be the result of IFNs produced at an earlier stage of tumor development or even through an ‘IFN-independent’ mechanism (35).

Although it is impossible to test directly that EBV infection is responsible for the ISG expression in NPC tumors, there are several lines of evidence that support this view. First is that EBV is 100% associated with nonkeratinizing NPC tumors (4) and can induce ISG expression in vitro (37–39). Second, although the ISGs overexpressed in NPC correlate with the global set induced in vitro by IFN, the correlation is much stronger with the subset induced in SLE patients (31). EBV infection is deregulated in SLE as evidenced by increased viral load, viral replication, and circulating antibodies (32, 33, 47). This has led some investigators to speculate that EBV may play an etiologic role in SLE, whereas we have suggested the reverse hypothesis (i.e., that the immunologic disruption of SLE causes deregulation of EBV). This prompted us to conclude that the striking similarity in ISG expression pattern between NPC tumors and SLE may be linked to the increased EBV activity observed in both diseases.

The level of EBV replication we detected in the tumors was low but could be important for tumor growth. Recent studies have implicated low-level EBV replication in tumor progression in the severe combined immunodeficient mouse tumor system (48). Similarly, in Kaposi's sarcoma, replication of the associated herpes virus is detected at very low levels that are required for tumorigenesis. Low-level chronic EBV replication in or adjacent to the tumor could also play a role in stimulating inflammation-driven tumor progression (49). Consistent with this, our arrays showed up-regulation of macrophage markers (vascular endothelial growth factor, MMD, CD68, CD16, and CCL2) especially in type III tumors.

Hierarchical clustering did not yield a clear partitioning of the tumor samples into type II and III. We also analyzed our biopsy samples using a novel computational method, consensus clustering (50). Although the data set is too small to derive definitive conclusions, the analysis did suggest that nonkeratinizing NPCs may be divided into two clearly resolved and novel subclasses independent of WHO classification or EBV gene expression levels. Marker gene analysis indicated that the two subsets may be distinguished based on differential expression of genes related to DNA metabolism, cell cycle, and apoptosis. A larger sample size is needed to establish the veracity of these preliminary observations, but they suggest that a molecular genetic partitioning of nonkeratinizing NPC may be possible.

In summary, expression profiling of nonkeratinizing NPC indicated that WHO ‘poorly’ differentiated (type II) and ‘undifferentiated’ (type III) subtypes have not originated from obviously distinct epithelial precursors. Rather, WHO histopathology is associated with distinct expression levels of biologically related ISG genes. Chronic infection with EBV in NPC tumors and/or in tissues within close proximity of the tumors leading to an antiviral state in the tumors could account for the large-scale alteration in ISG expression. Differential expression of both ISGs and EBV-encoded genes points to a complex interaction between EBV infection/replication and the ‘antiviral response’ in NPC tumor cells. Our findings support the idea that the histologic differences observed in NPCs are not the result of genetically distinct tumor types but are the result of external factors, such as exposure to IFNs and EBV activity.


   Acknowledgments
 
Grant support: USPHS grants AI 18757 and CA 65883 (D.A. Thorley-Lawson).

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.


   Footnotes
 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

Current address for D.M. Pegtel: Division of Cell Biology, Netherlands Cancer Institute, 1066 CX Amsterdam, the Netherlands.

5 For a detailed protocol of microarray normalization methods, see http://www.broad.mit.edu/MPR/CNS. Back

6 For more details, go to http://www.broad.mit.edu/gsea/doc/subramanian_tamayo_gsea_pnas.pdf. Back

7 http://www.broad.mit.edu/gsea/msigdb/msigdb_index.html. Back

Received 5/23/06. Revised 9/29/06. Accepted 11/21/06.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Shanmugaratnam K, Sobin LH. The World Health Organization histological classification of tumours of the upper respiratory tract and ear. A commentary on the second edition. Cancer 1993;71:2689–97.[CrossRef][Medline]
  2. Ali MY, Shanmugaratnam K. Cytodiagnosis of nasopharyngeal carcinoma. Acta Cytol 1967;11:54–60.[Medline]
  3. Lin HS, Lin CS, Yeh S, Tu SM. Fine structure of nasopharyngeal carcinoma with special reference to the anaplastic type. Cancer 1969;23:390–405.[CrossRef][Medline]
  4. Klein G. The relationship of the virus to nasopharyngeal carcinoma. In: Epstein MA, Achong BG, editors. The Epstein-Barr virus. Berlin: Springer-Verlag; 1979. p. 340–50.
  5. Krueger GR, Kottaridis SD, Wolf H, et al. Histological types of nasopharyngeal carcinoma as compared to EBV serology. Anticancer Res 1981;1:187–94.[Medline]
  6. Brooks L, Yao QY, Rickinson AB, Young LS. Epstein-Barr virus latent gene transcription in nasopharyngeal carcinoma cells: coexpression of EBNA1, LMP1, and LMP2 transcripts. J Virol 1992;66:2689–97.[Abstract/Free Full Text]
  7. Allen MD, Young LS, Dawson CW. The Epstein-Barr virus-encoded LMP2A and LMP2B proteins promote epithelial cell spreading and motility. J Virol 2005;79:1789–802.[Abstract/Free Full Text]
  8. Caldwell RG, Wilson JB, Anderson SJ, Longnecker R. Epstein-Barr virus LMP2A drives B cell development and survival in the absence of normal B cell receptor signals. Immunity 1998;9:405–11.[CrossRef][Medline]
  9. Pegtel DM, Subramanian A, Sheen TS, et al. Epstein-Barr-virus-encoded LMP2A induces primary epithelial cell migration and invasion: possible role in nasopharyngeal carcinoma metastasis. J Virol 2005;79:15430–42.[Abstract/Free Full Text]
  10. Hudnall SD, Ge Y, Wei L, et al. Distribution and phenotype of Epstein-Barr virus-infected cells in human pharyngeal tonsils. Mod Pathol 2005;18:519–27.[CrossRef][Medline]
  11. Pegtel DM, Middeldorp J, Thorley-Lawson DA. Epstein-Barr virus infection in ex vivo tonsil epithelial cell cultures of asymptomatic carriers. J Virol 2004;78:12613–24.[Abstract/Free Full Text]
  12. Niedobitek G, Agathanggelou A, Nicholls JM. Epstein-Barr virus infection and the pathogenesis of nasopharyngeal carcinoma: viral gene expression, tumour cell phenotype, and the role of the lymphoid stroma. Semin Cancer Biol 1996;7:165–74.[CrossRef][Medline]
  13. Capell DF. Pathology of nasopharyngeal tumours. J Laryngol Otol 1938;53:558–80.
  14. Agathanggelou A, Niedobitek G, Chen R, et al. Expression of immune regulatory molecules in Epstein-Barr virus-associated nasopharyngeal carcinomas with prominent lymphoid stroma. Evidence for a functional interaction between epithelial tumor cells and infiltrating lymphoid cells. Am J Pathol 1995;147:1152–60.[Abstract]
  15. Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass cancer diagnosis using tumor gene expression signatures. Proc Natl Acad Sci U S A 2001;98:15149–54.[Abstract/Free Full Text]
  16. Fleming ID, Cooper JS, Murphy GP, et al. American Joint Committee on Cancer cancer staging manual. 5th edition. Philadelphia: Lippincott-Raven; 1997.
  17. Shanmugaratnam KSL. World Health Organization international histological classification: histological tumours of the upper respiratory tract and ear, 2nd ed. Berlin, 1991.
  18. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999;286:531–7.[Abstract/Free Full Text]
  19. Pomeroy SL, Tamayo P, Gaasenbeek M, et al. Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 2002;415:436–42.[CrossRef][Medline]
  20. Hochberg D, Souza T, Catalina M, et al. Acute infection with Epstein-Barr virus targets and overwhelms the peripheral memory B-cell compartment with resting, latently infected cells. J Virol 2004;78:5194–204.[Abstract/Free Full Text]
  21. Tierney RJ, Steven N, Young LS, Rickinson AB. Epstein-Barr virus latency in blood mononuclear cells: analysis of viral gene transcription during primary infection and in the carrier state. J Virol 1994;68:7374–85.[Abstract/Free Full Text]
  22. Hayes DP, Brink AA, Vervoort MB, et al. Expression of Epstein-Barr virus (EBV) transcripts encoding homologues to important human proteins in diverse EBV associated diseases. Mol Pathol 1999;52:97–103.[Abstract]
  23. Prang NS, Hornef MW, Jager M, et al. Lytic replication of Epstein-Barr virus in the peripheral blood: analysis of viral gene expression in B lymphocytes during infectious mononucleosis and in the normal carrier state [see comments]. Blood 1997;89:1665–77.[Abstract/Free Full Text]
  24. Castelruiz Y, Larrea E, Boya P, et al. J. Interferon alfa subtypes and levels of type I interferons in the liver and peripheral mononuclear cells in patients with chronic hepatitis C and controls. Hepatology 1999;29:1900–4.[CrossRef][Medline]
  25. Goodbourn S, Didcock L, Randall RE. Interferons: cell signalling, immune modulation, antiviral response, and virus countermeasures. J Gen Virol 2000;81:2341–64.[Free Full Text]
  26. Mootha VK, Lindgren CM, Eriksson KF, et al. PGC-1{alpha}-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 2003;34:267–73.[CrossRef][Medline]
  27. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005;102:15545–50.[Abstract/Free Full Text]
  28. Shih WJ, Ohman-Strickland PA, Lin Y. Analysis of pilot and early phase studies with small sample sizes. Stat Med 2004;23:1827–42.[CrossRef][Medline]
  29. Der SD, Zhou A, Williams BR, Silverman RH. Identification of genes differentially regulated by interferon {alpha}, ß, or {gamma} using oligonucleotide arrays. Proc Natl Acad Sci U S A 1998;95:15623–8.[Abstract/Free Full Text]
  30. Sen GC. Viruses and interferons. Annu Rev Microbiol 2001;55:255–81.[CrossRef][Medline]
  31. Bennett L, Palucka AK, Arce E, et al. Interferon and granulopoiesis signatures in systemic lupus erythematosus blood. J Exp Med 2003;197:711–23.[Abstract/Free Full Text]
  32. Gross AJ, Hochberg D, Rand WM, Thorley-Lawson DA. EBV and systemic lupus erythematosus: a new perspective. J Immunol 2005;174:6599–607.[Abstract/Free Full Text]
  33. Kang I, Quan T, Nolasco H, et al. Defective control of latent Epstein-Barr virus infection in systemic lupus erythematosus. J Immunol 2004;172:1287–94.[Abstract/Free Full Text]
  34. Stark GR, Kerr IM, Williams BR, Silverman RH, Schreiber RD. How cells respond to interferons. Annu Rev Biochem 1998;67:227–64.[CrossRef][Medline]
  35. Mossman KL, Macgregor PF, Rozmus JJ, et al. Herpes simplex virus triggers and then disarms a host antiviral response. J Virol 2001;75:750–8.[Abstract/Free Full Text]
  36. Samuel CE. Antiviral actions of interferons. Clin Microbiol Rev 2001;14:778–809, table of contents.[Abstract/Free Full Text]
  37. Carter KL, Cahir-McFarland E, Kieff E. Epstein-Barr virus-induced changes in B-lymphocyte gene expression. J Virol 2002;76:10427–36.[Abstract/Free Full Text]
  38. Ruvolo V, Navarro L, Sample CE, et al. The Epstein-Barr virus SM protein induces STAT1 and interferon-stimulated gene expression. J Virol 2003;77:3690–701.[Abstract/Free Full Text]
  39. Zhang J, Das SC, Kotalik C, et al. The latent membrane protein 1 of Epstein-Barr virus establishes an antiviral state via induction of interferon-stimulated genes. J Biol Chem 2004;279:46335–42.[Abstract/Free Full Text]
  40. Niedobitek G, Young LS, Sam CK, et al. Expression of Epstein-Barr virus genes and of lymphocyte activation molecules in undifferentiated nasopharyngeal carcinomas. Am J Pathol 1992;140:879–87.[Abstract]
  41. Martel-Renoir D, Grunewald V, Touitou R, et al. I. Qualitative analysis of the expression of Epstein-Barr virus lytic genes in nasopharyngeal carcinoma biopsies. J Gen Virol 1995;76:1401–8.[Abstract/Free Full Text]
  42. Einav U, Tabach Y, Getz G, et al. Gene expression analysis reveals a strong signature of an interferon-induced pathway in childhood lymphoblastic leukemia as well as in breast and ovarian cancer. Oncogene 2005;24:6367–75.[Medline]
  43. Shou J, Soriano R, Hayward SW, et al. Expression profiling of a human cell line model of prostatic cancer reveals a direct involvement of interferon signaling in prostate tumor progression. Proc Natl Acad Sci U S A 2002;99:2830–5.[Abstract/Free Full Text]
  44. Ku CC, Zerboni L, Ito H, et al. Varicella-zoster virus transfer to skin by T cells and modulation of viral replication by epidermal cell interferon-{alpha}. J Exp Med 2004;200:917–25.[Abstract/Free Full Text]
  45. Lung ML, Chan KH, Lam WP, et al. In situ detection of Epstein-Barr virus markers in nasopharyngeal carcinoma patients. Oncology 1989;46:310–7.[Medline]
  46. Gan YY, Fones-Tan A, Chan SH. Molecular diagnosis of nasopharyngeal carcinoma: a review. Ann Acad Med Singapore 1996;25:71–4.[Medline]
  47. McClain MT, Heinlen LD, Dennis GJ, et al. Early events in lupus humoral autoimmunity suggest initiation through molecular mimicry. Nat Med 2005;11:85–9.[CrossRef][Medline]
  48. Hong GK, Gulley ML, Feng WH, et al. Epstein-Barr virus lytic infection contributes to lymphoproliferative disease in a SCID mouse model. J Virol 2005;79:13993–4003.[Abstract/Free Full Text]
  49. Mantovani A, Sozzani S, Locati M, et al. Infiltration of tumours by macrophages and dendritic cells: tumour-associated macrophages as a paradigm for polarized M2 mononuclear phagocytes. Novartis Found Symp 2004;256:137–45; discussion 46–8, 259–69.[Medline]
  50. Monti S, Tamayo P, Mesirov J, et al. Consensus clustering: a resampling-based method for class discovery and visualization of gene expression microarray data. Machine Learning 2003;52:91–118.[CrossRef]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Pegtel, D. M.
Right arrow Articles by Thorley-Lawson, D. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Pegtel, D. M.
Right arrow Articles by Thorley-Lawson, D. A.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online