A Comparison Study of Different PCR Assays in Measuring Circulating Plasma Epstein-Barr Virus DNA Levels in Patients with Nasopharyngeal Carcinoma

Nasopharyngeal carcinoma is distinct from other head and neck cancers by its epidemiology, histopathology, and clinical characteristics. It is geographically endemic with an incidence as low as <0.5 cases per 100,000 in Western countries and as high as 25 per 100,000 in South East Asia (1). The primary treatment for nasopharyngeal carcinoma is fractionated external beam radiotherapy. Recent clinical trials support the addition of concomitant chemotherapy in patients with advanced stage tumors (2–4). Despite a high rate of initial response to therapy, tumor relapse is subsequently noted in 30% to 50% of patients with advanced stage neoplasms (2, 4). Such relapses presumably reflect the presence of microscopic residual tumor that cannot be readily detectable with the present staging and surveillance procedures. Therefore, efforts have been focused on identifying useful molecular predictors for treatment response and tumor recurrence in this disease.

Nasopharyngeal carcinoma is consistently associated with Epstein-Barr virus (EBV). Most nasopharyngeal carcinoma patients have been found to harbor the EBV genome in their tumor tissues (1). Within the last few years, it has been shown that cell-free EBV DNA could be identified in the plasma or serum of nasopharyngeal carcinoma patients (5–7) by PCR-based techniques, which make it possible to detect small amounts of DNA in a wide array of tissues. Further studies have shown that molecular detection of EBV DNA in patient blood may serve as a potential noninvasive test for tumor detection and prognostication for nasopharyngeal carcinoma in endemic areas (5, 6, 8–10). In a large prospective study involving 170 nasopharyngeal carcinoma patients, Chan et al. reported that the level of posttreatment plasma EBV DNA were superior to pretreatment levels in predicting progression-free and overall survival in these patients (8). Similarly, Lin et al. reported that patients with elevated pretreatment plasma EBV DNA levels or persistently detectable posttreatment levels had a significantly higher risk of relapse and death than those without in 99 patients with newly diagnosed nasopharyngeal carcinoma (10).

The PCR assay method for all of the above studies used a primer/probe set that span the BamHI-W region of the EBV genome. However, this DNA sequence occurs in multiple repeats (5-11) in different EBV genomes (11). Allan et al. found that the repeat number varied considerably among the naturally occurring EBV isolates with a mean number of six repeats (12). Additionally, some cell lines can harbor multiple viral genomes with differing numbers of repeats and the repeat number can change dramatically during the lytic phase of viral replication. Two commonly accepted PCR assay methods for measuring EBV load during infection includes one using a polymerase-1 gene (Pol-1) primer/probe set and the second using a primer/probe set for the latent membrane protein-2 gene (Lmp2), both of which are single-copy genes (13). In this study, we compared the performance of these three different assays in quantifying EBV DNA copies in a cell line and a subset of nasopharyngeal carcinoma patients with pretreatment plasma samples. In cell line studies, where single-gene copy plasmids were used for the standard curve, we found that the Pol-1 assay yielded approximately the expected DNA concentrations, whereas BamHI-W yielded higher concentrations. Similarly, in patients, the BamHI-W assay consistently yielded higher plasma EBV DNA levels than the other two assays. However, within the study patient cohort, there was a strong linear correlation in pretreatment DNA levels detected by the three assays, suggesting that all three approaches are appropriate for assessing circulating EBV DNA. We also did serial measurements of pre-, mid-, and posttreatment EBV DNA levels using the Pol-1 assay in a larger cohort of nasopharyngeal carcinoma patients treated with either radiotherapy alone or chemoradiotherapy. Using this method, posttreatment but not pretreatment EBV levels significantly predicted for freedom-from-relapse and overall survival in patients from both endemic and nonendemic regions.

Fifty-eight patients with a confirmed diagnosis of nasopharyngeal carcinoma from the Stanford University Medical Center and Pamela Youde Nethersole Eastern Hospital, Hong Kong were recruited with informed consent between July 2001 and June 2004. The study was approved by the Institutional Review Boards at both institutions. Patients underwent the standard staging procedures, which included nasopharyngoscopy and biopsy of the nasopharynx, magnetic resonance imaging or computed tomography of the nasopharynx and neck, chest X-ray, and laboratory studies. In patients with T₄ or N₃ tumors or in those with elevated liver function tests, chest and abdominal computed tomography, liver ultrasound, and bone scan were also done. Patients with evidence of distant metastasis were excluded from this study. Radiation and chemotherapy treatment were done per individual institutional protocols. At Stanford University Medical Center, stage III to IV patients were treated with concomitant chemoradiotherapy followed by adjuvant chemotherapy. A stereotactic radiosurgical boost of 8 to 12 Gy was also administered in patients with T_2b-4 tumors. At Pamela Youde Nethersole Eastern Hospital, patients with T_3-4N_0-1 tumors were randomized to either conventionally fractionated radiotherapy alone + chemotherapy or accelerated fractionated radiotherapy + chemotherapy and those with N_2-3 tumors were randomized to conventionally fractionated radiotherapy or the same radiation with concomitant and adjuvant chemotherapy. Brachytherapy or stereotactic radiosurgical boost were used in patients with persistent primary tumor. Patients were followed regularly with clinical examination and nasopharyngoscopy every 1 to 3 month(s) after treatment completion. The frequency of follow-up imaging studies was left at the discretion of the treating physician. Patients who developed signs or symptoms suspicious for recurrence underwent further imaging studies and biopsies as deemed appropriate.

Approximately 3.5 mL of blood was collected into EDTA anticoagulant before treatment, halfway into radiotherapy, at completion of radiation treatment, then every 3 to 6 months after completion of all therapy. The blood was centrifuged for 20 minutes at 3,000 rpm, plasma recovered, and frozen at 20°C until processing. DNA is isolated from 1 mL of plasma samples using the QIAamp Blood Mini Kit (QIAgen, Inc., Valencia, CA), eluting the column with 100 μL of sterile, deionized-distilled water. Then, 10 μL of eluant was used per well.

NAMALWA and HTB-60, which are EBV-positive cell lines, were purchased from the American Type Culture Collection (Manassas, VA) and cultured under the conditions specified. EBV plasmids used to construct the standard curve for the Pol-1 assay were prepared by cloning fresh PCR product into a pCR 2.1 vector using the TOPO-TA cloning kit as per the manufacturer's instructions (Invitrogen, Carlsbad, CA). The plasmid was purified using the QIAprep mini-prep kit (QIAgen), and its identity confirmed by sequencing.

Pol-1 assay. The Pol-1 assay, which amplifies a segment of the EBV Pol-1 gene, was adapted from Kimura et al. (13). The PCR reaction mix contains the following in a total volume of 50 μL per reaction: DNA template (10 μL of column eluant for plasma), 1× AmpliTaq Gold buffer II, 5 mmol/L MgCl₂, 125 μmol/L each deoxynucleotide triphosphate, 1.25 units TaqGold (Applied Biosystems, Foster City, CA), 10 pmol each primer and 5 pmol probe (Operon Technologies, Alameda, CA). The primer and probe sequences are as follows: CGGAAGCCCTCTGGACTTC (forward), CCCTGTTTATCCGATGGAATG (reverse), FAM-TGTACACGCACGAGAAATGCGCC-TAMRA (probe).

Each 96-well plate included patient samples (quadruplicate) run in parallel with plasmid standards for constructing a standard curve (duplicate) and appropriate controls (duplicate). The Pol-1 standard curve spanned 7-logs and was comprised of serial 10-fold dilutions containing 10⁷ down to 10¹ plasmid targets per well. An EBV-positive cell line (HTB-60) was the source of the positive control; DNA shown to test negative for EBV (from an anonymous patient) was the source of the negative control. A “no template control” was also run on each plate. The reactions were run at one cycle at 95°C for 10 minutes and 40 two-step cycles of 95°C for 15 seconds and 60°C for 1 minute on ABI7700 Sequence Detector (Applied Biosystems) according to the manufacturer's instructions. Fluorescence data were collected in real-time and analyzed with Sequence Detection System Software (version 1.9). Results were expressed as the number of copies of EBV genomes per milliliter of plasma. Samples showing fluorescence signals less than the signal of the 10¹ standard were outside the linear range of the assay and considered to have zero copies.

BamHI-W assay. The protocol for the BamHI-W assay, which amplifies a segment of the BAMH1 repeat region, was done as described by Lo et al. (6). For in vitro studies, EBV plasmids as described above were used to construct the standard curve. For analysis of patient plasma samples, serial dilutions of genomic DNA extracted from cultured NAMALWA cells were used to construct the standard curve. The standards were run in duplicate and in parallel with patient samples and controls. Analysis of the real-time PCR fluorescence data was as described for the Pol-1 assay.

Lmp2 assay. A 239-bp fragment of the Lmp2 gene (Genbank accession no. V01555) was amplified with specific primers for EBV as per the manufacturer's instructions (Roche Molecular Systems, Pleasanton, CA). Approximately 5 μL of plasma was used in each reaction. The resulting amplicon was then detected via fluorescence using a pair of hybridization probes, having sequence-homology to portions of the Lmp2 gene of EBV, again according to the manufacturer's instructions. The hybridization probes, under stringent conditions (55°C), were designed to exclusively hybridize to the Lmp2 gene of EBV. Cycling conditions consisted of a 10-minute hold at 95°C; 45 cycles of 95° for 10 seconds, 55°C for 15 seconds, and 72°C for 15 seconds done in a Light Cycler (Roche Molecular Systems).

Simple regression was used to compare and correlate pretreatment EBV DNA levels detected by the three different assays. Pretreatment EBV DNA levels were correlated with clinical variables such as tumor stage, nodal stage, overall disease stage, histology, gender, and ethnicity using the Student's t test. Pretreatment EBV DNA levels were also correlated with WBC count and primary tumor volumes using simple regression. Freedom-from-relapse was defined as the time interval between the date of diagnosis and the date of relapse or the date of last follow up for censoring patients. Log-rank and Kaplan-Meier methods were used to analyze the relationship of pre- and posttreatment EBV DNA levels to freedom-from-relapse and overall survival. Cox proportional hazard model was used for multivariate analysis (14). The Statview statistical software (SAS Institute, Inc., Carey, NY) was used.

Comparison between Pol-1 and BamH1-W assays. We did a gene bank blast of the forward primer sequence spanning the BamHI-W region and found 11 identical matches in the B95-8 EBV genome (accession number v01555) and 7 matches in the HH4 EBV genome (accession number AJ507799). This is consistent with the data reported by Allan et al., who noted that the repeat number of this sequence varied considerably among naturally occurring EBV isolates (12). Next, we used Pol-1 and BamHI-W primer/probe sets to quantify the number of EBV DNA copies from known concentrations of HTB 60 cell lines. For each method, we determined the EBV concentration from standard curves prepared using plasmids constructed so as to contain a single copy of the gene being assayed. Five different concentrations of EBV genomes were tested (1 × 10¹-10⁵); each differed by a magnitude of 10. The Pol-1 assay yielded concentrations that were on the average 6.9-fold higher than the expected concentrations (range, 3.8- to 7.7-fold); whereas the BamHI-W2 assay yielded concentrations that were on the average 21.2-fold higher than expected (range, 1.7- to 62-fold). For all but one tested concentrations, the BamHI-W assay consistently yielded concentrations that were 1.5- to 8.4-fold higher than that detected by Pol-1. These data are consistent with the existence of multiple repeats of the BamHI-W region in the EBV genome in this cell line.

Measurements of circulating EBV DNA levels in the plasma samples of 23 representative nasopharyngeal carcinoma patients using Pol-1, BamHI-W, and Lmp2 assays showed that, in most cases, BamHI-W yielded the highest concentrations of circulating DNA compared with the other two assays for individual patients (Fig. 1). However, taken as the whole for the entire group, the three assays yielded similar results: patients who had the highest DNA concentration as determined by one method also had the highest levels by the other two methods; and those with undetectable levels by one method also had very low or undetectable levels by the other two (Fig. 1). Simple regression analysis revealed a strong correlation among the three assays. As expected, the r² value was highest at 0.99 for the Pol-1 and Lmp2 correlation (P < 0.0001) because both assays detect single-copy genes. The r² was lower at 0.66 for the Pol-1 and BamHI-W correlation (P < 0.0001) and was at 0.55 for the Lmp2 and BamHI-W correlation (P < 0.0001).

Relationship between clinical outcomes and Pol-1-detected circulating Epstein-Barr virus DNA. Between the two single-copy gene methods, Pol-1 generally detected higher levels of circulating EBV DNA than Lmp2. Therefore, we proceeded to perform serial measurements of circulating EBV DNA in a larger cohort of patients before, during, and after radiation treatment using Pol-1. Between July 2001 and June 2004, we enrolled 58 patients with newly diagnosed nasopharyngeal carcinoma at the Pamela Youde Nethersole Eastern Hospital and at the Stanford University Medical Center. Table 1 shows the patient and treatment characteristics. The median follow-up for live patients was 25 months (range, 16-31 months).

Pretreatment EBV DNA levels were measured using the Pol-1 assay in 58 patients. Pretreatment EVB DNA levels ranged from 0 to 6,600 copies. It was undetectable in 29 patients, >0 in 28 patients, and was invaluable in 1 patient due to sample loss. We evaluated the relationship between pretreatment EBV DNA levels and several clinical variables. The data are shown in Table 2. Pretreatment EBV DNA levels correlated significantly with the overall stage and nodal stage groupings. In addition, patients with T_3-4 tended to have higher pretreatment EBV DNA levels compared with those with T_1-2 tumors, although the difference was not statistically significant (P = 0.11).

Midradiation EBV DNA levels were available in 41 patients, of which only 6 were >0. In four of these six patients, the midradiation levels were higher than the pretreatment levels, suggesting contribution from tumor lysis in response to therapy. Postradiation EBV DNA levels, which were obtained either immediately after treatment or within the first 3 months of follow-up, were available in 46 patients, of which 12 were >0. Eight of these 12 patients eventually relapsed compared with 3 of 34 patients with undetectable posttreatment levels. The follow-up for the four patients with elevated posttreatment DNA and without relapse ranged from 22 to 26 months. The estimated 2-year freedom-from-relapse rates were 92% for patients with undetectable posttreatment EBV DNA and 37% for those with detectable levels (P < 0.0001, Fig. 2A). Pretreatment EBV levels, when stratified according to undetectable versus detectable, failed to predict for treatment relapse in these patients (Fig. 2B).

Univariate analyses for freedom-from-relapse using the log rank tests were carried out evaluating the prognostic significance of pertinent clinical and laboratory variables. These included age, gender, WHO histology, T stage, N stage, overall stage, pre- and posttreatment EBV DNA levels. Only posttreatment EBV DNA level (P < 0.0001) and WHO histology (WHO I-II versus III, P = 0.03) reached significant levels on univariate analysis. N stage (N_0-1 versus N_2-3, P = 0.06) and overall stage (II-III versus IV, P = 0.07) were of borderline significance. Multivariate analysis using a Cox proportional hazard model incorporating WHO histology, N stage, overall stage and posttreatment EBV DNA levels showed that overall stage [favoring stage II-III, P = 0.02, hazard ratio = 5.3 (1.33-20.8)] and posttreatment DNA levels [favoring undetectable, P = 0.0006, hazard ratio = 21.7 (3.7-125)] were the only independent predictors for freedom-from-relapse.

Table 3 shows the time interval from initial EBV DNA detection after treatment to the time of relapse and the location of relapse in patients who failed and had at least one posttreatment or follow-up EBV DNA measurements. Eleven of these 13 patients had serial follow up measurements. As shown, the time from EBV DNA detection after treatment to that of relapse documentation ranges from 1.7 to 20 months (median 6.7 months). Distant relapses were generally detected earlier than local relapses.

As of the last follow up, seven patients have died, five from recurrent or metastatic nasopharyngeal carcinoma, one from a chemotherapy complication, and one from a C4 fracture of unclear etiology. The 2-year overall survival was 94% for patients with undetectable posttreatment EBV DNA compared with 55% for those with detectable levels (P = 0.002, Fig. 2C). Pretreatment EBV DNA level was not a prognostic factor for survival (P = 0.73). We did not perform multivariate analysis due to the small number of events.

Since it was first pioneered by Lo et al. (6), the DNA fragment corresponding to the BamHI-W region in the EBV genome has been used extensively in PCR for quantifying circulating EBV DNA in the plasma or serum of nasopharyngeal carcinoma patients in endemic regions (5, 8, 10, 15). Many of these studies have shown that pretreatment levels correlated with tumor stage (16), the likelihood of recurrence and survival (8, 10, 17), and the presence of persistent or metastatic disease (6). In addition, posttreatment EBV DNA profiles have been noted to be a strong predictor for relapse and survival in at least two large studies (8, 10). However, as shown in Table 4, the reported values for the medians and ranges of pretreatment circulating EBV DNA levels in newly diagnosed nasopharyngeal carcinoma patients varied considerably from study to study using the same primer/probe set and assay variables. Similarly, the proposed predictive values above which the level would portend a worse prognosis differ for different studies with the proposed level ranging from 1,500 to 40,568 DNA copies/mL for pretreatment and 0 to 500 copies/mL for posttreatment levels. In addition, controversies exist in the microbiology literature with regards to the use of the BamHI-W assay in assessing EBV viral load (13, 18–20). Although it is found to be more sensitive than other assays in detecting EBV DNA, it is thought to be not suitable for quantitating the EBV viral load due to its variable repeat numbers in different EBV laboratory isolates. To address these discrepancies, we investigated in detail the expression of the BamHI-W region in the EBV genome. We noted that multiple repeats of this DNA region exist in different EBV genomes and that the repeat number is different for different EBV strains. These data suggest that the absolute DNA values that are detected by the BamHI-W assay depend not only on the level of circulating EBV DNA but also on the number of BamHI-W repeats harbored in the viral genome. For example, for the sample tumor and viral burden, the measured circulating EBV DNA level using the BamHI-W assay may be 5-fold higher in a patient infected with an EBV strain that harbors five BamHI-W repeats compared with a patient infected with a strain that has only one repeat. Although similar detection sensitivity rates were reported for the BamHI-W and EBNA-1 assays with a correlation coefficient of 0.92 and a P value < 0.0005 (6), no actual values were presented in that manuscript. Therefore, in this study, we compared the performance of BamHI-W to Pol-1 and Lmp2 genes, which exist as single-copy genes in all EBV strains and which are commonly used to quantify the EBV load during active infection. Overall, we found that the BamHI-W assay consistently yielded higher EBV DNA levels in both cell lines and human plasma compared with the other two assays. This is consistent with the hypothesis that this DNA region could exist as multiple repeats in different EBV genomes. However, other variables may also contribute to the differences among assays. One source of interassay variability in real-time PCR is the difference in the PCR efficiency among the methods used, especially when the quantitation technique is relative rather than absolute. Here we employ absolute standard curves with empirically determined linearity limits; in this way, methodologic differences in PCR efficiencies among the three assays are minimized. Even in this situation, PCR efficiency differences can still play a role in variability because the standards (genomic, recombinant, or synthesized) may be more or less efficient than the patient samples. This standard versus sample relationship may not be consistent for all three assays. To investigate this possibility, we compared the PCR efficiencies of various templates—plasmid standards, genomic DNA standards, and cell line dilutional samples—for the Pol-1 and BamHI-W assays (data not shown). We found that there is a difference in amplification efficiency of ∼10% between the standards and the cell line DNA for each of these two methods. There may also be variability in amplification efficiency from sample-to-sample, although this would be inherent to the sample rather than the method and should hold consistently across all three methods. Thus, the variability in PCR efficiency may account for a small portion of the differences in the results noted for the assays.

It is noteworthy, however, that for the entire patient cohort, there was a significant correlation between the three approaches. As expected, the correlation coefficient was highest for the assays designed for the single-copy genes (Pol-1 and Lmp2). A lower but highly significant correlation coefficient was also observed when BamHI-W was compared with either Pol-1 or Lmp2, suggesting that the variable number of the BamHI-W repeats found in laboratory EBV strains may have a lesser impact on clinical plasma samples than previously predicted. Hence, the three assays seem to be similar in detecting EBV DNA levels with the most sensitive assay being that for the BamHI-W region.

Similar to prior studies (16), pretreatment EBV DNA using the Pol-1 assay significantly correlates with N stage and the overall stage of the tumors. However, unlike prior studies using the BamHI-W assay, pretreatment EBV DNA levels as detected by the Pol-1 assay was not a significant predictor for either freedom-from-relapse or overall survival in this study. This is probably due to the lower sensitivity of this method in detecting circulating EBV DNA. Only 50% of our patients had detectable EBV DNA levels at diagnosis, a result that is similar to those reported for EBNA-1 assays (21). Recently, Hsiao et al. showed that an increase in the number of PCR cycles from 35 to 50 cycles during EBNA-1 testing resulted in improved detection sensitivity from 35% to 75% but at the expense of specificity (22). An increase in the number of PCR cycles during Pol-1 testing (from 40 to 50 cycles) may be useful in improving DNA detection rate in these patients.

Serial measurements during radiation treatment showed a rapid reduction in the level of circulating EBV DNA, with most patients attaining undetectable levels within 3 weeks of initiating treatment. In a few patients, midtreatment levels were higher than pretreatment levels; however, the increase in EBV DNA level at midtreatment was not associated with relapse in these few patients (data not shown). These data suggest that this transient increase in EBV DNA levels may be related to tumor lysis from therapy and does not portend a worse outcome. The rapid decline in DNA levels on therapy and its lack of prognostic significance also suggest that serial measurements during radiotherapy are not useful for prognostication in these patients.

Similar to others (8, 10), we found that posttreatment EBV DNA levels measured either immediately or up to 3 months after completion of radiotherapy was highly significant for predicting treatment outcomes in these patients. This is independent of stage or pretreatment levels. Patients with undetectable levels had an estimated 2-year freedom-from-relapse rate of 92% compared with 37% for patients with detectable levels. In addition, EBV DNA levels were detected several months prior to documentation of recurrence in most patients. These were not spuriously false-positive findings as serial follow-up measurements before pathologic and radiographic documentation of relapse showed persistently elevated EBV DNA levels in several patients. As shown in Table 4, the time interval from posttreatment DNA detection to diagnosis of distant recurrence was generally shorter than that for local recurrence. This phenomenon is probably due to the difficulty in diagnosing local recurrences after high dose radiation treatment owing to significant posttreatment fibrosis that precludes optimal surveillance with conventional clinical and radiographic evaluation. Makitie et al. found that positron emission tomography scans correlated better with posttreatment EBV DNA levels than magnetic resonance imaging scans in a very small group of patients (23). The role of positron emission tomography scans in detecting relapses should therefore be further investigated in patients with rising EBV DNA levels after radiotherapy.

To date, most of the studies evaluating the role of circulating EBV DNA for prognostication and posttreatment surveillance in nasopharyngeal carcinoma patients have focused on patients from endemic regions. A unique aspect of our study is its inclusion of a nonendemic patient population. Although the number of nonendemic patients is small, circulating EBV DNA seems to be useful for monitoring recurrence in these patients as well. A larger confirmatory study in this patient group is needed to validate the benefit of EBV DNA as a surveillance tool in these patients.

We thank Yun Chuen Lo (Pathology Laboratory, Pamela Youde Nethersole Eastern Hospital) and Dr. Kuen Chan (Clinical Oncology, Pamela Youde Nethersole Eastern Hospital) for collection and initial processing of the plasma samples from Hong Kong.