Characterization of Epstein-Barr virus–infected B cells in patients with posttransplantation lymphoproliferative disease: disappearance after rituximab therapy does not predict clinical response

Epstein-Barr virus (EBV) is a widespread gamma herpesvirus associated with several human malignancies, including nasopharyngeal carcinoma, Burkitt's lymphoma, Hodgkin's disease, and post-transplantation lymphoproliferative disease (PTLD).1-4 In the blood of healthy carriers, the presence of EBV-infected cells can be demonstrated by the spontaneous outgrowth of EBV-immortalized lymphocytes or by polymerase chain reaction (PCR).5,6 Cell fractionation studies show that the PCR-positive population of cells are resting, memory B lymphocytes.7,8 Reverse transcription PCR (RT-PCR) studies show that these infected cells express a highly restricted set of viral latency genes.9-11 In contrast, in vitro EBV-immortalized lymphoblastoid cells (LCLs) proliferate and express the full set of viral latency genes.12 Thus, the population of infected lymphocytes in healthy carriers is very different from LCLs immortalized in vitro. The absence of a population of immortalized LCLs in healthy carriers is thought to reflect immune surveillance.9 

PTLD is an EBV-associated process that occurs in organ and bone marrow transplant recipients as a complication of immunosuppression. Tumor cells often express the full set of viral latency antigens, paralleling the pattern of viral gene expression in LCLs.13 Several groups have reported that EBV loads in the peripheral blood lymphocytes of patients with PTLD are increased at the onset of disease and that viral loads fall with effective treatment.14-18 In the studies described here, we investigated the character of EBV-infected peripheral blood mononuclear cells (PBMCs) in patients with PTLD and in healthy controls. We also assessed the impact of therapy with the humanized anti-CD20 antibody rituximab on viral load and the relationship between viral load and tumor progression.

PTLD patients were seen at the Johns Hopkins Hospital and The Ohio State University. All patients were EBV seropositive prior to transplantation. Patients were described as having either active PTLD or PTLD in remission according to whether physical examination and imaging showed evidence of disease at the time of analysis. The transplanted organ, histology of the tumor, and other clinical information are summarized in Table 1. Healthy controls were platelet donors known to be EBV⁺ by serology whose leukopheresis products were collected under a human investigations committee–approved protocol.

PBMCs were isolated by density gradient centrifugation by means of Ficoll-Hypaque 1.077 (Biochrom, Berlin, Germany) and cryopreserved immediately. In some experiments, CD19⁺ B cells were enriched by positive selection by means of immunomagnetic beads (Dynal, Oslo, Norway).

The following cell lines were used: Namalwa, an EBV⁺Burkitt's lymphoma cell line that contains 2 EBV genomes per cell20; CA46, an EBV⁻ Burkitt's lymphoma cell line; B95.8, an EBV⁺ marmoset LCL; and Rael, an EBV⁺ Burkitt's lymphoma cell line. Cell lines were maintained in standard RPMI medium (RPMI 1640, 2 mM L-glutamine, 10 mM Hepes, 100 IU/mL penicillin, 100 μg/mL streptomycin, 10% vol/vol fetal bovine serum).

EBV load (copy number of EBV genomes in PBMCs) was determined by a quantitative competitive PCR (QC-PCR) method by means of a detection kit from BioSource International (Camarillo, CA). Briefly, cells (10⁶ PBMCs or mixtures of EBV⁺ and EBV⁻ cells) were spiked with 200 copies of the Internal Calibration Standard (ICS), a DNA sequence with flanking primer-binding sites identical to those of the EBV sequence to be detected. DNA was isolated by means of the QIAamp Blood Kit (QIAGEN Inc, Valencia, CA) and eluted in 100 μL dH₂O. We used 40 μL DNA in amplification reactions. The PCR primers (5′ primer,5′-GTGGTCCGCATGTTTTGATC, including nucleotide positions 6780-6800; 3′ primer,5′-GCAACGGCTGTCCTGTTTGA, including nucleotide positions 6969-6950), one of which is biotinylated, amplify a region of the EBER1 gene that is highly conserved.21 PCR involved an initial denaturation for 2.5 minutes at 95°C, followed by 40 cycles of 30 seconds at 94°C, 30 seconds at 60°C, and 1 minute at 72°C, with a final extension for 15 minutes at 72°C. The PCR products were denatured and hybridized to either ICS or EBV sequence-specific probes prebound to microwells. The bound PCR products were detected by addition of a streptavidin–horseradish peroxidase (HRP) conjugate followed by the HRP substrate, 3,3′,5,5′-tetramethylbenzidine. The reaction was terminated by addition of a stop solution. The optical density (OD) of microwells was then read in a 7520 Microplate Reader (Cambridge Technology, Cambridge, MA). The copy number of EBV DNA was calculated with the use of the OD of the ICS amplification product as a standard. For quantitation of viral loads greater than 5000 per sample, DNA was diluted 1:100 in dH₂O, and 200 copies of ICS were added to each 100 μL of diluted DNA before PCR.

The frequency of EBV-infected cells was determined by PCR of serial dilutions of PBMCs. At each dilution, 8 to 24 replicates were prepared. DNA was isolated as above and amplified with primers (5′ primer,5′-CTTTAGAGGCGAATGGGCGCCA, including nucleotide positions 14 068-14 089; 3′ primer,5′-TCCAGGGCCTTCACTTCGGTCT, including nucleotide positions 14 583-14 562) for the BamH-W repeat of B95.8 EBV to give a product of 516 bp.22 Time-release PCR with AmpliTaq Gold DNA polymerase (PerkinElmer, Norwalk, CT) was used to maximize sensitivity.23 The reaction involved an initial denaturation for 3 minutes at 95°C, followed by 60 cycles of 1 minute at 95°C, 30 seconds at 64°C, and 1 minute at 72°C, with a final extension for 15 minutes at 72°C. The PCR products were electrophoresed on a 1.8% agarose gel and transferred in 0.4 N NaOH onto a HyBond N⁺ membrane (Amersham, Piscataway, NJ). The membrane was hybridized with a [γ-³²P]adenosine triphosphate–labeled internal oligonucleotide probe (5′-GTTGCTAGGCCACCTTCTC, including nucleotide positions 14 262-14 280) in Rapid-Hyb buffer system (Amersham) at 42°C overnight. The membrane was then washed and autoradiographed for 20 minutes.

The percentage of samples without EBV sequences as determined by PCR with the use of BamHI-W primers was plotted as a function of the input number of PBMCs. The resulting curve was fit with an empiric model of the form f(x) = 1/(1+(x/c)^b), where x is the average number of PBMCs in which an EBV-infected cell can be found. This curve has a maximum of 1, a minimum of 0, and a slope of b at the “mid-effect” value of c. The mid-effect value is the input number of PBMCs per sample when the percentage of negative samples is 50%. Assuming the fitted curve approximated a Poisson probability model, we estimated the frequency of EBV-infected cells by interpolation of the best-fit curve at f(x) = 0.37.24,25 For curve fitting and estimation of the frequency of EBV-infected cells, Sigma Plot regression curve fitter software (Version 4.0; SPSS Inc, Chicago, IL) was used. Differences in EBV load and frequency of EBV-infected cells between patients with active PTLD and healthy carriers or patients in remission were tested by means of the Wilcoxon nonparametric test.

RT-PCR primers and internal probes for EBV transcripts are listed in Table 2. Total RNA was extracted from PBMCs by means of TriZol (Gibco BRL, Gaithersburg, MD). In each case, 5 μg RNA was used as the template for amplification. Oligo d(T)₁₆ and the GeneAmp RNA PCR kit (PerkinElmer) were used for reverse transcription and PCR. PCR was carried out by means of AmpliTaq Gold and an initial denaturation for 9.5 minutes at 95°C, followed by 40 cycles of 30 seconds at 94°C, 30 seconds at optimal annealing temperature, and 60 seconds at 72°C, with a final extension for 10 minutes at 72°C. PCR products were electrophoresed, transferred onto Hybond-H⁺ membrane, and hybridized with internal oligonucleotide probes as described above.

Five patients with active PTLD who failed conventional treatment as listed in Table 1 received 4 weekly infusions of rituximab at a dose of 375 mg/m². Blood was drawn before treatment and once every month after the first dose of rituximab. In 2 patients, blood was also drawn once or twice per week within the first month in order to pinpoint the onset of response to rituximab. PBMCs were isolated and EBV loads measured by means of QC-PCR as described earlier.

QC-PCR was used to determine EBV load (copy number of EBV genomes in PBMCs). The accuracy and reproducibility of this assay were determined by measuring the copy number of EBV genomes from mixtures of EBV⁺ and EBV⁻ Burkitt's lymphoma cell lines, Namalwa and CA46, respectively. In contrast to many EBV⁺cell lines, there are no viral episomes in Namalwa, and thus, the copy number per cell is stable. A log-linear relationship was found between the actual and the measured copy number of EBV genomes (Figure1). Linear regression analysis yielded the standard curve, which was used for calculation of the actual copy number of EBV genomes. The curve is described by the equation: log(added-EBV) = 1.58 × log(measured-EBV) − 1.71.

EBV load in the PBMCs of 13 healthy carriers and 14 PTLD patients was determined by EBV QC-PCR (Figure 2; Table3). Of the 14 patients, 8 had active PTLD and 6 were in remission after chemotherapy or withdrawal of immunosuppression. Viral DNA was detected in 12 of 13 healthy carriers and in each of the patients with PTLD. In healthy EBV carriers, viral load ranged from undetectable to 1921 EBV genomes per 10⁶PBMCs (mean, 335). In patients with active PTLD, EBV load ranged from 2930 to 72 537 EBV genomes per 10⁶ PBMCs (mean, 18 539). The difference was significant (P = .0002). In patients with PTLD in remission, EBV load ranged from 109 to 2742 EBV genomes per 10⁶ PBMCs (mean, 1216), significantly lower than viral load in patients with active PTLD (P = .002).

Experiments were carried out to determine whether the increased EBV load in PTLD patients reflects an increase in the frequency of EBV-infected cells or an increase in the copy number of viral genomes per infected cell. PBMCs were serially diluted; DNA was isolated; and PCR for the BamH-W sequences of the viral genome was used to determine the presence or absence of viral DNA in any particular dilution. In contrast to the EBER1 primers used for QC-PCR, the BamH-W primers amplify a region that is repeated up to 11 times in the viral genome. As a result, BamH-W PCR is more sensitive than EBER1 PCR. A representative limiting dilution DNA PCR experiment is presented in Figure 3. Poisson statistical analysis of the limiting dilution DNA PCR studies was used to approximate the frequency of EBV-infected cells in PBMCs in healthy carriers and in patients with active PTLD. The frequencies of EBV-infected cells are summarized in Table 3. Frequencies averaged 9 infected cells per 10⁶ PBMCs (range, 0.05 to 24) in healthy controls and 271 per 10⁶ PBMCs (range, 32 to 684) in patients with active PTLD. The difference was significant (P = .008). EBV load and the frequency of infected cells were correlated (Spearman R = 0.95; Figure 4).

The copy number of EBV genomes per infected cell was measured to determine whether increased copy numbers per cell contributed to the increased viral loads in patients with active PTLD. Copy numbers per cell were measured by QC-PCR at dilutions of PBMCs that were estimated to have less than 1 EBV-infected cell. Four healthy carriers and 3 patients with active PTLD were studied. At least 15 samples with detectable viral load at dilutions of PBMCs were measured for each individual. In both healthy carriers and patients with active PTLD, the majority of infected cells harbored fewer than 50 EBV genome copies per cell (Table 4). By visual inspection, the distribution of viral genomes per cell is indistinguishable between healthy donors and PTLD patients (Figure5).

To further characterize the status of EBV infection in peripheral blood of healthy carriers and patients, latent and lytic EBV transcripts were analyzed by RT-PCR in 2 healthy carriers and 2 patients with active PTLD. EBV transcripts analyzed included Epstein-Barr nuclear antigen (EBNA) 1, EBNA2, latent membrane protein (LMP) 2A, BZLF1, BLLF1, and BART. The sensitivity of oligonucleotide primer-probe combinations used in RT-PCR analysis was determined by assay of cell mixtures containing a standard number of EBV⁻ CA46 cells and serial 10-fold dilutions (10⁻² to 10⁻⁶) of the virus-productive B95.8 cells (Figure 6; Table5). The RNA extracted from aliquots (2 × 10⁶ cells each) of these mixed populations was analyzed by RT-PCR. Since the frequency of cells harboring EBV is increased in patients with active PTLD, PBMCs from healthy carriers were enriched for CD19⁺ cells in order to achieve similar frequencies of virus-infected cells in specimens from healthy carriers and in patients. About 10% of PBMCs were CD19⁺ B cells by immunomagnetic selection, so the frequencies of EBV-infected cells were approximately 240 and 40 per 10⁶ CD19⁺ cells for healthy carriers H1 and H2, respectively. This is close to 266 and 254 per 10⁶ PBMCs of patients P1 and P2, respectively. The pattern of viral gene expression in peripheral blood was indistinguishable between healthy carriers and those patients with active PTLD. The abundant transcript EBER1 was detected in all individuals examined. LMP2A was detected in 1 of 2 healthy carriers and 1 of 2 patients (Figure 7). However, EBNA1 and EBNA2 were not detected in any of the individuals tested, nor were the lytic transcripts BZLF1 and BLLF1. BART was detected in 1 of 2 healthy carriers and in both patients with active PTLD. Results of RT-PCR are summarized in Table 5.

Monoclonal antibodies directed against B-cell markers have been used in the treatment of PTLD.28 Recently, rituximab, a monoclonal antibody targeting CD20, a pan–B-cell marker, has been available for clinical use.29-31 We followed the EBV load in PBMCs of 5 patients with active PTLD treated with rituximab. A dramatic fall in EBV load was seen after therapy in each of the patients studied (Table 6). Two patients with serial measurements in the days after therapy showed that the decline occurred within 24 to 72 hours (Figure8). The rapid disappearance of virus-infected cells is similar to that described for the disappearance of B cells from the periphery following rituximab therapy.32-34 Although all of the patients' tumors expressed CD20, tumors progressed in 3 of 5 patients, including patient P1 whose dramatic fall in viral load is illustrated in Figure 8A. Only one patient's tumor responded to therapy. Curiously, the responding patient (P4) had the highest residual EBV load following rituximab therapy. Note that patient P6 with primary central nervous system lymphoma was inevaluable for tumor response because cranial irradiation was used as the primary treatment and rituximab was used as consolidation. These results provide in vivo evidence that the increased peripheral blood viral load in PTLD patients is harbored in CD20⁺ cells and that these cells are readily eliminated from the circulation in patients with PTLD, even when there is clinical tumor progression.

This investigation shows that an increased frequency of EBV-infected cells, rather than an increased copy number of viral genomes per infected cell, accounts for the increased viral load in the lymphocytes of patients with PTLD. Furthermore, virus-infected cells in the blood of PTLD patients are indistinguishable from those in healthy carriers and differ from LCL-like lymphoblasts in terms of viral gene expression. In addition, in patients with PTLD, a fall in EBV load does not predict clinical response following rituximab therapy.

EBV load has been measured by PCR-based assays by several investigators. These methods include quantitative-competitive PCR (QC-PCR),14,15 semiquantitative PCR,16-18,35,36 and real-time PCR.37-39 Using these techniques, groups of investigators measured EBV load in PTLD patients and in healthy carriers. Their reports all demonstrate increased viral load in patients, but there is striking variability in absolute values across studies (Table 7). The QC-PCR strategy was selected in the current study because it allows the test sample to be co-amplified in the same PCR reactions with a competitor DNA, thus permitting sample, primer pair, and target sequence variation to be internally controlled. Our method differs from the QC-PCR approach used by 2 other groups14,15 in that rather than using series of increasing quantities of competitor DNA, we used a fixed quantity of competitor DNA for co-amplification followed by comparison with a standard curve. The approach used here requires less material and yields a continuous measure. EBV load in patients with PTLD and in healthy carriers reported in this study is similar to what was reported by 2 other groups that used QC-PCR and the group that used real-time PCR. This study confirmed previous reports that PTLD is associated with an increased viral load in PBMCs.

The increased EBV load might reflect an increase in the number of EBV-infected cells, an increase in the copy number of EBV genomes per infected cell, or both. By limiting dilution DNA PCR, we show that the frequencies of EBV-infected cells are increased in the blood of patients with PTLD. In fact, the frequencies of virally infected cells correlate nicely with EBV loads (Figure 4). This is consistent with recent findings of Babcock et al.40 In a similar analysis, they reported 1 to 43, and 4 to 1670 EBV-infected cells per 10⁶ B cells in healthy controls and immunosuppressed patients. Others have reported that the rate of spontaneous LCL outgrowth also increased in the blood of PTLD patients.41 

The findings of increased EBV load and increased frequency of EBV-infected cells in the blood of PTLD patients lead to the question of whether the copy number of EBV genomes per cell differs in healthy seropositive individuals and in patients with PTLD. Our data show that the distribution of copy number per cell is indistinguishable between patients with PTLD and healthy carriers (Table 2; Figure 5). Most EBV-infected cells in patients as well as healthy seropositive individuals carry fewer than 50 EBV genomes per cell. Thus, the increased peripheral blood EBV load in patients with PTLD cannot be accounted for by an increased number of viral genomes per infected cell.

RT-PCR analysis showed no difference in the pattern of viral gene expression between healthy carriers and patients with PTLD, with LMP-2A being the only protein-coding transcript detected. Failure to detect other transcripts is unlikely to be a sensitivity issue in that our RT-PCR controls detected 1 positive cell among 10⁴ to 10⁶ negative cells while the lowest frequency of EBV-infected cells in our test samples for RT-PCR is approximately 1 in 10⁴ cells. This pattern of expression is different from viral gene expression of LCLs, which express the full spectrum of EBV latent cycle antigens, namely, the EBNAs 1, 2, 3A, 3B, 3C, -LP, and LMPs 1, 2A, and 2B.12 

Thus. EBV-infected cells in peripheral blood of patients with PTLD carry, in general, fewer than 50 copies of viral genomes per cell, express a highly restricted set of viral latency genes, and appear phenotypically as resting memory B cells8,40—the same population of infected cells found in healthy carriers. However, the frequency of infected cells is increased in patients with PTLD. Thus, there is no expansion of LCL-like immunoblasts in the peripheral blood of PTLD patients, as had been thought previously. Instead, virus-infected cells in the blood of patients with PTLD are similar to those found in healthy seropositive individuals.

Analysis of the frequency of infected cells and the copy number of EBV genomes per infected cell demonstrates that an increase in the frequency of infected cells accounts for most, if not all, of the increased EBV load in the PBMCs of PTLD patients. Proliferation of infected cells in the blood seems unlikely to account for the increased number of EBV-infected cells insofar as our evidence suggests that these cells are not expressing the viral proteins that drive proliferation. This conclusion is consistent with that of Babcock et al40 showing that EBV-infected cells in PTLD patients, as in healthy controls, are resting memory B cells. The increased numbers of EBV-infected B cells in PTLD patients may therefore reflect increased numbers of infectious events. If so, the site of infection is not likely to be the peripheral blood because neither lytic transcripts nor increased viral genome copy numbers per cell were detected. The locus of virion production and B-cell infection remains a subject of speculation in healthy individuals and PTLD patients. Babcock et al40 proposed that there is increased virion production in the lymphoid tissue of PTLD patients.

Our analysis of patients with active PTLD and PTLD in remission (after chemotherapy or withdrawal of immunosuppression) confirms a relationship between disease activity and viral load. However, this relationship disappears in patients treated with rituximab. EBV-infected lymphocytes in peripheral blood differ in their sensitivity to rituximab from tumor cells of PTLD insofar as the EBV-infected cells in the blood promptly and dramatically decline with therapy, while the response of tumor cells is variable. EBV-infected lymphocytes also differ from tumor cells in that virus-infected lymphocytes are resting cells that express a restricted set of viral antigens, while tumor cells actively proliferate and commonly express the full set of EBV-latency antigens.13 These differences suggest that virally infected cells in peripheral blood belong to a separate compartment from tumor cells. Thus, monitoring viral load in peripheral blood does not predict clinical response of patients with PTLD tumors.

The authors thank Dr Kathryn Kloppenstein for her assistance.