Long-term treatment with methotrexate or tumor necrosis factor α inhibitors does not increase epstein-barr virus load in patients with rheumatoid arthritis

Authors


Abstract

Objective

We previously demonstrated that patients with rheumatoid arthritis (RA) have a 10-fold systemic Epstein-Barr virus (EBV) overload, very similar to that observed in healthy organ transplant recipients. Our objective was to monitor EBV load over time in patients with RA receiving methotrexate, infliximab, or etanercept to detect possible immunosuppression-associated EBV dysregulation, as described in posttransplant lymphoproliferative disease.

Methods

The EBV load in the peripheral blood mononuclear cells (PBMCs) from 19 patients receiving methotrexate, 68 patients receiving infliximab, and 48 patients receiving etanercept was monitored for durations ranging from 6 months to 5 years using a real-time polymerase chain reaction assay previously developed for that purpose. The effect of treatment duration on EBV load and the link between the Disease Activity Score in 28 joints and EBV load were analyzed by generalized estimating equations.

Results

Methotrexate tended to decrease EBV load over time, but this did not reach significance. Tumor necrosis factor α (TNFα) inhibitors did not significantly modify EBV load over time. Finally, high disease activity was significantly associated with high EBV load.

Conclusion

Long-term usage of methotrexate or TNFα inhibitors in patients with RA does not significantly influence EBV load in PBMCs.

INTRODUCTION

Patients with rheumatoid arthritis (RA) have an impaired immune response to Epstein-Barr virus (EBV). These patients have high-titer antibodies to EBV antigens (1). T lymphocytes from their peripheral blood are less efficient at controlling the outgrowth of EBV-infected B cells (2). Patients with RA have more EBV-infected B cells than normal individuals (3). Disease activity is associated with lower T cell responses to the EBV replication protein gp110 (4). Finally, we recently demonstrated that patients with RA have a 10-fold systemic EBV overload, very similar to that observed in healthy organ transplant recipients (5). Remarkably, both patients with RA and solid-organ transplant recipients are at increased risk to develop lymphoma (6–8). In the absence of immunosuppressive therapy, severity of the disease and inflammation seem to increase the risk of lymphoma in patients with RA (9). Current treatment of RA routinely includes potentially immunosuppressive medications such as methotrexate and tumor necrosis factor α (TNFα) inhibitors and is suspected to increase the risk of developing lymphoma. In solid-organ transplant recipients receiving immunosuppressants, emergence of lymphoma can be predicted by monitoring EBV load in peripheral blood mononuclear cells (PBMCs). EBV load >1,000 copies per 500 ng of PBMC DNA is considered a limit above which patients are at risk of developing EBV-associated posttransplant lymphoproliferative disorder, a condition characterized by polyclonal EBV-positive B lymphocyte proliferation, which can evolve into EBV-positive B cell lymphoma (7, 8).

We investigated whether long-term usage of methotrexate or TNFα inhibitors in patients with RA, whose EBV control is already impaired, could increase EBV load and therefore put these patients at risk of developing lymphoma, as already demonstrated in solid-organ transplant recipients. Therefore, we monitored EBV load in the PBMCs of 144 patients with RA receiving methotrexate or TNFα inhibitors (infliximab or etanercept) for periods ranging from 6 months to 5 years (mean followup duration 23 months for methotrexate, 28 months for infliximab, and 17 months for etanercept) using a real-time fluorescent polymerase chain reaction (PCR) assay that we developed previously (5).

We found that long-term treatment of patients with RA with methotrexate tended to decrease EBV load in PBMCs, but this did not reach significance. Similarly, long-term treatment with infliximab or etanercept did not significantly modify EBV load in patients with RA.

PATIENTS AND METHODS

Patients and controls.

A total of 144 patients with RA (satisfying the 1987 American College of Rheumatology [formerly the American Rheumatism Association] criteria [10]) from the rheumatology ward at Hôpital La Conception, Marseille, France, were followed for times ranging from 6 months to 5 years. Nineteen patients received methotrexate (7.5–12.5 mg/week); 68 patients received infliximab (3 mg/kg for 8 weeks), associated or not associated with methotrexate; 48 patients received etanercept (25 mg twice per week), associated or not associated with methotrexate; and 9 patients received etanercept after infliximab. Three patients were first included in the methotrexate group, then in the infliximab group. Four patients were first included in the methotrexate group, then in the etanercept group. None of the patients received any prophylactic antiviral therapy. Baseline characteristics of patients are presented in Table 1. Peripheral blood samples were obtained every 6 months for EBV DNA assay.

Table 1. Characteristics of patients*
CharacteristicMethotrexate (n = 19)Infliximab (n = 68)Etanercept (n = 57)All (n = 144)
  • *

    Values are the mean ± SD unless otherwise indicated. SE = shared epitope.

Women, no. (%)15 (78.9)51 (74.9)47 (82.5)112 (77.7)
Age, years55.9 ± 13.560.0 ± 12.358.5 ± 14.359 ± 12.8
Disease duration, years10.4 ± 7.214.8 ± 8.812.3 ± 11.113.2 ± 9.6
SE, no. positive/no. tested (%)13/19 (68.4)51/68 (74.6)52/57 (91.2)116/144 (80.6)

DNA preparation.

Human genomic DNA was isolated from 10 ml of heparinized blood. Mononuclear cells were isolated by isopyknic centrifugation through Ficoll-Histopaque (Sigma, St. Louis, MO) and processed through Qiagen Genomic-tips 100/G (Qiagen, Courtaboeuf, France) according to the Qiagen genomic DNA handbook. DNA was resuspended in 10 mM Tris, pH 8 and was quantified by real-time PCR with a LightCycler (Roche, Mannheim, Germany) as previously described (11).

Quantification of EBV copy number.

Approximately 500 ng of DNA from peripheral blood lymphocytes was used for EBV DNA assay by quantitative real-time PCR as described previously (5). Briefly, a 214-bp segment of the highly conserved long internal repeat region 1 of EBV was amplified by quantitative real-time PCR using hybridization probes (Tib-Molbiol, Berlin, Germany) and a LightCycler. The Raji cell line was used as an external EBV standard. Each sample was tested in duplicate and assayed twice. For final results, EBV copy numbers in test samples were calculated for 500 ng of DNA.

Statistical analysis.

The effect of treatment duration on EBV load and the link between the Disease Activity Score in 28 joints (DAS28) and EBV load were tested using generalized estimating equations (12), taking into account the interdependence of the observations made for the same individual. EBV load was the dependent variable and treatment duration or DAS28 was the independent variable.

The gamma law was used to describe the distribution of EBV load. To check the sensitivity of estimations to this hypothesis, results were compared with those obtained using the normal law as the distribution law. The link function used was the canonical link for the family (gamma or normal).

Two different correlation structures were used: exchangeable (the correlation between observations for the same individual at different times is constant) and independent (the correlation between observations for the same individual at different times is independent). The sandwich robust variance estimator was used. Analysis was performed using Stata software, version 9 (StataCorp, College Station, TX).

RESULTS

EBV loads are expressed in viral copy number per 500 ng of PBMC DNA (Table 2). A total of 739 samples were tested. For most patients the first quantification was performed after the beginning of the treatment. Results of a previous study are also given in Table 2 for comparison (135 samples from 84 patients with RA treated with various disease-modifying antirheumatic drugs [DMARDs], methotrexate, infliximab, or etanercept and 100 healthy controls were tested using the same assay as in our study, developed for the purpose of testing EBV load [5]). All quantifications for all patients are shown in Figure 1: loads at first and last quantification are given, load increase is in red, load decrease is in green, and load stability is in blue, taking into account the former load. Time 0 is the first day of therapy.

Table 2. Epstein-Barr virus (EBV) load in patients with rheumatoid arthritis receiving different treatments*
 Previous study (5)Methotrexate (n = 19)Infliximab (n = 68)Etanercept (n = 48)Etanercept, after infliximab (n = 9)All (n = 144)
 Treatment (n = 84)No treatment (controls) (n = 100)
FirstLastFirstLastFirstLastFirstLastFirstLast
  • *

    EBV load range, median, and mean are expressed in viral copy number per 500 ng of DNA and are given for the first and the last quantifications of the study. Results of a previous study (5) are given for comparison. NT = not tested.

  • Methotrexate, infliximab, etanercept, and miscellaneous disease-modifying antirheumatic drugs.

EBV range (% positive)0–185 (88)0–34 (89)1–80.2 (100)0–21.3 (58)0–186 (85.3)0–542 (69.1)0–66 (31)0–246 (64.6)0–542 (71.5)0–82.6 (85.7)0–542 (63.4)0–542 (65.3)
EBV median8.840.69.61.24.22.402.10.171.902.02.1
EBV mean15.61.918.13.912.016.34.717.740.612.613.616.4
Followup duration, mean ± SD monthsNTNT22.9 ± 13.328.3 ± 11.917.1 ± 9.016.7 ± 6.322.6 ± 12.6
Figure 1.

Variation of Epstein-Barr virus load in patients with rheumatoid arthritis receiving A, methotrexate, B, infliximab, and C, etanercept. Load is expressed in viral copy number per 500 ng of DNA. Load at first quantification is in the white squares. Load at last quantification is also given. Load increase is in red, load decrease is in green, and load stability is in blue, taking into account the former load. An empty white box means that no quantification was performed. Time 0 is the first day of therapy.

Methotrexate tends to decrease EBV load.

Nineteen patients received methotrexate. In all of these patients, EBV was detectable at the onset of the study. EBV loads ranged from 1 to 80 copies per 500 ng of DNA, the mean load was 18 copies per 500 ng of DNA, and the median load was 9.6 copies per 500 ng of DNA (Table 2).

Over time, EBV load decreased in 11 patients, was stable in 7 others, and increased slightly in 1 patient. Indeed, Figure 1A shows mostly blue squares (load stability) and green squares (load decrease). At the end of the study, EBV was detectable in only 53% of patients. EBV loads ranged from 0 to 21 copies per 500 ng of DNA, the mean load was 3.8 copies per 500 ng of DNA, and the median load was 1.2 copies per 500 ng of DNA. These data suggest that methotrexate decreases EBV load over time, almost bringing it down to the levels observed in controls (mean 1.9, median 0.6) (Table 2). Nonetheless, this trend was not significant when the data were analyzed with generalized estimating equations: P = 0.61 under the hypothesis of normal law and exchangeable correlation, P = 0.73 under normal law and independent correlation, P = 0.51 under gamma law and exchangeable correlation, and P = 0.71 under gamma law and independent correlation.

In most patients, TNF inhibitors do not influence EBV load.

Sixty-eight patients received infliximab. At the onset of the study, 85% of these patients had detectable EBV. EBV loads ranged from 0 to 186 copies per 500 ng of DNA, the mean load was 12 copies per 500 ng of DNA, and the median load was 4.2 copies per 500 ng of DNA (Table 2). EBV load was stable in 54 patients receiving treatment, decreased in 8 patients, and increased in 6 patients. Figure 1B shows slightly more green squares (load decrease) than red squares (load increase) in accordance with overall load stability. At the end of the study, 63% of patients had detectable EBV. EBV loads ranged from 0 to 542 copies per 500 ng of DNA, the mean load increased to 16 copies per 500 ng of DNA, and the median load decreased to 2.4 copies per 500 ng of DNA. In 1 patient, EBV load increased from 7 to 542 copies per 500 ng of DNA. This patient's treatment was switched to etanercept and EBV load decreased sharply afterwards. Analysis of the evolution of EBV loads over time in patients receiving infliximab, by generalized estimating equations, did not demonstrate any significant change: P = 0.16 under the hypothesis of normal law and exchangeable correlation, P = 0.57 under normal law and independent correlation, P = 0.06 under gamma law and exchangeable correlation, and P = 0.50 under gamma law and independent correlation.

Forty-eight patients received etanercept. At the onset of the study, 31% had detectable EBV. EBV loads ranged from 0 to 66 copies per 500 ng of DNA, the mean load was 4.7 copies per 500 ng of DNA, and the median load was 0 copies per 500 ng of DNA (Table 2). EBV load was stable in 33 patients receiving treatment, decreased in 6 patients, and increased in 11 patients (Figure 1). Figure 1C shows slightly more red squares (load increase) than green squares (load decrease) in accordance with overall load stability. At the end of the study, 64% of patients had detectable EBV. EBV loads ranged from 0 to 246 copies per 500 ng of DNA, the mean load increased to 17.7 copies per 500 ng of DNA, and the median load increased to 2.1 copies per 500 ng of DNA. Analysis of the evolution of EBV loads over time in patients receiving etanercept, by generalized estimating equations, did not demonstrate any significant change: P = 0.38 under the hypothesis of normal law and exchangeable correlation, P = 0.67 under normal law and independent correlation, P = 0.09 under gamma law and exchangeable correlation, and P = 0.63 under gamma law and independent correlation.

Nine patients received etanercept after infliximab and were analyzed as a separate group. In these patients, after infliximab and before etanercept treatment, EBV loads ranged from 0 to 542 copies per 500 ng of DNA, the mean load was 81.2 copies per 500 ng of DNA, and the median load was 0.17 copies per 500 ng of DNA. After etanercept treatment, EBV loads ranged from 0 to 82 copies per 500 ng of DNA, the mean load decreased to 12.6 copies per 500 ng of DNA, and the median load increased to 1.9 copies per 500 ng of DNA.

Disease activity is associated with increased EBV load.

Whenever possible, disease activity was evaluated by calculating the DAS28 score (13). Both EBV load and DAS28 score were available for 529 blood samples. We tested whether EBV loads correlated with DAS28 scores using generalized estimating equations (12). EBV load was the dependent variable and DAS28 activity was the independent variable. Two different correlation structures were used: exchangeable and independent.

We found that higher DAS28 scores were associated with higher EBV loads (Figure 2). This correlation was significant for each hypothesis (normal law or gamma law) made on the distribution of EBV loads in patients with RA receiving treatment: P = 0.033 under the hypothesis of normal law and exchangeable correlation, P = 0.0047 under normal law and independent correlation, and P ≤ 0.0001 for gamma law and either exchangeable correlation or independent correlation.

Figure 2.

Epstein-Barr virus (EBV) load was assayed in 529 blood samples from patients with rheumatoid arthritis and plotted against the Disease Activity Score in 28 joints, showing positive correlation between disease activity and EBV load in peripheral blood lymphocytes.

Other factors.

Sex, age, and HLA–DR genotype were not significantly different between the 3 groups (Table 1) and did not influence either EBV load over time (data not shown).

DISCUSSION

In patients with RA, impaired control of EBV infection results in high EBV load in PBMCs (5). This increase in EBV load (10 times that of healthy controls) is similar to that observed in healthy transplant recipients receiving immunosuppressants (7, 8) and has also been described in patients with systemic lupus erythematosus (14). Of interest, both transplant recipients and patients with RA develop lymphoma more often than healthy controls (6–8, 15). In transplant recipients, lymphoma usually contain the EBV genome and are preceded by polyclonal expansion of EBV-positive B cells. At that time, EBV load in peripheral blood B cells progressively increases. A load of 1,000 copies of EBV per 500 ng of PBMC DNA is considered a level above which transplant recipients are at increased risk of developing lymphoma (7, 8).

The EBV status of lymphoma developing in patients with RA is still controversial. A recent study of 343 lymphoma that occurred between 1964 and 1995 in patients with RA from Sweden, before TNF blockers were used, demonstrated that although most lymphoma were B cell derived, only 12% contained the EBV genome (16). This finding is consistent with our observation that methotrexate decreases EBV load over time in patients with RA (Figure 1 and Table 2). This number is very different from that observed in lymphoma developing in transplant recipients, which are almost always EBV positive. However, because TNF blockers were not used before 1995, this study does not indicate whether lymphoma developing in patients treated with TNF suppressors are similarly EBV negative. Our results shed some light on this issue. Indeed, we investigated whether long-term treatment of RA with methotrexate, a first-line DMARD, or TNF inhibitors, a second-line treatment, could impair EBV control and facilitate the emergence of EBV-positive B cell malignancies.

Long-term treatment of 19 patients with methotrexate decreased EBV load to a level observed in controls. This suggested partial EBV clearing from patients' PBMCs. Therefore, our data do not support the hypothesis that methotrexate may impair the control of EBV replication and do not explain the 4 cases of EBV-positive, immunosuppression-type lymphoma that occurred in patients with RA receiving methotrexate and that regressed after withdrawal of methotrexate (17).

Long-term followup of 128 patients with RA treated with TNF inhibitors demonstrated stability of EBV load over time in most patients. EBV load reached 542 copies per 500 ng of DNA in only 1 patient receiving infliximab and reached 250 copies per 500 ng of DNA in 1 patient receiving etanercept. These increases account for higher means in these groups. In the patient reaching 542 copies, we found no evidence of B cell proliferative disease or lymphoma.

Therefore, in most patients with RA, methotrexate and TNF inhibitors do not impair control of EBV replication as much as immunosuppressive agents in transplant recipients. However, we have to acknowledge the fact that lymphoma occurring in patients with RA treated with TNF blockers are very uncommon. Indeed, while following 18,572 patients for 3 years, Wolfe and Michaud only observed 29 cases of lymphoma, which is roughly 1 case of lymphoma in 640 patients, over 3 years (15). Therefore, our study may fail to detect a very rare population of fragile (possibly genetically) patients who are at risk of developing lymphoma when treated with methotrexate or TNF blockers.

Our study also highlights the correlation between disease activity (as assessed by the DAS28 index) and EBV load. This correlation may only reflect the fact that disease activity is associated with hyperactivation of B cells, which in turn may facilitate the replication of EBV and lead to high EBV loads. Alternatively, disease flares may be associated with lowered EBV-specific immunity. This has already been observed when studying T cell responses to gp110, a glycoprotein of the late replicative cycle whose recognition is critical in controlling EBV replication (4).

In short, we have monitored EBV load in the PBMCs of 144 patients with RA treated with methotrexate and/or TNF inhibitors. EBV load was stable over the years. The main prediction carried by this result is that lymphoma developing in patients with RA treated with TNF inhibitors should differ from EBV-positive posttransplant lymphoma.

AUTHOR CONTRIBUTIONS

Dr. J. Roudier had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Balandraud, J. Roudier, C. Roudier.

Acquisition of data. Balandraud, Guis, Auger, C. Roudier.

Analysis and interpretation of data. Balandraud, J. Roudier, C. Roudier.

Manuscript preparation. Balandraud, J. Roudier, C. Roudier.

Statistical analysis. Meynard.

Ancillary