Abstract
Infection with Epstein-Barr virus (EBV) results in lifelong infection of B cells in the peripheral blood and episodic shedding of virus from the oropharynx. We followed patients treated with rituximab (anti-CD20 monoclonal antibody) and found that several had both undetectable B cells and EBV in the blood, but shed EBV from the throat. While some models postulate that EBV traffics from the B cells in the blood to the throat where it is subsequently shed, our findings indicate that circulating EBV in B cells is not necessary for the virus to persist in, and to be shed from, the oropharynx.
Epstein-Barr virus (EBV) establishes latency after primary infection and can be detected in throat washings or saliva of at least half of healthy seropositive adults [1], and from a greater percentage of immunosuppressed persons [2]. In latently infected persons, EBV is found in memory B cells in the blood [3]; the virus reactivates in the oropharynx, releasing virus into the saliva.
The role of oropharyngeal epithelial cells in the EBV life cycle is controversial. In situ hybridization demonstrated EBV in epithelial cells from patients with infectious mononucleosis [4]. However, later studies of throat washings and tonsils from infectious mononucleosis patients and healthy persons showed that EBV was present in B cells exclusively [5]. Patients with X-linked agammaglobulinemia lack mature B cells and have no evidence of EBV infection; therefore, mature B cells are essential for acquisition and/or maintenance of EBV [6]. However, latent EBV infection has been found in tonsillar epithelial cells cultured from asymptomatic donors [7]. Asymptomatic EBV carriers have EBV strains in the blood and oropharynx that often differ [8]. These findings suggest that B cells are necessary but not sufficient for acquisition and maintenance of EBV infection, and that infection also occurs in epithelial cells.
Studies of EBV in seropositive patients receiving bone marrow transplants showed that some patients became seronegative after transplant and subsequently acquired EBV from their donor or from a close contact [9]. Most of these patients received cyclophosphamide and total body irradiation. These findings imply that EBV resides in the hematopoietic cells that are ablated by the bone marrow transplantation process, either by cytotoxic chemotherapy, irradiation, or graft-versus-host reactivity. Epithelial cells were presumed to remain after chemotherapy, and any EBV still in these cells was insufficient to continue the latent infection.
Here, we investigated whether selective depletion of circulating B cells results in loss or reduction of EBV shedding from the oropharynx. Rituximab, a monoclonal antibody against CD20, causes rapid and profound depletion of B cells from the blood. Its half-life varies from days to weeks, and it can be detected in the serum months after administration [10]. In human studies using standard (375 mg/m2) or higher single doses of rituximab, B cell numbers are markedly reduced, but are still detected in lymph nodes [11, 12]. Here, we measured the level of EBV in throat washings and blood in patients receiving rituximab.
PATIENTS, MATERIALS, AND METHODS
Patient selection
Patients receiving rituximab (10 mg/kg) for lymphoma or cryoglobulinemia at the National Institutes of Health (NIH) Clinical Center were enrolled in the study. Informed consent was obtained from patients, and the study was approved by the Institutional Review Board of the National Institute of Allergy and Infectious Diseases.
Sample collection
Throat washes were collected by having the patient gargle for 10 seconds with 2 separate 10 ml aliquots of preservative-free normal saline. The 20 ml sample was centrifuged at low speed, and the pellets and supernatants stored separately at −70°C. Heparinized blood was collected and peripheral blood mononuclear cells (PBMC) were separated by Ficoll-Hypaque gradient centrifugation and stored at −70°C.
EBV DNA detection in throat wash and blood samples
DNA was extracted from throat wash pellets using the Easy DNA Kit (Invitrogen, Carlsbad, CA). Quantitative real-time PCR amplification was performed with 100 ng of each DNA sample using a TaqMan PCR kit and a Model 7700 Sequence Detector (Applied Biosystems, Foster City, CA). Primers were obtained from Invitrogen and fluorescent probes from Synthegen (Houston, TX). Primers EBVW-F1 (5′ GGACCACTGCCCCTGGTATAA 3′) and EBVW-R2 (5′ TTTGTGTGGACTCCTGGGG 3′) were used to amplify the BamH1 W region of the EBV genome and the product was detected with fluorogenic probe EBVW (5′ [6FAM]-TCCTGCAGCTATTTCTGGTCGCATCA-[TAMRA] 3′). Primers Bcl2-F (5′ CCTGCCCTCCTTCCGC 3′) and Bcl2-R (5′ TGCATTTCAGGAAGACCCTGA 3′) were used to amplify the human diploid Bcl2 gene as a control and the PCR product was detected with a Bcl2 probe (5′ [6FAM]-CTTTCTCATGGCTGTCC-[TAMRA] 3′). Each experiment included standard curves made with dilutions of plasmid DNA containing the EBV BamH1 W or Bcl2 sequences. The number of EBV BamHI W copies per cell (N) was calculated as N = (2 × W)/B, where W = EBV BamH1 W copy number, and B = Bcl2 copy number. The limit of detection of the reaction was 5 copies per well. Values for EBV DNA in throat washes were expressed as number of copies per 106 cells. EBV detection in blood samples was performed by the Department of Laboratory Medicine of the NIH Clinical Center as described previously [13]. Values were expressed as number of copies per 106 PBMCs, with the lower limit of detection varying when patients had very low total PBMC counts.
Flow cytometry
Aliquots of PBMCs were thawed, washed, incubated with antibodies to CD19 or CD20 (BD Biosciences, San Jose, CA), and analyzed by flow cytometry using a FACSCalibur (BD Biosciences).
RESULTS
Patients receiving rituximab have EBV DNA in throat washes
In the first study, we determined whether EBV DNA could be detected in throat washes in patients who had recently received rituximab. We recruited patients already receiving rituximab as part of their therapy for lymphoma. Most treatment protocols included administration of rituximab at 3-week intervals along with cycles of etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin (EPOCH-R). Because of the rapid effect and long half-life of rituximab, we reasoned that samples could be obtained at any point within the treatment cycle. We obtained 1 to 3 throat washes and blood samples from each of 5 patients, with the timing determined by patient convenience. Samples were obtained after patients had received 1 to 5 cycles of rituximab therapy. We used real-time PCR to detect EBV DNA in cell pellets obtained by centrifugation of throat washes, since cell pellets have a greater sensitivity to detect EBV than throat wash supernatants [2]. Three of 5 patients showed detectable EBV DNA in their throat washes, 1 to 21 days after their last rituximab doses (Table 1). Of the two patients who did not shed EBV DNA, one (patient 3) was known to be EBV seropositive, while the other (patient 1) was not tested. No patient had a detectable level of EBV DNA in PBMCs while receiving rituximab. Five samples, from two patients receiving rituximab, had ≤0.3% CD19+ B cells, indicating that rituximab had effectively depleted B cells (data not shown).
Table 1.
Shedding of EBV from the oropharynx in patients receiving rituximab in the first study.
| Patient no. | Diagnosis | Treatment | Week of RTX therapy | EBV copies/106 PBMC* | EBV copies/106 throat wash cells |
|---|---|---|---|---|---|
| 1 | DLBCL | EPOCH-R | Pre | <100 | |
| 3 | 0 | ||||
| 4 | 0 | ||||
| 6 | <90 | 0 | |||
| 2 | Burkitt | EPOCH-R | Pre | 520 | |
| lymphoma | |||||
| 9 | 34,300,000 | ||||
| 12 | <40 | 34,300,000 | |||
| 14 | <140 | 172,000,000 | |||
| 3 | B cell | EPOCH-R | Pre | <160 | |
| lymphoma | 0 | ||||
| 1 | 0 | ||||
| 4 | 0 | ||||
| 4 | DLBCL | EPOCH-R | Pre | 93,000 | |
| 3 | <400 | 31,000,000 | |||
| 4 | <430 | ||||
| 5 | DLBCL | EPOCH-R | 5 | 672,000,000 | |
| HIV | 6 | <300 |
Abbreviations: DLBCL, diffuse large B-cell lymphoma; EPOCH-R, etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin, rituximab; HIV, human immunodeficiency virus
Values with a “less than” sign were undetectable. The detection limit varied with the patient’s total white blood cell count on that day.
Amount of EBV DNA shed from the oropharynx varies among patients before and after rituximab
Having detected EBV DNA in throat washes in patients receiving rituximab, we determined whether the amount of shedding changes in response to therapy. We initiated a second study of 12 patients who had not yet started rituximab therapy and collected blood and throat washings on at least one occasion prior to therapy, at least once during each cycle of therapy, and 3 months after the last cycle.
Before the start of therapy, the level of EBV DNA in the oropharynx varied among patients, as has been observed previously [2, 4]. Seven patients (patients 6, 7, 8, 9, 11, 14, 15) who shed EBV before therapy continued to shed EBV episodically at high levels (>1,000 EBV DNA copies/106 throat wash cells) during rituximab therapy (Table 2). One patient (patient 11) maintained high levels of shedding at followup visits up to 8 weeks after the last rituximab dose. One patient who shed EBV before therapy (patient 12) shed low levels of EBV (<200 EBV copies/106 throat wash cells) after therapy. Two patients (patients 10, 13) who had no detectable shedding before therapy shed low levels of EBV (≤300 EBV copies/106 throat wash cells) after receiving rituximab. One patient (patient 17) shed no detectable EBV throughout the study, but was found to be seropositive for EBV. All of these patients were receiving cytotoxic chemotherapy in addition to rituximab and the chemotherapy might have affected EBV shedding. Since cytotoxic chemotherapy might either reduce shedding (by killing epithelial cells), or enhance shedding (by reducing immunosurveillance by cytotoxic T cells or reactivating EBV DNA from latently infected B cells), we studied a patient who was receiving rituximab, but not cytotoxic chemotherapy. Patient 16 received rituximab for cryoglobulinemia and was also found to shed EBV from the oropharynx during therapy.
Table 2.
Shedding of Epstein-Barr virus from the oropharynx, and presence of CD20+ B cells in the blood before and after rituximab (RTX) therapy in the second study.
| Patient no. | Diagnosis | Treatment | Week of RTX therapy | EBV copies/106 PBMC* | EBV copies/106 throat wash cells | % CD20+ lymphocytes | Total lymphocyte count |
|---|---|---|---|---|---|---|---|
| 6 | HIV, DLBCL | EPOCH-R | Pre | 590 | 49,800,000 | 8.61 | 1,480 |
| 1 | <90 | 0 | |||||
| 3 | 0 | 0.09 | 1,210 | ||||
| 6 | 23,800 | 0 | 1,650 | ||||
| 7 | 1,270 | ||||||
| 8 | <100 | 0 | |||||
| 11 | <80 | 0 | |||||
| Post | <110 | ||||||
| Post | <110 | 3.42 | 949 | ||||
| 7 | DLBCL | EPOCH-R | Pre | <70 | 3 | 7.19 | 420 |
| 6 | <100 | 90,300 | 0.26 | 1,380 | |||
| 7 | 75 | ||||||
| 10 | <170 | 2 | |||||
| 13 | 542 | 0.12 | 1,080 | ||||
| 15 | <110 | 0 | |||||
| Post | <90 | 8 | |||||
| Post | <100 | 396 | 0.04 | 835 | |||
| 8 | Hodgkin lymphoma | EPOCH-R | Pre | <70 | 7 | 7.63 | 792 |
| 4 | <80 | 1,640 | 0.23 | 1,140 | |||
| 8 | <120 | 94 | 0.28 | 138 | |||
| 10 | <170 | 592 | 0.74 | 576 | |||
| 13 | 832,000 | 0.1 | 662 | ||||
| Post | <180 | 1 | 0.18 | 477 | |||
| Post | <100 | 55 | 0.19 | 220 | |||
| 9 | DLBCL | EPOCH-R | Pre | 80 | 16,500 | 8.66 | 2,390 |
| 3 | <70 | 4,800 | 0.17 | 2,530 | |||
| 6 | <200 | 4 | 1,550 | ||||
| 9 | 410 | 1,250 | |||||
| 15 | <90 | 77,100 | 0 | 929 | |||
| Post | <70 | ||||||
| Post | <200 | 3 | 0.03 | 2,010 | |||
| 10 | Follicular lymphoma | EPOCH-R | Pre | 0 | |||
| 3 | <110 | 0 | 0.02 | 1,550 | |||
| 6 | 19 | 0 | 1,670 | ||||
| 9 | <220 | 0 | |||||
| 12 | <80 | 0 | |||||
| Post | <70 | 4 | |||||
| Post | 119 | 0 | 2,340 | ||||
| 11 | Mantle cell lymphoma | B+EPOCH-R | Pre | 20 | 860 | 76.39 | 804 |
| 19,300 | 58.86 | 165 | |||||
| 11,600 | |||||||
| 3 | <90 | ||||||
| 4 | 3,540 | ||||||
| 6 | <160 | 68.45 | 812 | ||||
| 9 | <150 | 234 | |||||
| 12 | <100 | 1,180,000 | |||||
| 13 | 56,100 | ||||||
| 15 | <90 | 43,500 | |||||
| Post | 387,000 | 42.35 | 612 | ||||
| Post | <160 | 47,000 | |||||
| Post | 270 | 1,430 | 12.06 | 1,530 | |||
| 0.19 | 1,210 | ||||||
| 12 | Burkitt lymphoma | EPOCH-R | Pre | 950 | 1,520 | ||
| 3 | 0 | ||||||
| 6 | <130 | 0 | |||||
| 9 | <170 | 0 | |||||
| 12 | <210 | 0 | 0.02 | 3,360 | |||
| 15 | <100 | 0 | |||||
| Post | <120 | 165 | |||||
| Post | <150 | 0 | 5.4 | 401 | |||
| 13 | DLBCL | EPOCH-R | Pre | <110 | 0 | 0 | 627 |
| 3 | <90 | 0 | |||||
| 8 | <130 | 300 | |||||
| 11 | <290 | 0 | |||||
| 14 | <140 | 0 | |||||
| 17 | <90 | 0 | |||||
| Post | <150 | ||||||
| Post | <180 | ||||||
| Post | 0 | 0 | 3.1 | 732 | |||
| 0 | 955 | ||||||
| 14 | Mantle cell lymphoma | Pre | 280 | 26,000 | 0 | 754 | |
| 50,200 | |||||||
| 3 | <420 | 13,400 | |||||
| 5 | <130 | 5,520 | |||||
| 7 | <210 | 0 | |||||
| 10 | <110 | 0 | |||||
| 13 | <120 | 0 | |||||
| Post | <120 | 0 | |||||
| Post | <90 | 0 | 62.4 | 966 | |||
| 1,420 | |||||||
| 15 | Mantle cell lymphoma | B+EPOCH-R | Pre | <80 | 0 | ||
| 2,750 | |||||||
| 388 | 10.1 | 1,130 | |||||
| 199 | |||||||
| 3 | 230 | 0 | |||||
| 3 | 0 | ||||||
| 6 | <60 | 0 | |||||
| 9 | <100 | 1,370 | 956 | ||||
| 12 | <60 | 0 | |||||
| 15 | <80 | 0 | |||||
| Post | 0 | 0 | 626 | ||||
| Post | 0 | 0 | 695 | ||||
| 16 | Cryoglobulinemia | Rituximab | Pre | <70 | 0 | 1,940 | |
| 1 | <110 | 37 | |||||
| 2 | 798 | ||||||
| 3 | <80 | 0 | |||||
| Post | 0 | ||||||
| 17 | Burkitt lymphoma | EPOCH-R | Pre | <130 | 0 | ||
| 3 | <130 | 0 | |||||
| 5 | 0 | ||||||
| 6 | <60 | ||||||
| 8 | <70 | 0 | |||||
| 11 | <80 | 0 | |||||
| Post | 0 | 0 | |||||
| Post | 0 | 0 | |||||
| Post | 0 | 0 |
Abbreviations: DLBCL, diffuse large B-cell lymphoma; EPOCH-R, etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin, rituximab; B+EPOCH-R, bortezomib followed by EPOCH-R; HIV, human immunodeficiency virus; RTX, rituximab
Values with a “less than” sign were undetectable. The detection limit varied with the patient’s total white blood cell count on that day.
Patients receiving rituximab have EBV DNA in throat washes in the absence of detectable circulating B cells
CD20+ B cells were measured in the peripheral blood in 10 patients before and during rituximab therapy. In most patients rituximab rapidly reduced and virtually eliminated circulating CD20+ B cells after the first dose, and eliminated detectable EBV DNA in PBMCs (Table 2). Four patients continued to shed high levels of EBV from the throat (>1,000 EBV DNA copies/106 throat wash cells) and had no detectable EBV DNA in their PBMCs and no or very few (<0.3%) circulating CD20+ cells during rituximab therapy (patients 6, 7, 9, 14 [Table 2]). Three patients who shed low levels of EBV from the throat (≤300 EBV copies/106 throat wash cells) after rituximab had no detectable EBV DNA in PBMCs, and low levels (0.2%) of circulating CD20+ B cells during rituximab therapy (patients 10, 12, 13). Patient 8 shed high levels of EBV from the throat, had no detectable EBV in PBMCs, but had somewhat higher levels of CD20+ B cells (0.74%) in the blood while receiving rituximab. Two patients (patients 11, 15) with mantle cell lymphoma continued to have substantial numbers of CD20+ B cells in the peripheral blood despite rituximab therapy. To determine if EBV DNA in throat washes was in virions or was released as free viral DNA from infected cells, supernatant fractions corresponding to cell pellets were used for PCR directly, or were treated with DNase to digest free DNA, heated to inactivate DNAse, and PCR was performed. Only one sample had detectable EBV DNA, of which only 20% was resistant to DNAse and therefore was encapsidated in virions indicative of virus replication (data not shown).
DISCUSSION
We have shown that patients receiving rituximab therapy shed EBV DNA from the oropharynx at times when they have no or very few detectable circulating CD20+ B cells. In contrast to previous results with myeloablative therapy followed by bone marrow transplantation, no cases were found in which EBV shedding was completely eradicated by rituximab. Our results suggest that EBV persists in a cell population that is not depleted by rituximab, such as oropharyngeal epithelial cells or B cells within lymphoid tissue, in which it can survive until recovery of circulating B cells occurs.
Detection of similar EBV strains in peripheral blood lymphocytes and in oral hairy leukoplakia lesions from the same patient suggests that EBV may be spread from the blood to the oropharynx [14]. At least three pathways have been proposed for transit of EBV from cells circulating in the blood to epithelial cells where productive infection occurs [14, 15]. First, EBV could traffic from B cells in the blood to the oropharynx where the virus reactivates and infects epithelial cells. Second, EBV in pre-Langerhans cells circulating in the blood might traffic to the oropharynx where they mature into EBV-infected Langerhans cells and infect epithelial cells [14]. Third, EBV in circulating B cells may infect circulating monocytes, which might subsequently infect epithelial cells [15].
In addition to circulating EBV-infected cells trafficking to the oropharynx, EBV might persist in B cells in lymphoid tissues in the throat. Since rituximab does not entirely deplete B cells from lymph nodes [11, 12], persistence of EBV-infected B cells in lymphoid tissue may allow the virus to persist during rituximab therapy. EBV is present at similar levels in memory B cells in the peripheral blood and in the adenoids and tonsils of Waldeyer’s ring [3]; therefore mucosal lymphoid tissues in the oropharynx could be a site of persistent infection during rituximab therapy. While we cannot presently determine whether EBV persistence in the oropharynx is a result of trafficking to the site by non-B cells and/or persistence of EBV-infected B cells in oropharyngeal lymphoid tissues, our results indicate that circulating EBV-infected B cells are not required for persistent shedding of EBV from the oropharynx.
Acknowledgments
This study was supported by the intramural research program of the National Institute of Allergy and Infectious Diseases, National Cancer Institute, and the NIH Clinical Center. We thank Margaret Brown and Thomas Fleisher for assistance with flow cytometry and Michael Sneller, Margaret Shovlin, and Therese White for helping with patient recruitment.
Footnotes
The authors report no conflicts of interest.
References
- 1.Ling PD, Lednicky JA, Keitel WA, et al. The dynamics of herpesvirus and polyomavirus reactivation and shedding in healthy adults: a 14-month longitudinal study. J Infect Dis. 2003;187:1571–80. doi: 10.1086/374739. [DOI] [PubMed] [Google Scholar]
- 2.Haque T, Crawford DH. PCR amplification is more sensitive than tissue culture methods for Epstein-Barr virus detection in clinical material. J Gen Virol. 1997;78 ( Pt 12):3357–60. doi: 10.1099/0022-1317-78-12-3357. [DOI] [PubMed] [Google Scholar]
- 3.Laichalk LL, Hochberg D, Babcock GJ, Freeman RB, Thorley-Lawson DA. The dispersal of mucosal memory B cells: evidence from persistent EBV infection. Immunity. 2002;16:745–54. doi: 10.1016/s1074-7613(02)00318-7. [DOI] [PubMed] [Google Scholar]
- 4.Sixbey JW, Nedrud JG, Raab-Traub N, Hanes RA, Pagano JS. Epstein-Barr virus replication in oropharyngeal epithelial cells. N Engl J Med. 1984;310:1225–30. doi: 10.1056/NEJM198405103101905. [DOI] [PubMed] [Google Scholar]
- 5.Karajannis MA, Hummel M, Anagnostopoulos I, Stein H. Strict lymphotropism of Epstein-Barr virus during acute infectious mononucleosis in nonimmunocompromised individuals. Blood. 1997;89:2856–62. [PubMed] [Google Scholar]
- 6.Faulkner GC, Burrows SR, Khanna R, Moss DJ, Bird AG, Crawford DH. X-Linked agammaglobulinemia patients are not infected with Epstein-Barr virus: implications for the biology of the virus. J Virol. 1999;73:1555–64. doi: 10.1128/jvi.73.2.1555-1564.1999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.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. doi: 10.1128/JVI.78.22.12613-12624.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sitki-Green D, Covington M, Raab-Traub N. Compartmentalization and transmission of multiple epstein-barr virus strains in asymptomatic carriers. J Virol. 2003;77:1840–7. doi: 10.1128/JVI.77.3.1840-1847.2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gratama JW, Oosterveer MA, Lepoutre JM, et al. Serological and molecular studies of Epstein-Barr virus infection in allogeneic marrow graft recipients. Transplantation. 1990;49:725–30. doi: 10.1097/00007890-199004000-00014. [DOI] [PubMed] [Google Scholar]
- 10.Vieira CA, Agarwal A, Book BK, et al. Rituximab for reduction of anti-HLA antibodies in patients awaiting renal transplantation: 1. Safety, pharmacodynamics, and pharmacokinetics. Transplantation. 2004;77:542–8. doi: 10.1097/01.tp.0000112934.12622.2b. [DOI] [PubMed] [Google Scholar]
- 11.Maloney DG, Liles TM, Czerwinski DK, et al. Phase I clinical trial using escalating single-dose infusion of chimeric anti-CD20 monoclonal antibody (IDEC-C2B8) in patients with recurrent B-cell lymphoma. Blood. 1994;84:2457–66. [PubMed] [Google Scholar]
- 12.Genberg H, Hansson A, Wernerson A, Wennberg L, Tyden G. Pharmacodynamics of rituximab in kidney allotransplantation. Am J Transplant. 2006;6:2418–28. doi: 10.1111/j.1600-6143.2006.01497.x. [DOI] [PubMed] [Google Scholar]
- 13.Feng WH, Cohen JI, Fischer S, et al. Reactivation of latent Epstein-Barr virus by methotrexate: a potential contributor to methotrexate-associated lymphomas. J Natl Cancer Inst. 2004;96:1691–702. doi: 10.1093/jnci/djh313. [DOI] [PubMed] [Google Scholar]
- 14.Walling DM, Ray AJ, Nichols JE, Flaitz CM, Nichols CM. Epstein-Barr virus infection of Langerhans cell precursors as a mechanism of oral epithelial entry, persistence, and reactivation. J Virol. 2007;81:7249–68. doi: 10.1128/JVI.02754-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tugizov S, Herrera R, Veluppillai P, Greenspan J, Greenspan D, Palefsky JM. Epstein-Barr virus (EBV)-infected monocytes facilitate dissemination of EBV within the oral mucosal epithelium. J Virol. 2007;81:5484–96. doi: 10.1128/JVI.00171-07. [DOI] [PMC free article] [PubMed] [Google Scholar]