Virus Genes

, Volume 41, Issue 2, pp 174–180

Roles of TRAF2 and TRAF3 in Epstein-Barr virus latent membrane protein 1-induced alternative NF-κB activation

Article

DOI: 10.1007/s11262-010-0505-4

Cite this article as:
Song, YJ. & Kang, MS. Virus Genes (2010) 41: 174. doi:10.1007/s11262-010-0505-4

Abstract

Epstein-Barr virus (EBV) latent membrane protein 1 (LMP1)-induced NF-κB activation is essential for EBV-transformed B cell survival. LMP1 has two C-terminal cytoplasmic domains referred to as C-Terminal Activation Regions (CTAR) 1 and 2 that activate the alternative and canonical NF-κB pathways, respectively. While CTAR2 activates TRAF6, IKKβ and IKKγ-dependent canonical NF-κB pathway, CTAR1 interacts with TRAF2 and TRAF3 and activates NIK and IKKα-dependent alternative NF-κB pathway involving p100 processing into functional p52. Using IKKα−/−, IKKβ−/−, IKKγ−/−, TRAF2−/−, TRAF3−/−, TRAF6−/−, and NIKaly/aly mouse embryonic fibroblasts (MEFs), potential roles of these proteins in LMP1-induced alternative NF-κB activation were investigated. Deficiency in IKKα or functional NIK, but not in IKKβ, IKKγ, or TRAF6, severely impaired LMP1-induced p100 processing. Notably, p100 was constitutively processed in TRAF2−/− or TRAF3−/− MEFs independently of LMP1 suggesting that TRAF2 or TRAF3 may play a regulatory role in p100 processing. Subsequently, TRAF2 or TRAF3 over-expression in HEK293 cells significantly blocked LMP1-induced p100 processing. The LMP1 CTAR1 expression in 293HEK cells activated the alternative p65/p52 complex while CTAR2 failed to do so. Taken together, LMP1 activates alternative NF-κB pathway through functional NIK and IKKα that is regulated by TRAF2 or TRAF3.

Keywords

Epstein-Barr virus Latent membrane protein 1 NF-κB p100 processing 

Abbreviations

EBV

Epstein-Barr virus

TNF

Tumor necrosis factor

TNFR

TNF receptor

TRAF

TNFR-associated factor

Introduction

NF-κB is a family of transcription factors that mediate a wide range of cellular functions including cell proliferation, differentiation, and apoptosis (reviewed in [1, 2]). The NF-κB family includes RelA (p65), RelB, c-Rel, p105/p50 (NF-κB1), and p100/p52 (NF-κB2) that form homo or heterodimers to transactivate gene expression. In a canonical pathway for NF-κB activation, the p65/p50 complexes are retained in the cytoplasm by inhibitor of κB (IκB) proteins. Upon activation, IκB proteins are phosphorylated by IκB kinase β (IKKβ) and degraded via the ubiquitin–proteasome pathway allowing the p65/p50 complexes to translocate into the nucleus. An alternative pathway for NF-κB activation involves NF-κB inducing kinase (NIK)- and IKKα-mediated proteolytic processing of p100 to produce p52 and nuclear translocation of the RelB/p52 complexes.

Members of the Tumor Necrosis Factor (TNF) Receptor (TNFR)-associated factors (TRAFs) are adaptor proteins that interact directly or indirectly with members of the TNFR superfamily (reviewed in [3]). TRAFs function as signal transducers to activate transcription factors of the NF-κB and AP1 family upon TNFR ligation. To date, seven members (TRAF1 to 7) of the TRAF family have been identified in mammals.

Epstein-Barr virus (EBV) latent infection membrane protein 1 (LMP1) is essential for EBV-infected B lymphocyte conversion to proliferating lymphoblastoid cell lines (LCLs), and functionally mimics CD40, a member of the TNFR superfamily (reviewed in [4]). LMP1 consists of a 24 amino acid (aa) cytoplasmic N-terminus, six hydrophobic transmembrane domains (aa 25–186), and a 200 aa cytoplasmic C-terminus (aa 187–386). LMP1 self-aggregates in plasma membrane lipid rafts through transmembrane domains and constitutively activate NF-κB, p38 Mitogen-Activated Protein Kinase (MAPK), and c-Jun N-terminal Kinase through two C-terminal cytoplasm signaling domains referred to as C-Terminal Activation Regions (CTAR) 1 and 2 [5, 6, 7, 8, 9, 10, 11, 12]. CTAR1 engages TRAF1, 2, 3, and 5 through a consensus PXQXT motif found in the CD40 and activates the alternative pathway for NF-κB activation. CTAR2 engages TNFR-associated death domain protein (TRADD) and Receptor Interacting Protein 1 (RIP), and activate the canonical pathway for NF-κB activation [13, 14, 15, 16, 17]. Since NF-κB activation is essential for EBV–LCL survival [18, 19], delineation of LMP1-induced NF-κB activation pathway may contribute to the discovery of potential inhibitor(s) to treat EBV-associated cancers.

This study was designed to delineate the mechanism of LMP1-induced alternative NF-κB activation. Using various knock out (KO) mouse embryonic fibroblasts (MEFs), we investigated the roles of TRAF2, TRAF3, TRAF6, NIK, IKKα, IKKβ, and IKKγ in LMP1-induced p100 processing and found previously unrecognized regulatory roles of TRAF2 and TRAF3 in LMP1-induced alternative NF-κB activation. In addition, we found that LMP1–CTAR1-induced NF-κB activation comprises the alternative p65/p52 complex.

Materials and methods

Cells, retrovirus, plasmids, and transfections

TRAF3−/− MEF was a gift from Dr. Genhong Cheng (UCLA), and IKKα−/−, IKKβ−/−, IKKγ−/−, TRAF2−/− and TRAF6−/− MEFs was previously described [20]. NIKaly/aly MEF was a gift from Dr. Tasuku Honjo (Kyoto University). Retrovirus expressing GFP or LMP1 has been previously described [21]. The pcDNA3FLAG-LMP1 wild type (WT), 1-231 only (CTAR1), and Δ187-351(CTAR2) have been previously described [21]. Effectene for transient transfection was used according to the manufacturer’s directions (Qiagen, Valencia, CA).

Sub-cellular fractionation and western blot analysis

Cells were collected, fractionated, and transferred to nitrocellulose membranes as previously described [22]. Polyclonal rabbit antibody to p100/p52 was a kind gift from Dr. Ulrich Siebenlist (NIH). Antibodies to p65 and alpha–tubulin were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and Sigma-Aldrich (St. Louis, MO), respectively. Enhanced chemiluminescence detection reagents (Pierce, Rockford, IL) and secondary peroxidase-labeled anti-mouse or anti-rabbit immunoglobulin G antibody (Amersham Biosciences, Piscataway, NJ) were used according to the manufacturer’s directions.

Immunoprecipitation

Five million cells were lysed with NP-40 lysis buffer [25 mM Tris–HCl (pH 7.5), 150 mM NaCl, 2 mM EDTA, 1% NP-40] containing 50 mM sodium fluoride, 1 mM sodium orthovanadate, and protease inhibitor cocktail (Roche, Indianapolis, IN). Cell lysates were precleared with protein A/G agarose beads (Santa Cruz, CA) and incubated at 4°C for overnight with anti-p65 antibody (Santa Cruz). Immune complexes were collected on protein A/G agarose beads and washed three times with NP-40 lysis buffer. Immunoprecipitates were eluted by addition of an equal volume of 2× sodium dodecyl sulfate–polyacrylamide gel electrophoresis loading buffer [100 mM Tris–HCl (pH 6.8) containing 4% SDS, 0.02% bromophenol blue, and 2% ß-mercaptoethanol] and subjected to western blot analysis with either anti-p100/p52 antibody or anti-p65 antibody.

Results

NIK and IKKα positively regulate LMP1-induced p100 processing to p52

In order to determine roles of IKKα and NIK in LMP1-induced p100 processing to p52, embryonic fibroblasts from wild type (WT), IKKα−/−, or alymphoplasia (aly) NIKaly/aly mice, a natural strain with a mutant NIK gene, were transiently transfected with pcDNA3–LMP1 expression vectors, and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 antibody at 48 h after transfection. In WT MEFs, LMP1 significantly induced p100 processing (Fig. 1a, compare lane 2 with lane 1). In contrast, LMP1-induced p100 processing to p52 was significantly impaired in IKKα−/− (Fig. 1a, compare lane 4 with lane 3) or NIKaly/aly MEFs (compare lane 6 with 5) agreeing with previous reports that suggest IKKα and NIK are critical for LMP1-induced p100 processing [23, 24, 25]. Consistent with previous reports that the canonical NF-κB activation induces p100 expression [23, 26], the increased p100 expression by LMP1 was detected in WT and IKKα−/−, but not in NIKaly/aly MEFs possibly due to cell senescence in this particular experiment (Fig. 1a, compare lanes 2, 4, and 6 with lanes 1, 3, and 5). The basal p100 was almost undetectable in WT, IKKα−/−, and NIKaly/aly MEFs (Fig. 1a, lanes 1, 3, and 5). However, endogenous p100 was clearly detected in other experiments (Supplementary Figure 1, lanes 1, 3, and 5). In any case, LMP1 failed to induce p100 processing to p52 in NIKaly/aly MEFs.
Fig. 1

LMP1-induced p100 processing in IKKα−/−, NIKaly/aly, IKKβ−/−, and IKKγ−/− MEFs. a Control WT (lanes 1 and 2), IKKα−/− (lanes 3 and 4), or NIKaly/aly (lanes 5 and 6) MEFs were transfected with either pSG5 () or pSG5-FLAG-LMP1 (+). After 48 h, equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 or anti-LMP1 antibody. b Control WT (lanes 1 and I), IKKβ−/− (lanes 3 and 4), or IKKγ−/− (lanes 5 and 6). MEFs were transduced with retroviruses expressing either GFP (−) or LMP1 (+), and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 (US), anti-LMP1, or anti-tubulin antibody at 6 days after transduction. The star denotes an induced p52 from p100 by LMP1. Additional non-specific (ns) bands were detected using polyclonal anti-p100/p52 antibody. These ns bands may represent incompletely processed p100 [43] or a ns binding of antibodies generated by insufficient blocking and/or washing of the membrane

In order to confirm the result obtained using transient transfection, WT, IKKα−/−, or NIKaly/aly MEFs were transduced with retroviruses expressing either GFP (−) or LMP1(+), and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 antibody at 6 days after transduction. Similar results to those obtained using transient transfection were obtained using retrovirus transduction, indicating critical roles of IKKα and NIK in LMP1-induced p100 processing (Supplementary Figure 1). Unlike in transient-transfected WT MEFs expressing LMP1 for 2 days, p100 rapidly underwent processing to p52 in retrovirus transduced WT MEFs expressing LMP1 for 6 days (Supplementary Figure 1, lane 2).

In order to further determine roles of IKKβ and IKKγ in LMP1-induced p100 processing, WT, IKKβ−/−, or IKKγ−/− MEFs were transduced with retroviruses expressing either GFP or LMP1, and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 antibody at 6 days after transduction. In IKKβ−/− or IKKγ−/− MEFs, LMP1-induced p100 processing to p52, despite only 25–50% of LMP1 expression in WT MEFs due to low transduction efficiency (Fig. 1b, compare lane 2 with lanes 4 and 6). Since IKKγ is dispensable for LMP1-induced canonical NF-κB activation [20], LMP1-induced p100 expression by activating a canonical NF-κB pathway in IKKγ−/− MEFs (Fig. 1b, compare lane 6 with lane 5). In this particular experiment, LMP1 did not strongly induce p100 expression in WT MEFs probably due to cell senescence (Fig. 1b, compare lane 2 with lane 1). Taken together, these data suggest that LMP1-induced p100 processing is dependent on NIK-IKKα, but not IKKβ-IKKγ.

TRAF2 and TRAF3 regulate LMP1-induced p100 processing

LMP1–CTAR1 interacts with TRAF2 and TRAF3 and mediates p100 processing. In order to investigate whether TRAF2 or TRAF3 plays an important role in LMP1-induced p100 processing, WT, TRAF2−/−, and TRAF3−/− MEFs were transduced with retroviruses expressing either GFP or LMP1, and p100 processing to p52 was determined. In WT MEFs, LMP1 expression strongly induced p100 processing (Fig. 2a, compare lane 2 with lane 1). In TRAF2−/− and TRAF3−/− MEFs, p100 was constitutively processed to p52 in the absence of LMP1 as previously reported (Fig. 2a, compare lanes 3 and 5 with lane 1) [27, 28, 29]. LMP1 failed to further induce p100 processing in TRAF2−/− and TRAF3−/− MEFs (Fig. 2a, compare lanes 4 and 6 with lanes 3 and 5) indicating that TRAF2 or TRAF3 may have an effect on LMP1-induced p100 processing. In order to further determine the role of TRAF2 or TRAF3 in LMP1-induced p100 processing, HEK293 cells were co-transfected with LMP1 and FLAG-tagged TRAF2 or TRAF3 expression vectors, and p100 processing was determined at 48 h after transfection. LMP1-induced p100 processing was significantly inhibited by the over-expression of TRAF2 or TRAF3 (Fig. 2b, compare lane 2 with lanes 4 and 6). The basal p100 level was decreased in TRAF2 or TRAF3 overexpressing cells due to uneven loading of samples confirmed by tubulin detection (Fig. 2b, compare lanes 1 and 2 with lanes 3–6). These data indicate that both TRAF2 and TRAF3 play negative roles in LMP1-induced p100 processing.
Fig. 2

The effect of TRAF2 or TRAF3 on LMP1-induced p100 processing. a Control WT (lanes 1 and 2), TRAF3−/− (lanes 3 and 4), or TRAF2−/− (lanes 5 and 6). MEF cells were transduced with retroviruses expressing either GFP (−) or LMP1 (+), and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52 antibody at 6 days after transduction. b HEK293 cells were co-transfected with either pSG5 (−) or pSG5-FLAG-LMP1 (+) and pcDNA3 (−) (lanes 1 and 2), pcDNA3-FLAG-TRAF2 (+) (lanes 3 and 4), or pcDNA3-FLAG-TRAF3 (+) (lane 5 and 6). After 48 h, equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52, anti-LMP1, or anti-tubulin antibody. The star denotes an induced p52 from p100 by LMP1. (ns non-specific)

TRAF6 is not required for LMP1-induced p100 processing

TRAF6 is essential for CTAR2-mediated NF-κB activation [20, 30]. Since TRAF6 can induce p100 processing [31], the potential role of TRAF6 in LMP1-induced p100 processing was determined. WT and TRAF6−/− MEFs were transduced with retroviruses expressing either GFP or LMP1, and p100 processing to p52 was determined at 6 days after transduction. In both WT and TRAF6−/− MEFs, LMP1 induced similar levels of p100 processing to p52 (Fig. 3a, compare lanes 1 and 2 with lanes 3 and 4). Thus, TRAF6 is not required for LMP1-induced p100 processing.
Fig. 3

The effect of TRAF6 on LMP1-induced p100 processing. a Control WT (lanes 1 and 2) or TRAF6−/− (lanes 3 and 4). MEFs were transduced with retroviruses expressing either GFP (−) or LMP1 (+), and equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52, anti-LMP1, or anti-tubulin antibody at 6 days after transduction. Cell extracts from p100−/− MEFs were used as a negative control to locate p100 and p52. b HEK293 cells were co-transfected with either pSG5 (−) or pSG5-FLAG-LMP1 (+) and pcDNA3 (−) (lanes 1 and 2) or pcDNA3-HA-A20 (+) (lanes 3 and 4). After 48 h, equal amounts of cell extracts were subjected to western blot analysis with anti-p100/p52, anti-LMP1, or anti-tubulin antibody. The star denotes an induced p52 from p100 by LMP1. (ns non-specific)

Since A20 is an ubiquitin-modifying enzyme which targets TRAF6 and blocks NF-κB activation from both CTAR1 and CTAR2 [32, 33, 34, 35], we further investigated the effect of A20 on LMP1-induced p100 processing. Human embryonic kidney (HEK) 293 cells were co-transfected with LMP1 and HA-tagged A20 expression vectors, and p100 processing was determined at 48 h after transfection. LMP1-induced p100 processing was not affected by the over-expression of A20 in HEK293 cells (Fig. 3b, compare lane 2 with lane 4). Thus, A20 had no effect on LMP1-induced p100 processing. Taken together, these data indicate that TRAF6 and A20 are dispensable for LMP1-induced p100 processing.

LMP1–CTAR1 activates alternative p65/p52 complex

LMP1–CTAR1 induces p100 processing and activates RelB/p52 heterodimers [23, 24, 25]. We further investigated whether LMP1–CTAR1 activates the alternative NF-κB pathway involving p65 and p52. HEK293 cells were transfected with vectors expressing either LMP1 aa 1-386 (WT), LMP1 deleted for CTAR2 (LMP1 aa 1-231) (denoted as CTAR1), or LMP1 deleted for CTAR1 (Δ187-351) (denoted as CTAR2), and nuclear localization of p65 or p52 was determined at 48 h after transfection. As previously reported, LMP1 WT and CTAR1 induced p100 processing and facilitated p52 nuclear localization (Fig. 4a, lanes 6 and 8). Interestingly, LMP1–CTAR1 also induced p65 nuclear localization (Fig. 4a, lane 8). In order to further determine whether LMP1–CTAR1 induces the formation of the alternative p65/p52 complex, HEK293 cells expressing either LMP1 WT, LMP1 aa 1-231 (CTAR1), or LMP1 Δ187-351 (CTAR2) were harvested, and cell lysates were immunoprecipitated with anti-p65 antibodies. The formation of p65/p52 complex was determined by western blot analysis with anti-p52 antibody. LMP1–CTAR1, but not CTAR2, induced p65 interaction with p100 and p52 indicating LMP1 CTAR1 mediates the formation of alternative p65/p100 and p65/p52 heterodimers (Fig. 4b, lanes 6 and 8). Taken together, LMP1–CTAR1 activates the alternative NF-κB pathway involving p65/p100 and p65/p52 complexes.
Fig. 4

LMP1 CTAR1 activates alternative p65/p52 complex. a HEK293 cells were transfected with pSG5 (−), pSG5-FLAG-LMP1 (WT), pSG5-FLAG-LMP1, Δ187-351 (CTAR2), or pSG5-FLAG-LMP1 aa 1-231 (CTAR1). After 48 h, cytoplasmic (lanes 14) or nuclear extracts (lanes 56) were subjected to western blot analysis with either anti-p65/RelA antibody or anti-p100/p52 antibody; p100 and tubulin were controls for cytoplasmic contamination of the nuclear fraction. b HEK293 cells were transfected with pSG5 (−), pSG5-FLAG-LMP1 (WT), pSG5-FLAG-LMP1, Δ187-351 (CTAR2), or pSG5-FLAG-LMP1 aa 1-231 (CTAR1). After 48 h, p65/RelA was immunoprecipitated, and the immunoprecipitates were subjected to western blot analysis with either anti-p100/p52 antibody (top) or anti-p65/RelA antibody (bottom)

Discussion

LMP1–CTAR1 interacts with TRAF1, 2, 3, and 5 and induces NIK/IKKα-dependent p100 processing to p52 [7, 8, 9, 11, 23, 24, 25, 36]. Since LMP1–CTAR1-induced NF-κB activation is critical for EBV-mediated transformation of B lymphocytes [8, 11, 36], delineation of LMP1-induced p100 processing may contribute to the discovery of novel therapeutic strategies in treating EBV-associated cancers. In this study, we determined roles of IKKα, IKKβ, IKKγ, NIK, TRAF2, TRAF3, and TRAF6 in LMP1-induced p100 processing to p52. Although roles of IKKα, IKKβ, IKKγ, and NIK were previously reported [23, 24, 25], we employed KO MEFs to revisit the roles of these proteins in LMP1-induced p100 processing and to validate the p100 processing assay. The novel findings in this study are that (i) the over-expression of TRAF2 or TRAF3 blocked the LMP1-induced p100 processing in HEK293 cells suggesting negative roles of these proteins in LMP1-induced alternative NF-κB activation, (ii) a deficiency of TRAF6, an essential adaptor protein in a canonical pathway of NF-κB activation, did not abolish LMP1-induced p100 processing in MEFs, (iii) the over-expression A20, a critical regulator of a canonical pathway of NF-κB activation, did not inhibit the LMP1-induced p100 processing; and (iv) LMP1–CTAR1, but not CTAR2, induced the formation of the alternative p65/p52 complex.

TRAF2 and TRAF3 negatively regulate p100 processing by inducing degradation of NIK [27, 28, 29]. TRAF3 recruits a TRAF2–cIAP1–cIAP2 E3 ubiquitin ligase complex to NIK and facilitates its degradation [29, 37]. Thus, lack of TRAF2 or TRAF3 enhances NIK protein levels and might induce robust p100 processing to p52 [28, 38]. The basal protein level of p100 in TRAF3−/− MEFs was significantly lower than in TRAF2−/− MEFs (Fig. 2a, compare lanes 3 and 4 with lanes 5 and 6). Decreased p100 expression may be due to impaired canonical pathway of NF-κB activation in TRAF3−/− cells [39, 40, 41, 42]. Furthermore, TRAF3 may be a major negative regulator of p100 processing by recruiting other E3 ubiquitin ligase complexes in addition to the TRAF2–cIAP1–cIAP2 to NIK. Since p100 processing to p52 was constitutively active in TRAF2−/− or TRAF3−/− MEFs independent of LMP1 expression (Fig. 2a, lanes 3 and 5), we employed the over-expression of TRAF2 or TRAF3 to determine roles of these proteins in LMP1-induced p100 processing. Interestingly, TRAF2 or TRAF3 over-expression in HEK293 cells significantly inhibited LMP1-induced p100 processing. LMP1 induces NIK–IKKα-dependent p100 processing possibly by binding to TRAF2 and TRAF3, down-regulating or sequestering TRAF2 and TRAF3 away from NIK and, subsequently, activating NIK. How LMP1 regulates TRAF2 and TRAF3 to activate NIK and IKKα is under investigation.

Although TRAF6 is essential for LMP1–CTAR2-induced canonical IKKβ activation [30], it is not required for LMP1–CTAR1-induced alternative p100 processing. Therefore, LMP1–CTAR1 or CTAR2 utilizes a unique set of adaptor proteins to activate the alternative or canonical NF-κB pathway. In consistent with the result obtained using TRAF6−/− MEFs, over-expression of A20, which down-regulates TRAF6, has no effect on LMP1-induced p100 processing. Interestingly, A20 inhibits both LMP1–CTAR1 and CTAR2-induced NF-κB activation [33]. Since LMP1–CTAR1 activates NF-κB activation by inducing p100 processing, how A20 blocks LMP1–CTAR1-induced NF-κB activation is unclear. LMP1–CTAR1 induces the nuclear localization of p65 and activates p65/p52 heterodimers in addition to RelB/p52 heterodimers. Therefore, A20 may affect the alternative p65/p52 complex to down-regulate LMP1–CTAR1-induced NF-κB activation. A20 may have additional functions to regulate the alternative NF-κB activation pathway downstream of p100 processing that are currently unknown.

Acknowledgments

We are grateful to Dr. Elliott Kieff for discussion. M.-S. K. was supported by the National Research Foundation of Korea, and The Korean Federation of Science and Technology Societies Grant by Korean Government (MEST, Basic Research Promotion Fund). Y.-J. S. was supported by Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (2010-0003301), and by the Kyungwon University Research Fund in 2010. This work was in part supported by 5R01CA085180-10 granted to Elliott Kieff, Channing Laboratory, Brigham and Women’s Hospital.

Supplementary material

11262_2010_505_MOESM1_ESM.ppt (2.4 mb)
Supplementary Figure 1. WT (lanes 1 and 2), IKKα−/− (lanes 3 and 4), or NIKaly/aly (lanes 5 and 6). MEFs were transduced with retroviruses expressing either GFP (−) or LMP1(+). At 6 days after transduction, equal amounts of cell extracts were subjected to western blotting with anti-p100/p52 antibody. (PPT 2436 kb)

Copyright information

© Springer Science+Business Media, LLC 2010

Authors and Affiliations

  1. 1.Department of Life ScienceKyungwon UniversitySeongnam-SiKorea
  2. 2.Institute for Biomedical Sciences, Samsung Medical CenterSungkyunkwan University School of MedicineGangnam-gu, SeoulKorea

Personalised recommendations