Advertisement

The Absence of Fucose but Not the Presence of Galactose or Bisecting N-Acetylglucosamine of Human IgG1 Complex-type Oligosaccharides Shows the Critical Role of Enhancing Antibody-dependent Cellular Cytotoxicity*

Open AccessPublished:November 08, 2002DOI:https://doi.org/10.1074/jbc.M210665200
      An anti-human interleukin 5 receptor (hIL-5R) humanized immunoglobulin G1 (IgG1) and an anti-CD20 chimeric IgG1 produced by rat hybridoma YB2/0 cell lines showed more than 50-fold higher antibody-dependent cellular cytotoxicity (ADCC) using purified human peripheral blood mononuclear cells as effector than those produced by Chinese hamster ovary (CHO) cell lines. Monosaccharide composition and oligosaccharide profiling analysis showed that low fucose (Fuc) content of complex-type oligosaccharides was characteristic in YB2/0-produced IgG1s compared with high Fuc content of CHO-produced IgG1s. YB2/0-produced anti-hIL-5R IgG1 was subjected to Lens culinaris aggulutin affinity column and fractionated based on the contents of Fuc. The lower Fuc IgG1 had higher ADCC than the IgG1 before separation. In contrast, the content of bisecting GlcNAc of the IgG1 affected ADCC much less than that of Fuc. In addition, the correlation between Gal and ADCC was not observed. When the combined effect of Fuc and bisecting GlcNAc was examined in anti-CD20 IgG1, only a severalfold increase of ADCC was observed by the addition of GlcNAc to highly fucosylated IgG1. Quantitative PCR analysis indicated that YB2/0 cells had lower expression level of FUT8 mRNA, which codes α1,6-fucosyltransferase, than CHO cells. Overexpression of FUT8 mRNA in YB2/0 cells led to an increase of fucosylated oligosaccharides and decrease of ADCC of the IgG1. These results indicate that the lack of fucosylation of IgG1 has the most critical role in enhancement of ADCC, although several reports have suggested the importance of Gal or bisecting GlcNAc and provide important information to produce the effective therapeutic antibody.
      ADCC
      antibody-dependent cellular cytotoxicity
      hIL-5R
      human interleukin 5 receptor
      Fuc
      fucose
      Man
      mannose
      CHO
      Chinese hamster ovary
      Fc
      the constant region of the antibody
      PBMC
      peripheral blood mononuclear cell
      LCA
      L. culinaris aggulutin
      PHA-E4
      P. vulgaris E4
      HPLC
      high performance liquid chromatography
      LTR
      long terminal repeat
      GnTIII
      N-acetylglucosaminyltransferase III
      Antibody-dependent cellular cytotoxicity (ADCC),1 a lytic attack on antibody-targeted cells, is triggered upon binding of lymphocyte receptors (FcγRs) to the constant region (Fc) of the antibodies. ADCC is considered to be a major function of some of the therapeutic antibodies, although antibodies have multiple therapeutic functions (e.g. antigen binding, induction of apoptosis, and complement-dependent cellular cytotoxicity) (
      • Lewis G.D.
      • Figari I.
      • Fendly B.
      • Wong W.L.
      • Carter P.
      • Gorman C.
      • Shepard H.M.
      ,
      • Clynes R.A.
      • Towers T.L.
      • Presta L.G.
      • Ravetch J.V.
      ).
      One IgG molecule contains two N-linked oligosaccharide sites in its Fc region (
      • Rademacher T.W.
      • Homans S.W.
      • Perekh R.B.
      • Dwek R.A.
      ). The general structure of N-linked oligosaccharide on IgG is complex-type, characterized by a mannosyl-chitobiose core (Man3GlcNAc2-Asn) with or without bisecting GlcNAc/l-fucose (Fuc) and other chain variants including the presence or absence of Gal and sialic acid. In addition, oligosaccharides may contain zero (G0), one (G1), or two (G2) Gal.
      Recent studies have shown that engineering the oligosaccharides of IgGs may yield optimized ADCC. ADCC requires the presence of oligosaccharides covalently attached at the conserved Asn297 in the Fc region and is sensitive to change in the oligosaccharide structure. In the oligosaccharide, sialic acid of IgG has no effect on ADCC (
      • Boyd P.N.
      • Lines A.C.
      • Patel A.K.
      ). The relationship between the Gal residue and ADCC is controversial. Boyd et al. (
      • Boyd P.N.
      • Lines A.C.
      • Patel A.K.
      ) have shown that obvious change was not found in ADCC after removal of the majority of the Gal residues. However, several reports have shown that Gal residues enhance ADCC (
      • Kumpel B.M.
      • Rademacher T.W.
      • Rook G.A.
      • Williams P.J.
      • Wilson I.B.
      ,
      • Kumpel B.M.
      • Wang Y.
      • Griffiths H.L.
      • Hadley A.G.
      • Rook G.A.
      ).
      Several groups have focused on bisecting GlcNAc, which is a β1,4-GlcNAc residue transferred to a core β-mannose (Man) residue, and it has been implicated in biological activity of therapeutic antibodies (
      • Lifely M.R.
      • Hale C.
      • Boyce S.
      • Keen M.J.
      • Phillips J.
      ). N-Acetylglucosaminyltransferase III (GnTIII), which catalyzes the addition of the bisecting GlcNAc residue to the N-linked oligosaccharide (
      • Narisimhan S.
      ), has been expressed in a Chinese hamster ovary (CHO) cell line with an anti-neuroblastoma IgG1 and resulted in greater ADCC (
      • Umana P.
      • Jean-Mairet J.
      • Moudry R.
      • Amstutz H.
      • Bailey J.E.
      ). Moreover, expression of GnTIII in a recombinant CHO cell line has led to the increase in ADCC of the anti-CD20 antibody (
      • Davies J.
      • Jiang L.
      • Pan L.Z.
      • LaBarre M.J.
      • Anderson D.
      • Reff M.
      ).
      Recently, Shields et al. have revealed the effect of fucosylated oligosaccharide on antibody effector functions, including binding to human FcγR, human C1q, human FcRn, and ADCC (
      • Shields R.L.
      • Lai J.
      • Keck R.
      • O'Connell L.Y.
      • Hong K.
      • Meng Y.G.
      • Weikert S.H.
      • Presta L.G.
      ). The Fuc-deficient IgG1s have shown 50-fold increased binding to FcγRIIIa and enhanced ADCC. Nevertheless, there are no data on comparison of the effect of Fuc, Gal, and GlcNAc or the combined effect of Fuc and bisecting GlcNAc.
      Here, we describe the correlation between glycosylation of human IgG1 and ADCC and demonstrate that Fuc showed the critical role for enhancing ADCC out of several sugar residues reported previously. We unexpectedly found that human IgG1 produced by rat hybridoma YB2/0 cells showed extremely high ADCC at more than 50-fold lower concentration of those produced by CHO cells. YB2/0-produced IgG1 had lower Fuc content than CHO-produced IgG1. IgG1 containing lower fucosylated oligosaccharides, which was fractionated by Lens culinaris aggulutin (LCA) lectin affinity chromatography, showed higher ADCC before separation. In contrast, the addition of bisecting GlcNAc to IgG1 enhanced ADCC much less effectively than defucosylation. The effect of bisecting GlcNAc was only observed in highly fucosylated IgG1. YB2/0 cells expressed a lower level of FUT8 (α1,6-fucosyltransferase gene) mRNA than CHO cells, and overexpression of FUT8 in YB2/0 led the increase of fucosylation of IgG1 and the decrease of ADCC.

      DISCUSSION

      In this study, we analyzed the molecular basis of extremely high ADCC of recombinant IgG1 produced by rat hybridoma YB2/0 cells, which produced IgG1 with at least 50-fold higher ADCC than that produced by CHO cells, one of the most widely used host cell lines for production of recombinant antibodies (Fig. 1). Our conclusion of the present study is that nonfucosylated oligosaccharide of YB2/0-produced IgG1 has a more critical role in enhancing ADCC than Gal-binding or bisecting GlcNAc-binding oligosaccharides according to the following evidence. First, monosaccharide composition and oligosaccharide profiling analysis showed that high content of nonfucosylated complex-type oligosaccharides were characteristic in YB2/0-produced anti-hIL-5R IgG1 (34%, Table II) and anti-CD20 IgG1 (91%, Table IV) compared with low content of those in CHO-produced anti-hIL-5R IgG1 (9%, data not shown) and anti-CD20 IgG1 (6%, data not shown). Second, ADCC assay of the anti-hIL-5R IgG1 separated by LCA affinity chromatography demonstrated that Fuc content of IgG1 was inversely correlated with ADCC (Fig. 3, A and B). Third, quantitative PCR analysis indicated that YB2/0 cells had a 10-fold lower expression level of FUT8 mRNA than CHO cells (Fig. 6 A). Fourth, overexpression of FUT8 in YB2/0 cells increased the content of fucosylated oligosaccharides and also decreased ADCC of anti-CD20 IgG1 (Fig. 6 B).
      l-Fuc residues in an α1,6-linkage to the GlcNAc of the reducing end (“core Fuc”) are relatively common in mammalianN-linked oligosaccharide. FUT8, considered to be the only gene that codes α1,6-fucosyltransferase, catalyzes the transfer of Fuc from GDP-Fuc to GlcNAc of the reducing end. Therefore, we focused on FUT8 as a key gene controlling the low Fuc content of IgG1 produced by YB2/0 and indicate that the cells produce low Fuc content IgG1 simply due to the low expression level of FUT8. Since biosynthesis ofN-linked oligosaccharides is controlled by a number of glycosyltransferases, their acceptors and substrates, etc., it remains to be determined whether there is possible involvement of the other factor in biosynthesis of nonfucosylated IgG1 in YB2/0 cells.
      The importance of nonfucosylated oligosaccharide on ADCC has been reported very recently by Shields et al. (
      • Shields R.L.
      • Lai J.
      • Keck R.
      • O'Connell L.Y.
      • Hong K.
      • Meng Y.G.
      • Weikert S.H.
      • Presta L.G.
      ). They have shown that nonfucosylated anti-Her2 humanized IgG1 and anti-IgE humanized IgG1 produced by a variant of CHO cells, Lec13, had enhanced ADCC relative to fucosylated IgG1s produced by normal CHO cells. However, they have only focused on Fuc, because no appreciable differences in the content of the other sugar residues have been found in Lec13-produced IgG1 and normal CHO-produced IgG1. Until now, the effects of Fuc, Gal, or bisecting GlcNAc on ADCC have been analyzed independently (
      • Boyd P.N.
      • Lines A.C.
      • Patel A.K.
      ,
      • Kumpel B.M.
      • Rademacher T.W.
      • Rook G.A.
      • Williams P.J.
      • Wilson I.B.
      ,
      • Kumpel B.M.
      • Wang Y.
      • Griffiths H.L.
      • Hadley A.G.
      • Rook G.A.
      ,
      • Lifely M.R.
      • Hale C.
      • Boyce S.
      • Keen M.J.
      • Phillips J.
      ,
      • Umana P.
      • Jean-Mairet J.
      • Moudry R.
      • Amstutz H.
      • Bailey J.E.
      ,
      • Davies J.
      • Jiang L.
      • Pan L.Z.
      • LaBarre M.J.
      • Anderson D.
      • Reff M.
      ,
      • Shields R.L.
      • Lai J.
      • Keck R.
      • O'Connell L.Y.
      • Hong K.
      • Meng Y.G.
      • Weikert S.H.
      • Presta L.G.
      ); therefore, comparison of the effect of each sugar residue or the combined effect of each sugar residue has not yet been reported.
      In this report, we could not find any correlation between the content of Gal and ADCC. A difference in the content of Gal between YB2/0-produced IgG1s and CHO-produced IgG1s was not correlated to ADCC (Table I). As a result of the separation of KM8399 using LCA lectin affinity chromatography, the compositions of G0, G1, and G2 of fraction II were very similar to that of KM8399 (Table II); nevertheless, ADCC of fraction II and KM8399 was quite different (Fig. 3 B). Our results show a good coincidence with the report of Boyd et al. (
      • Boyd P.N.
      • Lines A.C.
      • Patel A.K.
      ), in which obvious change was not found in ADCC after removal of the majority of the Gal residues of anti-CDw52 IgG1 produced by CHO cells. In contrast, Kumpel et al. (
      • Kumpel B.M.
      • Rademacher T.W.
      • Rook G.A.
      • Williams P.J.
      • Wilson I.B.
      ,
      • Kumpel B.M.
      • Wang Y.
      • Griffiths H.L.
      • Hadley A.G.
      • Rook G.A.
      ) reported that highly galactosylated anti-D-antigen IgG1s have higher ADCC, although that effect was only 2–3-fold.
      Two groups independently reported that increasing the level of bisecting GlcNAc of anti-neuroblastoma IgG1 and anti-CD20 IgG1 could enhance ADCC (
      • Umana P.
      • Jean-Mairet J.
      • Moudry R.
      • Amstutz H.
      • Bailey J.E.
      ,
      • Davies J.
      • Jiang L.
      • Pan L.Z.
      • LaBarre M.J.
      • Anderson D.
      • Reff M.
      ). GnTIII-transfected CHO cells produced anti-CD20 IgG1 with a high content of bisecting GlcNAc (48–71%), which showed a 10–20-fold enhancement of ADCC compared with that with no content of bisecting GlcNAc (0%). In the present study, we carefully examined the effect of bisecting GlcNAc in ADCC in comparison with that of Fuc. We prepared anti-hIL-5R IgG1 with different content of bisecting GlcNAc (0–30%). To avoid the effect of fucosylation, nonfucosylated IgG1s were depleted by LCA lectin affinity chromatography. To our surprise, we could not detect any correlation between ADCC and content of bisecting GlcNAc. One possible explanation of the discrepancy with the results of Umana et al. (
      • Umana P.
      • Jean-Mairet J.
      • Moudry R.
      • Amstutz H.
      • Bailey J.E.
      ) and Davieset al. (
      • Davies J.
      • Jiang L.
      • Pan L.Z.
      • LaBarre M.J.
      • Anderson D.
      • Reff M.
      ) might be that 30% content of bisecting GlcNAc is not enough to enhance ADCC. We next produced anti-CD20 IgG1 by LEC10 cells, a variant CHO cell that overexpressed GnTIII. The resultant IgG1 (74% bisecting GlcNAc and 100% Fuc) had shown only severalfold higher ADCC than normal CHO-produced IgG1 (0% bisecting GlcNAc and 94% Fuc); in contrast, YB2/0-produced IgG1 (19% bisecting GlcNAc and 91% non-Fuc) had 100-fold higher ADCC than LEC10-produced IgG1 (Fig. 5 B). These results suggest that an extremely high content of bisecting GlcNAc (74%) has a relatively weak effect for enhancing ADCC. More importantly, nonfucosylated oligosaccharide was shown to have a prominent effect in enhancement of ADCC of IgG1 compared with bisecting GlcNAc-containing oligosaccharide. We further evaluated the combined effect of bisecting GlcNAc with nonfucosylated oligosaccharides. YB2/0-produced anti-CD20 IgG1 was separated based on the content of bisecting GlcNAc-binding oligosaccharides (0, 8, 33, and 45%), which contained the same content of nonfucosylated oligosaccharides (around 90%). These four fractions did not show any significant difference in ADCC, indicating that the presence of bisecting GlcNAc-binding oligosaccharides, at least under 45%, does not have any additional effect in ADCC of highly nonfucosylated IgG1 (90%). To our knowledge, this is the first report that shows the effect of Fuc, Gal, and bisecting GlcNAc simultaneously and also shows the combined effect of Fuc and bisecting GlcNAc.
      The ADCC have been believed to be a result of specific killing of antigen-positive cells by natural killer cells through binding of the IgG Fc domain to FcγRIIIa. Recently, Shields et al. (
      • Shields R.L.
      • Lai J.
      • Keck R.
      • O'Connell L.Y.
      • Hong K.
      • Meng Y.G.
      • Weikert S.H.
      • Presta L.G.
      ) have revealed that binding of the Fuc-deficient IgG1 (produced by Lec13 cells) to FcγRIIIa was enhanced up to 50-fold. They have shown that improved binding to FcγRIIIa has translated into improved ADCCin vitro, using PBMC or natural killer cells. These results suggest that Fuc-deficient IgG1 may require a lower concentration of antibody on the surface of the target cell to activate an effector cell. There are a few possible explanations of why the antibodies with nonfucosylated oligosaccharides give rise to stronger binding to FcγRIIIa than those in which the glycoforms are absent. A core Fuc has been shown to influence the conformational flexibility of biantennary oligosaccharides (
      • Stubbs H.J.
      • Lih J.J.
      • Gustafson T.L.
      • Rice K.G.
      ,
      • Mukhopadhyay C.
      ). The oligosaccharides of IgG appear to be largely sequestered between the CH2 domains and may help to stabilize the CH2 domain (
      • Deisenhofer J.
      ). In the co-crystal structure of IgG1 Fc:FcγRIIIb, Fuc is orientated away from the interface and making no specific contacts with the receptor (
      • Radaev S.
      • Motyka S.
      • Fridman W.H.
      • Sautes-Fridman C.
      • Sun P.D.
      ); nevertheless, Harris et al. (
      • Harris L.J.
      • Larson S.B.
      • Hasel K.W.
      • McPherson A.
      ,
      • Harris L.J.
      • Skaletsky E.
      • McPherson A.
      ) have supposed that Fuc could have influence on the binding by the receptor. We speculated that the absence of Fuc provided a more suitable conformation for the binding of IgG1 to FcγRIII than the presence of bisecting GlcNAc or Gal, although structural analyses of a series of IgG1 with or without Fuc, bisecting GlcNAc, or Gal are needed for further discussion.
      Several recombinant monoclonal antibodies are being used as human therapeutics. Some of these are blocking monoclonal antibodies to receptors or soluble ligands and therefore may function without utilizing antibody effector functions. However, ADCC is still considered to be one of the most important anti-tumor mechanisms of clinically effective anti-Her2 humanized IgG1 and anti-CD20 chimeric IgG1 at least in animal models, since they are supposed to have multiple anti-tumor mechanisms (
      • Clynes R.A.
      • Towers T.L.
      • Presta L.G.
      • Ravetch J.V.
      ). More importantly, Cartron et al. (
      • Cartron G.
      • Dacheux L.
      • Salles G.
      • Solal-Celigny P.
      • Bardos P.
      • Colombat P.
      • Watier H.
      ) have reported recently that therapeutic activity of anti-CD20 chimeric IgG1 in patients with non-Hodgkin's lymphoma has been correlated with polymorphism in the FcγRIIIa gene. They have shown that FCGR3A-158V patients showed a better response to anti-CD20 chimeric IgG1 because they have higher ADCC against lymphoma cells. These reports have suggested that therapeutic antibodies with enhanced ADCC including the nonfucosylated IgG1 would result in the improvement of clinical response. Moreover, these findings may allow for use of the nonfucosylated IgG1 at lower doses with no reduction in efficacy. Antibody therapeutics effective in lower doses might reduce the cost of antibody therapy.

      REFERENCES

        • Lewis G.D.
        • Figari I.
        • Fendly B.
        • Wong W.L.
        • Carter P.
        • Gorman C.
        • Shepard H.M.
        Cancer Immunol. Immunother. 1993; 37: 255-263
        • Clynes R.A.
        • Towers T.L.
        • Presta L.G.
        • Ravetch J.V.
        Nat. Med. 2000; 6: 443-446
        • Rademacher T.W.
        • Homans S.W.
        • Perekh R.B.
        • Dwek R.A.
        Biochem. Soc. Symp. 1986; 51: 131-148
        • Boyd P.N.
        • Lines A.C.
        • Patel A.K.
        Mol. Immunol. 1995; 32: 1311-1318
        • Kumpel B.M.
        • Rademacher T.W.
        • Rook G.A.
        • Williams P.J.
        • Wilson I.B.
        Hum. Antib. Hybrid. 1994; 5: 143-151
        • Kumpel B.M.
        • Wang Y.
        • Griffiths H.L.
        • Hadley A.G.
        • Rook G.A.
        Hum. Antib. Hybrid. 1995; 6: 82-88
        • Lifely M.R.
        • Hale C.
        • Boyce S.
        • Keen M.J.
        • Phillips J.
        Glycobiology. 1995; 5: 813-822
        • Narisimhan S.
        J. Biol. Chem. 1982; 257: 10235-10242
        • Umana P.
        • Jean-Mairet J.
        • Moudry R.
        • Amstutz H.
        • Bailey J.E.
        Nat. Biotechnol. 1999; 17: 176-180
        • Davies J.
        • Jiang L.
        • Pan L.Z.
        • LaBarre M.J.
        • Anderson D.
        • Reff M.
        Biotechnol. Bioeng. 2001; 74: 288-294
        • Shields R.L.
        • Lai J.
        • Keck R.
        • O'Connell L.Y.
        • Hong K.
        • Meng Y.G.
        • Weikert S.H.
        • Presta L.G.
        J. Biol. Chem. 2002; 277: 26733-26740
        • Urlaub G.
        • Mitchell P.J.
        • Kas E.
        • Chasin L.A.
        • Funanage V.L.
        • Myoda T.T.
        • Hamlin J.
        Somatic Cell Mol. Genet. 1986; 12: 555-566
        • Campbell C.
        • Stanley P.
        J. Biol. Chem. 1984; 259: 13370-13378
        • Nakamura K.
        • Tanaka Y.
        • Fujino I.
        • Hirayama N.
        • Shitara K.
        • Hanai N.
        Mol. Immunol. 2000; 37: 1035-1046
        • Ohta S.
        • Honda A.
        • Tokutake Y.
        • Yoshida H.
        • Hanai N.
        Cancer Immunol. Immunother. 1993; 36: 260-266
        • Takaki S.
        • Murata Y.
        • Kitamura T.
        • Miyajima A.
        • Tominaga A.
        • Takatsu K.
        J. Exp. Med. 1993; 177: 1523-1529
        • Hase S.
        • Ibuki T.
        • Ikenaka T.
        J. Biochem. (Tokyo). 1984; 95: 197-203
        • Takahashi N.
        • Hotta T.
        • Ishihara H.
        • Mori M.
        • Tejima S.
        • Bligny R.
        • Akazawa T.
        Biochemistry. 1986; 25: 388-395
        • McGuire J.M.
        • Douglas M.
        • Smith K.D.
        Carbohydr. Res. 1996; 292: 1-9
        • Sasaki K.
        • Kurata K.
        • Funayama K.
        • Nagata M.
        • Watanabe E.
        • Ohta S.
        • Hanai N.
        • Nishi T.
        J. Biol. Chem. 1994; 269: 14730-14737
        • Hayashi H.
        • Yoneda A.
        • Asada M.
        • Ikekita M.
        • Imamura T.
        DNA Seq. 2000; 11: 91-96
        • Takahashi N.
        • Ishii I.
        • Ishihara H.
        • Mori M.
        • Tejima S.
        • Jefferis R.
        • Endo S.
        • Arata Y.
        Biochemistry. 1987; 26: 1137-1144
        • Parekh R.B.
        • Dwek R.A.
        • Sutton B.J.
        • Fernandes D.L.
        • Leung A.
        • Stanworth D.
        • Rademacher T.W.
        • Mizuochi T.
        • Taniguchi T.
        • Matsuda K.
        • Takeuchi F.
        • Nagano Y.
        • Miyamoto T.
        • Kobata A.
        Nature. 1985; 316: 452-457
        • Jefferis R.
        • Lund J.
        • Mizutani H.
        • Nakagawa H.
        • Kawazoe Y.
        • Arata Y.
        • Takahashi N.
        Biochem. J. 1990; 268: 529-537
        • Yanagidani S.
        • Uozumi N.
        • Ihara Y.
        • Miyoshi E.
        • Yamaguchi N.
        • Taniguchi N.
        J. Biochem. (Tokyo). 1997; 121: 626-632
        • Stubbs H.J.
        • Lih J.J.
        • Gustafson T.L.
        • Rice K.G.
        Biochemistry. 1996; 35: 937-947
        • Mukhopadhyay C.
        Biopolymers. 1998; 45: 177-190
        • Deisenhofer J.
        Biochemistry. 1981; 20: 2361-2370
        • Radaev S.
        • Motyka S.
        • Fridman W.H.
        • Sautes-Fridman C.
        • Sun P.D.
        J. Biol. Chem. 2001; 276: 16469-16477
        • Harris L.J.
        • Larson S.B.
        • Hasel K.W.
        • McPherson A.
        Biochemistry. 1997; 36: 1581-1597
        • Harris L.J.
        • Skaletsky E.
        • McPherson A.
        J. Mol. Biol. 1998; 275: 861-872
        • Cartron G.
        • Dacheux L.
        • Salles G.
        • Solal-Celigny P.
        • Bardos P.
        • Colombat P.
        • Watier H.
        Blood. 2002; 99: 754-758
        • Masuda K.
        • Yamaguchi Y.
        • Kato K.
        • Takahashi N.
        • Shimada I.
        • Arata Y.
        FEBS Lett. 2000; 473: 349-357