J. P. Deans, Department of Biochemistry and Molecular Biology, University of Calgary, 3330 Hospital Drive NW, Calgary, Alberta, Canada T2N 4N1. E-mail: firstname.lastname@example.org
CD20 is an effective target for therapeutic B-cell depletion with monoclonal antibodies. One proposed mechanism of action is direct cytotoxicity mediated via tyrosine kinase-dependent signalling pathways activated upon CD20 cross-linking. The association of CD20 with membrane microdomains known as lipid rafts, enriched in src-family tyrosine kinases and other signalling effectors, suggests an indirect mechanism of anti-CD20-induced apoptosis in which activation of src-family kinases occurs as a consequence of lipid raft clustering.
CD20 is one of the most reliable cell surface markers of human B lymphocytes. High expression on all normal and most malignant B cells, together with a low rate of antibody-induced internalization, has made CD20 an attractive target for immunotherapeutic depletion of B cells.1,2 After many years of preclinical studies the promise of immunotherapy with monoclonal antibodies (mAbs) was realized with the introduction of rituximab (Rituxan; Genentech, Inc, South San Francisco, CA, and IDEC Pharmaceuticals, San Diego, CA), a humanized mouse mAb directed against CD20.3,4 Approved in 1997 for treating low-grade non-Hodgkin's lymphoma, rituximab is also proving to be effective against other B-cell malignancies and in autoimmune diseases such as rheumatoid arthritis, autoimmune haemolytic anaemia and idiopathic thrombocytopenic purpura, in which pathogenic autoantibodies are clearly involved.5–8 The mechanism of rituximab action in vivo is not entirely clear, but is thought to involve complement-mediated lysis or phagocytosis, antibody-dependent cellular cytoxicity (ADCC), and direct effects of CD20-mediated signalling on cell growth and viability.9 Supporting a prominent role for ADCC are the observations that FcγRIIIa is required for effective killing by rituximab in a human tumour xenograft mouse model,10 and that resistance to rituximab in some patients may be linked to a polymorphism in FcγRIIIa affecting isotype preference.11
Although complement and ADCC are likely to be the major effectors of B-cell depletion in vivo,9 direct cytoxicity by rituximab and other CD20 mAbs on B-cell lines has been reported by many investigators.12–19 Evidence of apoptosis occurring in B cells freshly isolated from patients receiving rituximab supports a role for direct effects of CD20 signalling in vivo.20 Characterization of the signalling pathways leading to CD20-mediated apoptotic cell death is of considerable interest, particularly because not all patients respond to rituximab therapy and the mechanisms of resistance are unknown.
Structural features of cd20
The CD20 protein is 297 amino acids long with a molecular mass ∼33 000. The cDNA sequence predicted,21–23 and experimental data confirmed,24 that the protein has intracellular termini, spans the plasma membrane four times, and has a single, non-glycosylated25 extracellular loop of ∼43 residues between the third and fourth transmembrane regions. No tyrosine residues or recognized signalling motifs occur in any of the cytoplasmic regions, although there are a number of consensus sites for serine or threonine phosphophorylation.
CD20 almost certainly does not exist in the plasma membrane in monomeric form. After chemical cross-linking of surface proteins, CD20 migrates on sodium dodecyl sulphate–polyacrylamide gel electrophoresis non-reducing gels at 33 000, 70 000 and 140 000, indicating the existence of dimers and tetramers.26 In digitonin lysates CD20 is entirely organized into complexes of ∼200 000, comprised principally of CD20 with at least one additional protein component.27 Therefore, it is likely that CD20 exists as a tetramer, probably organized as a dimer of dimers, in a heterogeneous complex with other minor components.
There is a remarkably high degree of epitope diversity among CD20 mAbs, despite the small size of the extracellular region.27 This is, perhaps, not so surprising in the light of evidence that some CD20 epitopes require the integrity of the intact oligomeric complex and therefore could include residues contributed by neighbouring molecules.27 The secondary structure of the extracellular loop appears to depend on two residues, alanine and proline at positions 170 and 172, respectively (see Fig. 1). Mutation of these amino acids destroys the epitopes of all mAbs tested27 without disrupting the oligomeric complex (M. Polyak, unpublished data). The amino acid sequence of the extracellular region of human CD20 differs from that of murine CD20 at 16 of the approximate 43 residues, including alanine170 and proline172 (both positions have serine residues in the mouse sequence). Not surprisingly, antibodies directed against extracellular epitopes on human CD20 do not bind to murine B cells. Remarkably, however, replacement of only the two serine residues in the murine sequence located at positions equivalent to 170 and 172 in the human sequence, with alanine and proline, is all that is required to reconstitute epitopes recognized by many anti-human CD20 mAbs (ref. 27 and see Table 1). From this one can conclude that there are significant secondary structure differences between the extracellular loops of murine and human CD20, in addition to the obvious differences in amino acid sequence.
Table 1. Characteristics of selected CD20 mAbs
NT, not tested.
Epitope reconstituted in murine CD20 by S×S to A×P mutations
In the absence of a known ligand, much of our current understanding of CD20-activated signalling events is derived from the effects of ligating CD20 with mAbs (Table 1). In addition to rituximab, those antibodies that have been most characterized are B1, 1F5 and 2H7, each of which recognizes a distinct epitope.27 Among these mAbs, only 1F5 has the capacity to up-regulate c-myc expression and activate resting B cells.28–31 Other CD20 mAbs have inhibitory effects on proliferation and/or mitogen-driven antibody secretion.29,32–35 B1 – and to a lesser extent, rituximab – induces homotypic aggregation in B-cell lines, whereas 1F5, 2H7, and most other CD20 mAbs, do not.27,36 A variety of other biological responses to CD20 mAbs has been reported, including down-regulation of the B-cell receptor,37 shedding of CD23,38,39 increased expression of major histocompatibility complex class II and adhesion molecules,34,40 blocking of lymphocyte function-associated antigen-1-independent homotypic adhesion,41 rescue from apoptosis,42,43 and induction of apoptosis (see below). Little is known of the signalling pathways responsible for these effects or their relevance to the normal functioning of the CD20 complex.
CD20 signalling and apoptosis
One of the earliest indications that CD20 might be coupled to a tyrosine kinase-dependent signalling pathway was the observation that the induction of homotypic aggregation by the B1 mAb was prevented by tyrosine kinase inhibitors.36 Subsequently, up-regulation of c-myc expression by the activating mAb 1F5 was also shown to be tyrosine kinase-dependent.44 The 2H7 mAb, which exhibits neither of these biological activities, nevertheless induces tyrosine kinase activation leading to phosphorylation of multiple tyrosine kinase substrates including phospholipase C-γ2.44‘Hyper-cross-linking’, i.e. cross-linking of antigen-bound CD20 antibody with a secondary antibody, caused release of calcium from intracellular stores.44 These findings were confirmed by two groups who went on to demonstrate that CD20 hyper-cross-linking up-regulated Fas expression, increased caspase activity, and induced apoptosis in B-cell lines.12,16,45 Induction of apoptosis by CD20 mAbs has now been demonstrated in various B-cell lines by several investigators. Although not a universal finding, there is general agreement that hyper-cross-linking enhances, and is often essential for, the detection of apoptotic effects of CD20 mAbs (Table 2). In vivo, hyper-cross-linking is thought to occur via engagement of Fc receptors.12
CD20-mediated apoptosis was significantly reduced by calcium chelation as well as by caspase inhibitors.15,16,45 Of special interest is that apoptotic effects were prevented by specific inhibition of src-family kinases.16,45 CD20 is associated with src-family members Lyn, Fyn and Lck, along with a tyrosine phosphorylated protein, p75/80.44,46 As discussed below, the association of both CD20 and src-family kinases with membrane microdomains known as lipid rafts suggests the possibility that apoptotic effects of CD20 ligation may result from transactivation of src-family tyrosine kinases upon clustering of rafts (Fig. 2).
CD20 association with lipid rafts
Lipid rafts are liquid-ordered membrane microdomains, enriched in sphingolipids and cholesterol, which are thought to function as platforms for signal transduction by selectively compartmentalizing receptors and signalling effectors.47 Characteristically, proteins associated with lipid rafts are insoluble in the non-ionic detergent Triton X-100 and can be distinguished from the detergent-insoluble cytoskeleton by their low density on sucrose gradients. CD20 is soluble in 1% Triton X-100, but becomes insoluble after antibody ligation and floats on sucrose density gradients, consistent with the interpretation that it inducibly associates with lipid rafts.48 However, it is now apparent that CD20 is constitutively associated with lipid rafts (L. Ayer et al., manuscript in preparation). Although CD20 is normally soluble in 1% Triton X-100, it is found in low-density insoluble membranes isolated using several other non-ionic detergents and in low concentrations of Triton X-100. Low stringency detergent conditions are also necessary to detect localization of the high-affinity immunoglobulin E (IgE) receptor, FcERI, and CD40 to lipid rafts.49,50 The constitutive association of CD20 with rafts is dependent on cholesterol and on a short membrane-proximal cytoplasmic sequence (residues 219–225, see Fig. 1) previously shown to be critical for antibody-induced CD20 insolubility in 1% Triton X-100.24
CD20-associated p75/80 has been identified as PAG (M. Kalia, unpublished data), a ubiquitous, highly tyrosine-phosphorylated adaptor protein localized exclusively to lipid rafts.51,52 Also known as Csk-binding protein (Cbp), PAG recruits Csk to lipid rafts to maintain resident src-family tyrosine kinases in the inactive state. Those src-family kinases found in rafts include Lyn, Fyn and Lck, which are dually acylated with myristate and palmitate, but not Blk which is modified only with myristate.53 Co-precipitation with CD20 of PAG, Lyn, Fyn and Lck – but not Blk46– strongly suggests that the association is a result of mutual localization to lipid rafts. Consistent with this conclusion is the finding that co-precipitation of src-family kinases with CD20 does not occur efficiently in all B-cell lines, and can thus be assumed to be indirect.54
Is apoptosis secondary to raft aggregation?
As described above, evidence strongly favours an indirect association between CD20 and src-family kinases via lipid rafts. In view of this, and that CD20-mediated apoptosis is src-family kinase-dependent and requires cross-linking – in many cases, hyper-cross-linking – the apoptotic response seems most likely to result from transactivation of src-family tyrosine kinases upon clustering of rafts (Fig. 2). This model predicts that cross-linking any raft-associated molecule would have similar effects. This appears to be the case, because cross-linking glycosphingolipid GM1 or glycosyl-phosphatidylinositol-linked proteins, which are characteristically localized to lipid rafts, also induces an apoptotic response in B cells.55
Apoptotic effects of CD20 mAbs are not observed in all CD20+ cell lines and are inconsistently observed in the same cell line by different investigators (Table 2). The model of CD20-mediated B-cell death via aggregation of lipid rafts offers an explanation for such variable results. Although there is no information available yet on the regulation of lipid raft formation during B-cell development and activation, it is clear from work in other cell systems that such regulation occurs. For example, the raft composition of plasma membranes in naïve T cells is much lower than that of activated T cells, and caveolae, specialized raft microdomains found in non-haematopoietic cells, increase dramatically in number during differentiation of adipocytes.56–59 The lipid composition of plasma membranes may vary among B-cell lines, or during cell cycling, or in different cell culture conditions between laboratories. The integrity of lipid rafts depends on both sphingolipids and membrane cholesterol, and CD20 association with rafts is clearly cholesterol-dependent.60 Therefore, the ability of CD20 mAbs to induce activation of src-family kinases is expected to vary with the level of cholesterol – and lipid rafts – in the plasma membrane.
The function of CD20 has not yet been fully elucidated, but evidence of its participation in calcium influx is growing. The original observations of Bubien et al. demonstrating enhanced calcium conductance in cells expressing CD20 ectopically,26 were confirmed and extended by Kanzaki et al.61–63 Unfortunately, the generation of CD20-deficient mice did not yield an obvious phenotype,64 perhaps because of the coexpression in B cells of several members of the newly described MS4A family of related genes.65,66 Additionally, there are amino acid differences between murine and human CD20 in the sequence known to be critical for CD20 raft association (residues 219–225), raising the possibility that CD20 may localize and function differently in murine and human B cells.
As described earlier, it is now clear that CD20 is constitutively associated with lipid rafts. Whatever the function of CD20, it is evidently performed in the environment of lipid raft microdomains. What, then, is the relevance, if any, of antibody-induced translocation of CD20 from the soluble to the insoluble fractions of 1% Triton X-100 cell lysates? Highly purified Fab fragments can also induce this change, and cross-linking Fab with secondary (anti-light-chain) antibodies has no additional effect (H. Li, unpublished data). From this it can be concluded that cross-linking is not required. All CD20 mAbs induce CD20 translocation, but they do so to varying degrees (refs 27,48 and see Table 1). This variability appears to be independent of the isotype and affinity of the mAb, but rather is an epitope-dependent characteristic.27 These observations indicate the sensitivity of a specific region of the extracellular loop to interactions that could be mediated either by a ligand or by lateral engagement with a cell surface receptor. Translocation of CD20 from rafts that are insoluble only in a low concentration of Triton X-100 (or other non-ionic detergents at 1%), to rafts that are insoluble in 1% Triton X-100, may reflect an increase in the affinity of CD20 for lipid rafts and/or stabilization of the rafts. The consequence of this is presently unknown, but co-localization of the B-cell receptor with CD20 (R. Petrie, in press) suggests their functional interaction within the environment of lipid rafts, possibly localizing calcium influx to the site of antigen engagement.
A final note
Not all anti-CD20 induced cellular responses can be attributed to lipid raft aggregation. Those that are should be inducible by the majority of CD20 mAbs. Some responses, such as c-myc up-regulation and homotypic aggregation, are only induced by one or a few mAbs. Both c-myc induction and homotypic aggregation are abrogated by pretreatment with tyrosine kinase inhibitors, but the required kinase activity may be downstream of the initiating signalling events. The molecular mechanisms underlying the activation of these signalling pathways are not known but could involve antibody-mediated effects on the open probability of the calcium channel that is regulated or formed by CD20. It is possible that this could also contribute to CD20-mediated apoptosis, independently of the indirect mechanism via aggregation of lipid rafts proposed here. Modulation of membrane cholesterol, on which the integrity of lipid rafts depends, should distinguish these alternative mechanisms. Finally, it will be important to test the prediction that the size or abundance of lipid rafts in B-cell membranes is a contributing factor to the susceptibility or resistance of patients to CD20-mediated B-cell depletion.
The work of the authors described in this article was supported by the Canadian Institutes of Health Research (CIHR, formerly MRC). J. P. D. is a senior scholar of the Alberta Heritage Foundation for Medical Research (AHFMR). Thanks to members of the Deans laboratory, Cathlin Mutch and Ryan Petrie, and to Drs Kamala Patel and Ric Woodman, for their careful reading and critical appraisal of this article.