Origin and properties of soluble CD21 (CR2) in human blood

Abstract

By analysis with a panel of CD21 MoAbs it is shown that a large part of the soluble CD21 in human blood plasma is of the long isoform (CD21L), as judged by comparison with antigen produced by mouse L cells transfected with CD21L-cDNA and reactivity with the restricted CD21 MoAb R4/23. This is compatible with the hypothesis that soluble CD21 in the blood is mainly derived from follicular dendritic cells (FDC). Cells from a human keratinocyte cell line transfected with cDNA from the Burkitt lymphoma cell line Raji also produced soluble CD21L (sCD21L), whereas the short form of sCD21 (sCD21S) was the major component of sCD21 produced by the B lymphoblastoid cell line LICR-LON-HMy and the T cell line Jurkat. Confocal studies of FDC isolated from human tonsil revealed that CD21 was present in the cytoplasm. On gel filtration sCD21 from untreated serum has an apparent size considerably greater than the 130 kD found by SDS–PAGE analysis. This may be partly accounted for by the non-globular shape of the molecule, but may also indicate, as reported by others, that in its native state sCD21 is complexed with other proteins. However, no evidence of complexing with sCD23 or C3d could be found.

INTRODUCTION

Complement receptor type 2 (CR2), the cell surface receptor for the iC3b, C3dg and C3d components of complement and for the Epstein–Barr virus (EBV), is found on most mature B lymphocytes and 50% of bone marrow B cells, but it is not detectable on pre-B cells or plasma cells. Immunohistological studies indicate that CR2 is moderately expressed on mantle zone B cells and strongly on marginal zone B cells [1] but poorly on germinal centre B cells [2,3]. CR2 is a glycoprotein of molecular mass ≈ 145 kD containing a large extracellular portion made up of 15 or 16 structurally similar complement control proteins (CCP), each of about 60 amino acids, a single transmembrane region and a cytoplasmic tail of 34 amino acids [4,5]. The two N-terminal CCP contain the C3dg and EBV binding sites. Complexed N-linked oligosaccharides contribute about 35 kD to a peptide backbone of 110 kD, but are not required for binding to C3d or for membrane insertion [5,6].

A soluble form of CR2 (sCR2) is found in normal human serum and in increased amount in the sera of patients with chronic lymphocytic leukaemia of B cell type [7]. The sCR2 in serum and culture supernates appears to consist of the entire extracellular portion of surface CR2 and has C3d and EBV binding properties [8]. Surprisingly, the sCR2 in serum does not appear to be derived solely or even principally from B cells, since it is present in undiminished amount in sera from patients with agammaglobulinaemia [9]. From which cell type does it originate? About half the T cells in blood express very low levels of surface CD21 [9,10]; CD21 is also present on thymocytes [11] and fetal astrocytes [12]. T cells probably produce very little soluble CR2. We have been unable to produce sCR2 by stimulation of peripheral T cells in culture, but have confirmed that the Jurkat T cell line expresses CD21 on the surface [13] and produces small amounts of sCR2. We considered that follicular dendritic cells (FDC) were a more likely source of blood sCR2 [9]. Until recently this hypothesis has been difficult to evaluate. FDC are present in a network in germinal centres (particularly the light zone) with long cytoplasmic processes which reach into the follicular mantle zone [14]. They exhibit strong CR2 expression.

Studies on CR2 gene products revealed that there is variable expression of two transcription products of the CR2 gene; an exon encoding a 60 amino acid CCP (exon 11; then known as 10a) can be alternatively spliced [15]. A roughly equal presence of the two species was found in mRNA from a human tonsil (presumably mainly derived from B cells and FDC) but mRNA lacking the 11th CCP was the predominant form in seven B lymphoblastoid cell lines (B-LCL) tested [15]. It is not known if there is any external control of the splicing mechanism, but there is evidence for persistent cell type specificity.

It has recently been shown that human FDC specifically express the long isoform of CD21 (CD21L) consisting of all 16 CCP, in contrast to B cells which characteristically express the short form lacking the 11th CCP [16]. A MoAb (7D6) was produced which selectively stained FDC networks on tonsillar and spleen sections and also stained isolated FDC, but did not stain cells in fetal thymus or liver or cells in bone marrow or peripheral blood or cells from most B-LCL. Two other MoAbs previously reported to be FDC-specific (DRC-1 and KiM4) also recognized the long form of CD21 [16]. The DRC-1 MoAb (also known as Dend-1 and R4/23) was thoroughly investigated immunohistologically by Naiem et al. [17], who concluded that R4/23 detects an antigen present in high density on FDC and at a lower density on a B cell subset. B cells in the FDC network were unreactive, whereas B cells in the FDC-deficient area just outside the mantle and some B cells in the outer zone of the white pulp were reactive. It would be anticipated that soluble forms of CR2 would similarly exist in long and short isoforms, and this study confirms that this is so. We have also studied the intracellular distribution of CR2 in FDC.

MATERIALS AND METHODS

Monoclonal antibodies

The CD21 MoAbs BU-32, BU-33, BU-36, BU-42, BU-80 and BU-81 were produced in this laboratory and authenticated in the IVth and Vth International Leucocyte Workshops. The CD23 MoAb MHM6 was kindly donated by Dr M. Rowe (Department of Cancer Studies, University of Birmingham, UK) and was used in conjunction with a MoAb to a different CD23 epitope, EBVCS-5 (a gift from Dr B. Sugden, McArdle Laboratory, Madison, WN). A culture supernate containing the putative FDC-specific MoAb R4/23 [17] was a gift from Dr David Mason (Oxford, UK). The properties of a different FDC-specific MoAb, BU-10 (2BF11) have already been described [18]. Three CD55 MoAbs (BU-14, BU-84 and BU-97) were also tested. The CD19 MoAb used to test for CD19 in complexes with sCD21 was BU-12.

Sheep anti-human C3d antibody was prepared from a polyclonal sheep antiserum.

Affinity purification of CD21 antigens

A column (14 × 1.4 cm) was prepared by coupling purified BU-32, BU-33 and BU-35 MoAbs to cyanogen bromide (CNBr)-activated Sepharose 4B. Antigen was eluted with 3 m KCNS and dialysed against PBS. KCNS (3 m) was preferred to low pH buffer as the elutant in order to avoid antigen aggregation. This high salt concentration is likely to disrupt pre-existing non-covalently bound complexes which may or may not reform on restoration of isotonicity.

Gel filtration

This was performed on columns of Sephacryl S-300 beads (Pharmacia-LKB, Uppsala, Sweden) and the composition of the running buffer was 0.05 m phosphate buffer pH 7.5 containing 0.2 m NaCl and 0.02% sodium azide. Culture supernates were tested directly and after affinity-purification of sCD21. Sera were tested by direct addition to the column without prior purification. Triton extracts of whole cells were fractionated in buffer solutions containing 0.02% Triton X-100. Samples (0.05 ml) of fractions from all columns were titrated for CD21 and CD23 using a 1:1 mixture of sheep erythrocytes coated with BU-32 and BU-35 (CD21) or MHM6 and EBVCS-5 (CD23).

Radioactive labelling and autoradiographic detection of sCD21 antigen after SDS–PAGE analysis

Washed cells (10⁸) of the B-LCL LICR-LON-HMy which strongly expresses CD21 were suspended in 100 ml of methionine-free RPMI 1640 medium containing 10% fetal calf serum (FCS); after addition of 500 μCi of ³⁵S-methionine (Amersham, Aylesbury, UK) the cells were grown for 2 days at 37°C. The cells were spun down and the supernate harvested, brought to 0.05% azide and stored at 4°C. The cell pellet was washed and dispersed in 4 ml of 0.2% Triton X-100 containing protease inhibitors. After 1 h insoluble material was removed by centrifugation at 12 000 g for 10 min. CD21 antigen was isolated from culture supernates and cell extracts by affinity chromatography, dialysed for 2 days against 1.2 mm sucrose and freeze-dried for SDS–PAGE analysis. Gels were stained with coomassie blue, washed and dried, then apposed to x-ray film with an intensifying screen and developed for 2 weeks.

Detection and measurement of CD21 by haemagglutination

Washed sheep erythrocytes from a selected animal were coated with purified CD21 MoAb (usually BU-32 or BU-35) by a modified chromic chloride technique [19]. For fuller investigation of the properties of CD21 antigens and MoAbs, use was made of the fact that sheep erythrocytes coated with a single CD21 MoAb are able to bind sCD21 antigen through the single epitope expressed, but are not agglutinated because no cross-linkage is possible. Inclusion of a second CD21 MoAb binding to a different epitope topographically distinct from that bound by the sheep erythrocyte-bound MoAb will agglutinate the cells (Fig. 1). The indicator MoAb on the sheep erythrocytes in this synergy test system was usually BU-32, a MoAb previously shown to synergize with all 27 known CD21 MoAbs from the Vth workshop [19,20]. The test is extremely sensitive and reproducible. By standardizing the amount of soluble MoAb the test could also be adapted to the titration of antigen.

Fig. 1.

Diagram showing the basis for the detection and measurement of sCD21 by haemagglutination. S, Sheep erythrocytes coated with CD21 MoAb no. 1. The free antibody depicted refers to CD21 MoAb no. 2 in solution. The sCD21 antigen shown is sCD21L and complement control protein (CCP) no. 11 is shaded.

Cell lines and transfectants

The B-LCL LICR-LON-HMy and the Jurkat T cell line were grown in RPMI 1640 medium supplemented with 10% FCS. An SV40-transformed human adult skin keratinocyte cell line which had been transfected with CD21 DNA from Raji Burkitt lymphoma cells was grown in RPMI 1640 medium containing 10% FCS and the selective antibiotic geneticin. This line was kindly provided by Professor L. Young (Department of Cancer Studies, University of Birmingham, UK). Supernates from this line contained high concentrations of sCD21. The cell line L7D6, kindly provided by Dr Y.-J. Liu (Schering-Plough Laboratory for Immunological Research, Dardilly, France) is a CD21L cDNA transfectant of mouse L cells. Mouse Ltk⁻ cells (L cells) stably expressing the antigen recognized by MoAb 7D6 (selected because it specifically stained FDC networks on tonsillar and spleen sections) were generated by cotransfection with a neomycin-resistant plasmid by the calcium phosphate method. After culture in G418, surviving cells were selected for 7D6 antigen expression by flow cytometry.

sCD21 in human blood

Blood from eight normal subjects was collected in EDTA anticoagulant and centrifuged as soon as possible to minimize the contribution of sCD21 released from B cells. Antigen was affinity-purified from EDTA plasma on the triple CD21 MoAb column, eluted with 3 m KCNS, dialysed against PBS, brought to 2% FCS and filtered to sterilize.

RESULTS

Size and general properties of sCD21

Autoradiographs of affinity-purified sCD21 from a B-LCL metabolically labelled by culture with ³⁵S-methionine and analysed in 12.5% and 7% SDS gels revealed a principal band at ≈ 130 kD and another at ≈ 30 kD (Fig. 2).

Fig. 2.

Autoradiographs of SDS gels from ³⁵S-labelled sCD21 from a B lymphoblastoid cell line (B-LCL). Lane a, molecular mass standards, 12.5% gel, markers for lane b: glutamic dehydrogenase (53 kD), transferrin (76 kD), β-galactosidase (116 kD) and α₂-macroglobulin (170 kD); lane b, affinity-purified ³⁵S-sCD21 run in a 12.5% gel. Band 1 = 30 kD, band 2 = 130 kD; lane c, the same preparation of affinity-purified ³⁵S-sCD21 run for a shorter time in a 7% gel. Standards not shown. Band 2 = 130 kD.

In order to assess directly the state of sCD21 in B-LCL supernates and sera, gel filtration was performed on Sephacryl bead columns. Surprisingly it was found that sCD21 from all sources exhibited an apparent molecular mass of ≈ 320 kD; membrane CD21 was of a slightly larger size (Figs 3 and 4). Two explanations for this unexpectedly high figure were considered: (i) the reference standard proteins were giving a misleading result because they were all globular proteins whereas the sCD21 has a ‘string-of-beads’ conformation [4–6]; (ii) in its native state the sCD21 in sera and culture supernates is possibly complexed with other proteins [21–23].

Fig. 3.

Gel filtration of affinity-purified CD21 from a B lymphoblastoid cell line (B-LCL). ○, CD21 from culture supernate; •, CD21 from detergent extraction of washed cells. Molecular mass markers: 5 = bovine albumin (67 kD), 4 = catalase (232 kD), 3 = ferritin (440 kD), 2 = thyroglobulin (696 kD), 1 = void volume.

Fig. 4.

Gel filtration patterns of untreated sera from two patients with chronic lymphocytic leukaemia ((a) and (b)). Sephacryl S-300 beads, bed vol. 262 ml. Samples (0.05 ml) were titrated for sCD21 (○) and sCD23 (•); the result shown is the log₂ titre. Standards: 6 = ribonuclease (13.7 kD), 5 = bovine albumin (67 kD), 4 = aldolase (158 kD), 3 = catalase (232 kD), 2 = ferritin (440 kD), 1 = void volume.

Proteins which might be complexed with sCD21 include sCD23 [21] and C3 fragments [22,23]. sCD23 is present at only a very low level in normal sera, but sera from patients with B-type chronic lymphocytic leukaemia (B-CLL) contain relatively large amounts of sCD21 and sCD23. When two B-CLL sera were fractionated the sCD21 and sCD23 were clearly separated with no evidence of interaction (Fig. 4). C3d antigen could not be detected in affinity-purified sCD21. Tests for CD19 were also negative. It is, however, possible that some other protein may persistently complex with sCD21. The apparent larger size could not be accounted for by dimerization or trimerization of the sCD21, as this would be detected by the control (single MoAb) titration in the assay system.

Relative binding capacity of CD21 antigen from various sources to a panel of CD21 MoAbs

As shown in Table 1, all the CD21 MoAbs of the test panel reacted positively with all the antigen preparations except MoAb R4/23, which reacted with blood plasma sCD21 and with the antigens produced by the Raji transfectant and the L7D6 cell line, but failed to bind to the antigens from the B-LCL or Jurkat T cell lines. Since the MoAb in the test system (BU-32) has been shown to synergize with all known CD21 MoAbs, a negative result in the haemagglutination test indicates that the target epitope for the MoAb is absent from the antigen under test or unavailable. The reactions of R4/23 to the B-LCL and Jurkat antigens were not, however, completely negative, implying that a very small amount of the CD21L isoform may be produced by these cells. Both forms of mRNA have been found in all cells producing CR2, but there is marked quantitative variation in their expression [15]. It is also possible that there is some degree of antigenic cross-reactivity, as already suggested [16]. Further tests were performed with the same panel of CD21 MoAbs, but instead of using BU-32 as the synergizing MoAb, BU-35 was used instead (Table 2). This MoAb has been shown to synergize with only a proportion (17/27) of known CD21 MoAbs [19]. The MoAb R4/23 failed to synergize with BU-35 using any of the antigen preparations apart from a weak reaction with L7D6. This may indicate that the epitope recognized by R4/23 is on domain 11 and the epitope recognized by BU-35 is close to it, perhaps on an adjacent CCP. The poor synergy of BU-80 with BU-35 was confirmed. Another established FDC MoAb (BU-10) was completely negative in all tests with sCD21, indicating that the epitope recognized is not present on either isoform of CD21. Three CD55 MoAbs and five other FDC-binding MoAbs (not listed) also gave completely negative results.

Table 1.

Binding of various sCD21 antigens to a panel of CD21 and other MoAbs using BU-32 as the synergizing MoAb

Table 2.

Binding of sCD21 antigens to a panel of CD21 MoAbs using BU-35 as the synergizing MoAb

Measurement of long and short isoforms of sCD21 in cell supernates and blood plasma

Affinity-purified sCD21 preparations from cell lines and blood plasma were adjusted to give approximately the same titre against a 1:1 mixture of sheep erythrocytes coated with BU-32 and BU-35 MoAbs and the relative binding capacity of the five sCD21 antigens for MoAb R4/23 was determined by titration in a standard dilution of this MoAb (Table 3). All the CD21 antigens gave a similar titre with the BU-32/BU-35 combination, but very little reactivity of the B-LCL and T-Jurkat antigens with the BU-32/R4/23 combination was recorded. Assuming that the antigen produced by the L7D6 transfectant line consists entirely of the long form of sCD21 [16], it is evident that a high proportion of the sCD21 in blood plasma is also of the long form.

Table 3.

Titration of sCD21 antigen in cell supernates and blood plasma

CD21 is present in the cytoplasm of FDC

The function of CD21 is usually considered solely in terms of cell surface expression. However, when examined by confocal microscopy, serial sections of FDC isolated from human tonsil revealed that a major part of the cellular CD21 was present in the cytoplasm in the supranuclear region (Fig. 5).

Fig. 5.

Confocal fluorescence analysis of the cellular distribution of CD21 in a follicular dendritic cell (FDC) isolated from human tonsil. Eight serial optical sections collected at 0.5-μm increments in the ordinate (a–h; a = top surface, h = bottom surface). BioRad 600 fitted with a krypton-argon laser. Magnification × 400. Note the high level of CD21 in the supranuclear region.

DISCUSSION

Immunohistological and flow cytometric studies in several laboratories, including our own, have yielded results which can now be interpreted to mean that CD21L, the long isoform of CD21, is absent or present at a very low level on peripheral blood lymphocytes and on most B lymphocytes in lymphoid tissues, but is strongly expressed on FDC [14,16–18].

It is not practicable to measure the proportions of the long and short isoforms of soluble CD21 in an antigen preparation by their difference in size. Native CD21 is heavily and variably glycosylated and the broad band at ≈ 130 kD on SDS–PAGE analysis would mask the 6-kD difference contributed by CCP 11. Indeed, it is thought that increased glycosylation may account for the fact that CR2 on astrocytes and T cells is 10 kD larger than that expressed by Raji cells [12,13]. Clear cut results can, however, be obtained by serological analysis. Our results show that a high proportion of the soluble form of CD21 in blood plasma is of the long isoform, which may indicate that it originates from FDC. We have tested a different FDC-specific MoAb, BU-10, and a number of CD55 MoAbs as controls and found that none of these bind to sCD21. The highly selective quality of R4/23 MoAb for CD21L is most easily explained by assuming that it binds to determinants on the 11th CCP of CD21L, which is absent from CD21S, but this is not yet proven.

On immuno-electronmicroscopy of frozen sections of tonsil tissue CR2 has been found to be localized on the plasma membrane and dendritic processes of FDC [24]. The function of the surface CR2 is the retention and presentation of antigen in immune complexes. However, when FDC isolated from tonsil are examined by confocal microscopy CR2 is clearly shown to be present in the cytoplasm in the supranuclear region. This intracellular pool may be destined for cell surface expression. Soluble forms of cell surface receptors are usually generated by restricted proteolytic cleavage close to the cell membrane by endogenous enzymes rather than by alternative splicing of receptor mRNA, but there appear to have been no definitive studies on CR2. sCR2 is present in blood at about 10 times the level found in the culture supernate of a high producer B-LCL. In culture supernates and blood plasma it is remarkably stable; levels are not changed after storage for several months at 4°C. It remains in the blood as a functional molecule retaining C3d and EBV binding capacity [8] and unlike sCD23 does not pass through into the urine [25]. It might have a role in protection against EBV infection [26] and suppression of immune responses [27]. In studies on fetal astrocytes Gasque et al. [12] found that the amount of CR2 extracted from the cells was 5–10-fold greater than surface CR2. The CR2 was not redistributed globally but was localized in discrete patches on the cell body, particularly at the ends of processes.

The anomalous high molecular mass of sCD21 recorded in simple gel filtration analysis of sCD21 in serum and B-LCL culture supernates can partly be accounted for by the globular nature of the reference proteins, but it is likely that native sCD21 might be complexed with other proteins. Aubry et al. [21] have reported that CD23 interacts with N-linked oligosaccharide on the extracellular domains of sCD21 and there is another report that sCD23 and C3 fragments are associated with CD21 in normal serum [23]. We were unable to detect any interaction of sCD21 and sCD23 in leukaemic serum. The demonstration by Schendinger et al. [22] that there is an alternative pathway activation of complement in serum which leads to the deposition of C3 fragments on the surface CR2 of Raji cells would suggest that C3 fragments might be bound to soluble CD21, but we have not so far found evidence for this using CR2 affinity-purified from plasma, which may mean that binding occurs to surface but not to soluble CR2, or that the binding does not survive affinity purification.