Antibodies from patients and mice with autoimmune diseases react with recombinant hnRNP core protein A1

Journal of Autoimmunity

(1988) 1,73-83

Antibodies from Patients and Mice with Autoimmune Diseases React with Recombinant hnRNP Core Protein Al

LeeAnn Jensen, Edward L. Kuff, Samuel H. Wilson, Alfred D. Steinberg and Dennis M. Klinman National National

Institute of Arthritis,

Cancer Institute, National

Musculoskeletal

and Skin Diseases, and Bethesda, MD 20892,

Institutes of Health, USA

Sera from autoimmune patients and normal volunteers were tested for antibodies to the Al core protein of heteronuclear ribonucleoprotein (hnRNP) particles by ELISA and Western blot assays. The Al protein used in these studies was produced by recombinant DNA technology. Thirtyseven per cent of patients with systemic lupus erythematosus produced anti-Al antibodies, compared to 7% of normal controls. Several strains of autoimmune mice were also analysed. They spontaneously made hightitered responses to Al. Normal strains showed very low anti-Al response unless they were specifically immunized with the antigen. Following immunization, normal mice also made high-titered responses to the Al protein. These studies demonstrated that the Al protein itself is an autoantigenic component of hnRNP particles. Our studies uncovered no evidence of linkage between the production of anti-Al and of anti-DNA antibodies.

Introduction

Antibodies directed against native DNA and a variety of nuclear antigens are commonly found in the sera of patients with systemic lupus erythematosus [ 1, 21. The specificities of these antibodies have been investigated to better understand the physiopathology of this disease and for use as diagnostic tools. Lerner and Steitz discovered that anti-Sm and anti-U1 RNP (ribonucleoprotein) autoantibodies bound to a common set of RNA binding proteins which were associated with unique Correspondence to: Alfred D. Steinberg, M.D., Cellular Immunology Rm 9N 218, National Institutes of Health, Bethesda, MD 20892, USA.

Section, NIAMS,

Bldg 10,

73 0896-8411/88/010073+

11 $02.00/O

0 1988 Academic Press Limited

74

L. Jensenet al.

RNA species [3]. Since their discovery, many of the serologically defined autoantibodies (e.g., Anti-Ro, anti-La, and anti- Jo- 1) have been matched with important nucleoprotein ligands [4-61. Another component of the nucleus which induces autoantibody formation is heteronuclear ribonucleoprotein (hnRNP). Previous studies have shown that patients with mixed connective tissue disease have antibodies to hnRNl?‘s in their sera as well as antibodies to small nuclear RNP’s (snRNP’s) [7]. HnRNP’s constitute a set of six structurally related and evolutionarily conserved polypeptides [8,9]. They associate with preprocessed mRNA (hnRNA) in transcriptionally active cells to form the ‘beads on a string’ seen in electron micrographs [lo]. The hnRNP core proteins have molecular weights ranging from 30,000 to 43,000 D, are basic in charge, and are rich in the aminoacid glycine [ 11, 121. The hnRNP core proteins are extremely sensitive to protease treatment. This is of interest because proteolytic cleavage of self antigens can lead to exposure of cryptic (hidden) determinants capable of inducing autoimmune responses. For example, autoantibodies are produced against the cryptic determinant exposed on mouse red blood cells by treatment with the enzyme bromelain [ 131. We investigated the presence of autoantibodies to ultrapure recombinant hnRNP protein Al (initially known as helix destabilizing protein [ 141). Recombinant Al was purified by gel chromatography from bacteria transfected with a lambda gt 11 clone carrying the A 1 gene isolated from a newborn rat brain cDNA library [ 141. The use of purified recombinant Al antigen allowed detection of the entire spectrum of antibodies which interact specifically with that protein. It eliminated the possibility that other hnRNP proteins, or sites created or masked by the interaction of several hnRNP core proteins, were being recognized. We found that immunization of normal mice with Al protein led to the production of a large anti-Al antibody response. Sera from patients with systemic lupus erythematosus (SLE), juvenile rheumatoid arthritis (JRA), and polymyositis (PM), and from normal subjects were assayed for anti-hnRNP Al and anti-DNA antibodies. Sera from lupus patients and autoimmune mice frequently contained antibodies which bound to Al protein, as well as to DNA, whereas normal control sera did not. Our studies uncovered no evidence that the production of anti-Al and anti-DNA antibodies were causally related.

Materials and methods Patients and samples

Sera were collected by venipuncture from patients and normal volunteers, after informed consent, at the Clinical Center of the National Institutes of Health (NIH), Bethesda, MD. Lupus patients fulfilled the American Rheumatism Association (ARA) revised criteria for diagnosis of SLE [ 151. JRA patients were also diagnosed according to the ARA criteria [ 161. Patients with polymyositis met the criteria outlined by Bohan & Peter [ 171. Sera from normal volunteers were obtained through the generosity of the Clinical Center Blood Bank. Serum samples were stored at - 20°C prior to use.

Autoantibodies to recombinant hnRNP core protein Al

75

Animals

All mice were bred and raised at the NIH. Mice were bled by puncture of the retro-orbital sinus under light anesthesia. Blood was allowed to clot for two hours at room temperature, after which sera were separated, centrifuged to remove cellular debris, and stored at - 20°C until used.

Detection and quantitation

of antibodies

Serum antibodies reactive with ssDNA, hnRNP, bromelain-treated mouse red blood cells (BrMRBC) or murine thymocytes (MTC) were detected by ELISA assay as previously described [ 181. In brief, Immulon I microtiter plates (Dynatech) were coated with 50 pg/ml of recombinant hnRNP [l 11, 10 pg/ml of heat denatured ssDNA (Sigma) or 5 x lo5 formaldehyde-fixed mouse thymocytes or bormelaintreated red blood cells. Plates were then blocked with 1 “/bBSA in phosphate buffered saline (PBS). On the basis of preliminary experiments, serum samples were diluted 1 : 50 in all assays except the anti-ssDNA assay, where they were diluted 1 : 250. Binding of serum antibodies to antigen-coated plates was detected using the appropriate peroxidase-labeled or phosphatase-labeled goat anti-human or goat anti-mouse IgG, IgM or total Ig antisera (Kierkegaard-Perry). The relative concentration of antibody in test samples was determined by comparison to a standard curve generated using appropriate high titered antisera.

Western blots

One microgram of recombinant hnRNP Al and two micrograms of phosphorylase B (Boehringer-Mannheim) were electrophoresed in Laemmli sample buffer [19] in each lane of a 10% sodium dodecyl sulfate polyacrylamide (SDS-PAGE) minigel at 15 mAmps. Proteins were transferred to nitrocellulose (0.20 l.trnpore size) according to the technique of Towbin et al. [20] for 3 h at 60 V using a Bio-Rad Trans-Blot cell. Nitrocellulose strips were incubated with 3% BSA in PBS for 3 h at room temperature. This was followed by washes with water, and PBS containing 1 y0 normal goat serum plus 1% BSA. Strips were incubated with a 1 : 25 dilution of normal or patient serum in PBS-0.05a/0 Tween-20 (500 yl final volume) for 3 h at room temperature. After extensive washing with PBS-Tween 20, strips were developed by treatment with peroxidase-conjugated goat anti-human IgG (Cappel) followed by diaminobenzadine. Proteins on strips not treated with antiserum were visualized using Ponceau S staining (Sigma).

Immunization

Mice were immunized intraperitoneally with 10 micrograms of recombinant Al protein, either in PBS or in complete Freund’s adjuvant (CFA, Difco). They were boosted with 10 l.tg of antigen in PBS, 2 or 4 weeks later.

76

L. Jensen et al. Table

1. Relative

serum

autoantibody normal mice Relative

Strain’

Sex

NZB NZB.xid

Al

M M

610+55 4+1 1lOf 12 7fl

MRL/lpr

M

MRL/++

M

BXSB BXSB BXSB.xid

M F M

19*5 9*1 1

CBA/J CBA/N

F F

4+1 4*1

levels of autoimmune

serum antibody

and

levels

ssDNA

BrMRBC

52+ 12 2+_2

980& 110 4+3

105 + 14 40f 16

240+ 110 4+3

1,000f210 4+5

64+8 15_+19 l&l

250f 170 l&l If1

28f4 3_+5 0.2iO.l

8+11 3+1

3+3 l&l

2+1 l&l

MTC

34+ 18 0.3kO.2

‘Sera from 5-10 mice of each strain were tested by ELISA for antibodies reactive with the antigen panel. Average antibody levels k standard error are shown. Relative autoantibody levels were determined by comparison to a standard curve generated with high-affinity antiserum. All mice were 6 months old except NZB males (10 months) and BXSB males (9 months).

Results Elevated

levels of anti-Al

in the sera of autoimmune

mice

Sera from normal and autoimmune mice were analysed for the presence of antibodies reactive with hnRNP Al (Table 1). Autoimmune NZB mice had lOO-fold higher levels of anti-hnRNP antibodies than did congenic non-autoantibody-producing NZB.xid animals. MRL-&r/&r mice expressed 15 times higher levels of anti-Al antibodies than did MRL- + /+ mice. Male BXSB mice, which develop accelerated autoimmunity, were found to have modestly elevated anti-Al antibody levels compared to normal BXSB female mice. Other nonautoimmune mice, including those of the CBA/J, CBA/N and BXSB.xid strains, had minimal expression of anti-Al antibodies (Table 1). The presence of anti-Al antibodies was compared to that of other autoantibodies. As seen in Table 1, autoimmune mice produced high concentrations of anti-DNA, anti-bromelain-treated mouse red blood cell (BrMRBC) and anti-murine thymocyte (MTC) antibodies, while normal mice did not. This pattern correlated with that found for anti-hnRNP antibodies in these same mice.

Anti-Al

levels inpatients

with autoimmune

diseases

To determine whether the Al protein was recognized by autoantibodies in the sera of humans with autoimmune diseases, patients’ sera were compared to normal control sera for binding to Al in ELISA and Western blot assays. A typical Western blot is shown in Figure 1. The same antigen preparation was used in Western blot and

Autoantibodies to recombinant hnRNP core protein Al

77

-P

-A

12

34

Figure 1. Auto-antibodies to Al are detected on a Western blot developed with autoimmune human serum. One microgram of Al and 2 urn of phosphorylase B were electrophoresed on a 10% polyacrylamide gel. Proteins were transferred to nitrocellulose and processed as described in the text. Strip 1, anti-Al rabbit serum control; strip 2, human serum reactive with Al; strip 3, human serum without anti-Al antibodies; strip 4, Ponceau S strain of phosphorylase B and Al.

ELISA assays. Thirty-seven per cent of SLE patients had elevated anti-Al antibody levels, while 24% of JRA patients, 4% of PM patients, and 7% of controls had similarly elevated levels (Figure 2, Table 2). The isotypes of the anti-hnRNP antibodies were determined. Eighty-one per cent of the positive sera contained IgM anti-hnRNP antibodies. IgG anti-hnRNP antibodies were present (either alone or in combination with IgM) in 37% of the positive sera. There was no significant difference in the isotypes of the autoantibodies present in any of the patient groups or the normal controls. A review of the patients’ medical records was conducted to determine whether the presence of anti-hnRNP correlated with disease activity. We found that 65% of SLE patients with increased anti-hnRNP had active disease.

Ability

of Al to induce an antibody response

To test the immunogenicity of the Al protein, mice of several strains were immunized with purified Al in complete Freund’s adjuvant. Sera from these mice showed a substantial increase in anti-Al antibody levels by 2 weeks after immunization, and a further increase in anti-Al levels following secondary immunization (Figure 3). Moreover, Al was immunogenic even in the absence of adjuvant [Figure 4(a)] inducing a 30-fold increase in specific antibody.

78

L. Jensen et al.

Y .

..

.

:

G

,“,... .--y~------_

”

AU&U4 YL

NL

SLE

JRA

Figure 2. Anti-Al levels were measured by ELISA. Relative antibody concentration was determined by comparison to a standard curved generated using serially-diluted, high-titered patient serum. Samples were considered positive if the anti-Al level exceeded two standard deviations of the mean of the normal volunteers.

Table

2. Anti-Al

antibodies are made by patients with autoimmune disease Autoantibody

Group SLE

JW PM Control

Number of subjects

positive’

Al

DNA

170

63 (37)

75 (44)

34 (20)

71 24 98

17 (24) l(4) 7 (7)

12 (17) 0 (0) 11(11)

4 (6) 0 (0) 3 (3)

Al and DNA’

iSerum autoantibodies were measured by ELISA. Relative levels were determined by comparison to a standard curve generated with serially-diluted, hightitered patient serum. Samples which had levels exceeding two standard deviations of the mean of the normal donors were considered positive. Percent of the total response is shown in parentheses. 2The number of patients with anti-Al antibodies which simultaneously express elevated levels of anti-DNA antibodies are shown.

Autoantibodies to recombinant hnRNP core protein Al

79

Time (weeks)

Figure 3. Anti-Al levels increase with time following immunization. Mice were immunized with 10 ug of Al protein in complete Freund’s adjuvant and boosted two weeks later with 10 pg of antigen in PBS. Antibody levels were measured by ELISA at two week intervals. Values are the A,,, of sera diluted l/SO.

Figure 4. Induction of anti-Al antibodies is not accompanied by anti-DNA antibody production. Mice were immunized with 10 ug of Al protein in PBS (solid symbols), or were not immunized (open symbols), in groups of four. Animals were boosted at 4 weeks with 10 pg of antigen. Antibodies to Al (a) and DNA (b) were measured by ELISA. Relative anti-Al and anti-DNA concentrations were determined by comparison to a standard curve generated using serially diluted, high-titered antiserum.

Association of anti-Al

with anti-DNA

It has been proposed that autoantibodies arise as a consequence of abnormal idiotype-anti-idiotype regulation [25-281. HnRNP proteins bind to single-stranded DNA, as well as to RNA [ 141. Thus, an antibody to Al might also be an anti-DNAidiotype antibody. This anti-idiotype antibody might induce anti-DNA antibodies. To determine if there was an association between the expression of anti-DNA antibodies and anti-Al antibodies, we asked whether anti-DNA antibodies might be induced in animals immunized with Al. The results indicate that anti-DNA antibodies were not induced (Figure 4).

80

L. Jensenet al. Table 3. Multiple serum samples with antibodies to DNA, Al or both Number of patients studied Average number of samples per patient Average time between samples (months) Patients with only anti-Al ’ Patients with only anti-DNA’ Patients with both anti-Al and anti-DNA Patients without either Patients whose anti-DNA preceded anti-Al Patients whose anti-Al preceded anti-DNA

29

3 15 5 10 4 8 1 1

‘Serum anti-Al and anti-DNA antibody levels were measured by ELISA. Samples were considered positive if the autoantibody concentration was two standard deviations above the mean of the normal donors. Relative antibody levels were determined by comparison to serially-diluted, high-titered patient sera.

The expression of anti-DNA antibodies was also examined in normal and autoimmune human sera (Table 2). Of the 63 SLE patients expressing increased anti-Al antibodies, 54% simultaneously expressed elevated anti-DNA antibody levels (34 out of 63, Table 2). Although not statistically significant this frequency was greater than that of anti-DNA antibodies in patients not expressing anti-hnRNP antibodies. To pursue this issue further, we sought to determine whether there was a sequential relationship between the appearances of anti-Al antibodies and anti-DNA antibodies in individual SLE patients over a 3-year period. As shown in Table 3, only one of 10 patients with anti-DNA antibody activity became anti-Al antibody positive. Only one out of five positive patients for anti-Al antibodies developed anti-DNA antibodies.

Discussion Sera from patients and mice with autoimmune diseases were assessed for reactivity to recombinant Al protein, a specific component of the hnRNP complex. Sera from normal mice showed only background binding to plates coated with hnRNP Al protein, whereas sera from autoimmune NZB, MRL-Zpr/Zpr, and BXSB male mice all had significantly elevated titers of anti-Al antibodies. Thirty-seven per cent of humans with SLE and 24% of JRA patients had elevated levels of autoantibodies capable of binding to purified Al. These findings confirm and extend previous reports showing that hnRNP’s are recognized as autoantigens in 35% of SLE patients [21]. Moreover, these data represent the first demonstration that antihnRNl? antibodies are able to bind directly to an individual recombinant component of the hnRNP complex. In a report analysing the reactivity of sera from patients with mixed connective tissue disease (MCTD) [7], autoimmune sera were tested for activity against whole hnRNl? particles or nuclear extracts, rather than against pure components of these nuclear antigens. These authors concluded that the MCTD sera reacted with

Autoantibodies to recombinant hnRNP core protein Al

81

Ul -RNP and hnRNA, but not with any protein component of hnRNP. The possibility of a small amount of protein remaining in their protease-treated preparation was not ruled out. However it is unlikely that Al would have remained, since the Al protein is protease sensitive [22, 231. In contrast, we have demonstrated that the reactivity of SLE and JRA autoimmune sera with hnRNP includes reactivities against the Al protein component of hnRNP. The discrepancy between our results and those of Fitzler et al. [7] points out the advantage of using purified recombinant antigens for studies of this nature. What drives autoantibody production? It has been proposed that nucleosomes might be immunogenic [2]. It is also possible that anti-Al antibodies are generated in response to cross-reacting foreign antigens. In this report evidence is provided that specific antigenic stimulation can lead to production of anti-Al autoantibodies. Injection of 10 lrg of the protein without adjuvant induced high-titered anti-Al responses in non-autoimmune mice. This observation is of special interest due to the ease with which Al is degraded in vivo. Because cryptic determinants of Al may be hidden from the immune system, determinants exposed by degradation may be immunogenic rather than tolerogenic. In addition to specific immunization, polyclonal B-cell activation may be critical to autoantibody production [24]. We believe that the combination of polyclonal activation and specific immunization leads to most autoantibodies in SLE. It has also been proposed that idiotype-anti-idiotype networks induce and regulate autoantibody production [25-271. Since both hnRNP and anti-DNA antibodies bind to DNA, anti-(anti-DNA) antibodies might regulate the expression of anti-Al antibodies. This would be consistent with Jerne’s network hypothesis [28]. All of the autoimmune mice studied coexpressed these two autoantibodies. In contrast, although 54 “,&of the human lupus patients who had elevated anti-Al antibodies also had increased anti-DNA antibodies, coexpression of these autoantibodies was not always observed. There was no evidence that expression of one autoantibody led to the subsequent expression of the other, despite the ability to link anti-hnRNP and anti-DNA antibodies in a theoretical regulatory idiotype network. Furthermore, normal mice which were injected with Al did not develop anti-DNA antibodies. Therefore, we were unable to support the hypothesis that an idiotype network may represent an important mechanism for induction of anti-DNA antibodies following an immune response to Al. We are presently generating monoclonal anti-Al and anti-DNA antibodies from individual autoimmune or Al-stimulated normal mice to further investigate the possibility of coexpression.

Acknowledgements Dr Jensen and Dr Klinman are supported by Arthritis Foundation Fellowships.

References 1. Tan, E. M. 1982. Autoantibodies to nuclear antigens (ANA): their biology and medicine. Adv. Immunol. 33: 167-240 2. Hardin, J. A. and T. Mimori. 1985. Autoantibodies to ribonucleoproteins. Cl&. Rheum. Dis. 11: 485-505

82

L. Jensen et al.

3. Lerner, M. R. and J. A. Steitz. 1979. Antibodies to small nuclear RNAs complexed with proteins are produced by patients with systemic lupus erythematosus. Proc. Nat1 Acad. Sci. (USA) 76: 5495-5499 4. Lerner, M. R., J. A. Boyle, J. A. Hardin, and J. A. Steitz. 1981. Twonovel classes of small

5.

6.

7.

8.

9.

10.

11.

12.

ribonucleoproteins detected by antibodies associated with lupus erythematosus. Science 211: 400-402 Rinke, J. and J. A. Steitz. 1982. Precursor molecules of both human 5s ribosomal RNA and transfer RNAs are bound by a cellular protein reactive with anti-La lupus antibodies. Cell 29: 149-159 Rosa, M. D., J. P. Hendrick, M. R. Lerner, and J. A. Steitz. 1983. A mammalian tRNAHi”containing antigen is recognized by the polymyositis-specific antibody anti- Jo-l. NucZeic Acid Res. 11: 853-869 Fritzler, M. J., R. Ali, and E. M. Tan. 1984. Antibodies from patients with mixed connective tissue disease react with heterogeneous nuclear ribonucleoprotein or ribonucleic acid (hnRNP) ofthenuclear matrix.J. Zmmunol. 132: 1216-1221 Wilks, A. F. and J. T. Knowler. 1980. The analysis of the proteins of heteronuclear ribonucleoprotein particies on two-dimensional acrylamide gels. Electrophoresis 1: 155-158 Leser, G. P., J. Escara-Wilke, and T. E. Martin. 1984. Monoclonal antibodies to heterogeneous nuclear RNA-protein complexes: The core proteins comprise a conserved group of related po1ypeptides.g. Biol. Chem. 259: 1827-1831 Samarina, 0. P., E. M. Lukanidin, J. Molnar, and G. P. Georgiev. 1968. Structural organization of nucleoprotein complexes containing DNA-like RNA. J. Mol. Biol. 33: 25 l-263 Beyer, A. L., M. E. Christensen, B. W. Walker, and W. M. LeStourgeon. 1977. Identification and characterization of the packaging proteins of core 40s heteronuclear ribonuclear protein particles. CeZl11: 127-138 Fuchs, J. P., C. Judes, and M. Jacob. 1980. Characterization of glycine rich proteins from the ribonucleoproteins containing heteronuclear ribonucleic acid. Biochemistry 19: 1087-1094

13. Cunningham, A. J. 1974. Large numbers of cells in normal mice produce antibody to components of isologous erythrocytes. Nature (London) 252: 749-751 14. Cobianchi, F., D. N. SenGupta, B. Z. Zmudzka, and S. H. Wilson. 1986. Structure of rodent helix-destabilizing protein revealed by cDNA cloning. J. BioZ. Chem. 261: 3536-3543

15. Tan, E. M., A. S. Cohen, J. F. Fries, A. T. Masi, D. J. McShane, N. F. Rothfield, M. G. Schaller, N. Talal, and R. J. Winchester. 1982. The 1982 revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum. 25: 1271-1277 16. JRA Criteria Subcommittee of the Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association. 1977. Current proposed revisions of the JRA Criteria. Arthritis Rheum. 20 (suppl): 195-199 17. Bohan, A. and J. B. Peter. 1975. Polymyositis and dermatomyositis (first of two parts). N. EngZ. J. Med. 292: 344-347

18. Klinman, D. M. and A. D. Steinberg. 1987. Novel ELISA spot assays to quantitate B cells specific for T cells and bromelated red blood cell autoantigens. 3. Zmmunol. Methods 102: 157-164

19. Laemmli, U. F. 1970. Cleavage of structural proteins during the assembly of the head bacteriophage T4. Nature (London) 227: 680-685 20. Towbin, H., T. Staehelin, and J. Gordon. 1979. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: Procedures and some applications. Proc. Natl Acad.

Sci. (USA)

76: 4350-4354

21. Hardin, J. A., C. Gelpi, J. Downs, and T. Minori. 1985. Autoantibodies to hnRNP in patients with SLE. Chin. Res. 33: 507a (Abstr.) 22. Planck, S. R. and S. H. Wilson. 1985. Native species of helix destabilizing protein-l in mouse myeloma identified by antibody probing of western blots. Biochem. Biophys. Res. Comm. 131: 362-369

Autoantibodies

to recombinant

hnRNP core protein

Al

83

23. Pandolfo, M., 0. Valentini, G. Biamonti, and S. Riva. 1985. Single stranded DNA binding proteins derive from hnRNl? proteins by proteolysis in mammalian cells. Nucleic Acids Res. 13: 65776590 24. Klinman, D. M. and A. D. Steinberg. 1987. Systemic autoimmune disease arises from polyclonal B cell activation. 3. Exp. Med. 165: 1755-1760 25. Hahn, B. H. and F. M. Ebling. 1984. Suppression of murine lupus nephritis by administration of an anti-idiotype antibody to anti-DNA. 3. Immunol. 132: 187-190 26. Teitelbaum, D., J. Rauch, B. D. Stollar, and R. S. Schwartz. 1984. In viva effects of antibodies against a high frequency idiotype and anti-DNA antibodies in MRL mice. 3. Immunol. 132: 1282-1285 27. Zouali, M. 1983. Idiotypic-antiidiotypic interactions in systemic lupus erythematosus: Demonstration of oscillary levels of anti-DNA autoantibodies and reciprocal antiidiotypic activity in a single patient. Ann. Immunol. (Paris) 134C: 377-391 28. Jerne, N. K. 1974. Towards a network theory of the immune system. Ann. Immunol. (Paris) 125C: 373-389