Autoantibodies against mitochondrial glycerophosphate dehydrogenase in patients with IDDM

Autoantibodies against mitochondrial glycerophosphate dehydrogenase in patients with IDDM

Diabetes Research and Clinical Practice 38 (1997) 115 – 121 Autoantibodies against mitochondrial glycerophosphate dehydrogenase in patients with IDDM...

94KB Sizes 2 Downloads 95 Views

Diabetes Research and Clinical Practice 38 (1997) 115 – 121

Autoantibodies against mitochondrial glycerophosphate dehydrogenase in patients with IDDM Marta E. Fabregat a, Rosa Gasa a, Camino Rodriguez a, Anna Novials a, Teresa Gallart a, Willy J. Malaisse b,*, Ramon Gomis a b

a Endocrinology Unit, Hospital Clinic, Barcelona, Spain Laboratory of Experimental Medicine, Brussels Free Uni6ersity, 808 Route de Lennik, B-1070 Brussels, Belgium

Received 24 April 1997; received in revised form 15 August 1997; accepted 4 September 1997

Abstract The mitochondrial enzyme FAD-linked glycerophosphate dehydrogenase (mGDH) plays a key role in the recognition of D-glucose as a stimulus for insulin release from the pancreatic islet B-cell. This study reveals that autoantibodies against this enzyme are not uncommonly found in patients with insulin-dependent diabetes mellitus (IDDM) examined at the onset of the disease. Antibodies reacting with a recombinant mGDH fragment product were observed in the serum of four out of 15 type-1 diabetics, but in none of 15 control subjects. The serum of patients positive for the recombinant mGDH fragment also recognized native mGDH in a rat testis extract, provided that the enzymatic protein was first exposed to an anti-mGDH rabbit serum. Antibodies against mGDH were also found in four out 12 patients with autoimmune thyroiditis. These findings reveal that a mitochondrial enzyme, that represents an essential component of the islet B-cell glucose-sensing device, may act as an antigenic determinant in patients with IDDM or other autoimmune diseases. © 1997 Elsevier Science Ireland Ltd. Keywords: Autoantibodies; Mitochondrial glycerophosphate dehydrogenase; IDDM

1. Introduction Insulin-dependent diabetes mellitus (IDDM) is characterized by organ-specific autoimmunity. Antibodies reactive with pancreatic islets of Langerhans are indeed associated with IDDM * Corresponding author.

[1,2] as well as with prediabetes [3,4]. A standard test for the detection of islet cell antibodies (ICA) is based on their binding to pancreatic cryostat sections. This assay suffers, however, from the potential of antisera to bind to non-specific islet cell components. Recently, considerable progress has been made in identifying individual islet cell antigens involved in the autoimmune process.

0168-8227/97/$17.00 © 1997 Elsevier Science Ireland Ltd. All rights reserved. PII S 0 1 6 8 - 8 2 2 7 ( 9 7 ) 0 0 0 9 5 - 8

b

30 14 24 25 17 42 15 20 20 27 20 19 14 38 21

Age (years)

65 42 99 51 52 50 53 72 50 79 54 81 67 — 53

Weight (kg)

186 172 185 161 159 161 152 180 158 153 162 181 174 — 175

Height (cm)

18.6 14.2 28.9 19.7 20.6 19.1 23.0 22.2 20.0 33.7 20.7 24.7 22.0 — 17.2

Body mass index

8 8 3 6 8 16 3 10 2 8 6 2 6 5 8

Diabetes duration (week)

26 37 42 15 46 34 40 34 38 45 27 30 42 44 20

Treatment (Insulin, U)

C-peptide concentration (ng/ml) before and 10 min after the intravenous administration of 1.0 mg glucagon. The figure in parentheses refers to the type of diabetes (type 1 or 2).

M F M F F F F M M F M M M M M

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

a

Sex

Subject (no.)

Table 1 Clinical features of the IDDM patients IAA

− + − + + + − + + + + − + − +

Familial history

Father (1)b — — Brother (1) — Father (1) and Mother (2) Father (2) — Brother (1) — — — Maternal relative (1) — —

− − − + + + + − + − + − + + +

GAD antibodies

+ − + + − − − + − − − − − − −

mGDH antibodies

0.4 0.2 2.2 0.7 0.6 0.4 0.1 0.2 0.4 0.8 1.1 0.2 1.1 0.5 0.1

min 0

0.5 0.5 3.2 0.7 1.0 0.5 0.1 0.5 0.5 0.7 1.6 0.4 1.6 0.8 0.1

min 10

Glucagon testa

116 M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121

M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121

117

Table 2 Clinical features of patients with autoimmune thyroiditis Subject (no.)

Sex

Age (years)

TPO Aba (U/ml)

TG Abb (U/ml)

TSHR Abc (%)

Thyroid disease

mGDH antibodies

1 2 3 4 5 6 7 8 9 10 11 12

F F F F F F F F M F F F

34 36 61 45 39 26 31 69 18 63 27 47

414 \600 \600 \600 432 \600 190 \600 203 53 \600 N.D.

1256 134 247 2052 21 376 256 2002 3474 7 3681 N.D.

N.D.d 12 N.D. 5 23 61 N.D. 110 12 N.D. N.D. 68

Hashimoto Hyperthyroidism Hyperthyroidism N.D. Graves Hypothyroidism Hypothyroidism Hypothyroidism Hypothyroidism Hypothyroidism Hypothyroidism Hyperthyroidism

− − + − + + − + − − − −

a

Thyroperoxidase antibodies (normal range518 U/ml). Thyroglobulin antibodies (normal range 528 U/ml). c TSH receptor antibodies (normal range59%). d N.D., not determined. b

Thus, insulin [5,6], the 64-kDa glutamic acid decarboxylase [7–9], the glycolipid GT3 [10], a 52kDa antigen [11], carboxypeptidase H [12,13], the glucose-transporter GLUT-2 [14], an IA-2 antigen related to protein tyrosine phosphatase [15], as well as autoantigens that were cloned by expression screening of an islet cDNA library [16], were all proposed as islet cell antigens in patients with IDDM. This study reveals that FAD-linked glycerophosphate dehydrogenase (mGDH), the key enzyme of the glycerol phosphate shuttle, may also act as an antigenic determinant in patients with IDDM examined at the onset of the disease. This enzyme differs from most antigens so far identified by the facts that it represents a mitochondrial protein and that an inherited impairment of its catalytic activity may be a cause of islet B-cell dysfunction in NIDDM [17 – 19].

2. Materials and methods Fifteen insulin-dependent diabetic patients were selected on the basis of both a recent onset of the disease and the presence of ICA. The characteristics of these subjects are given in Table 1. In these

subjects, the intravenous administration of glucagon (1.0 mg) only caused, within 10 min, a modest increase in C-peptide plasma concentration, averaging (mean9 S.D.) no more than 0.259 0.28 ng/ml (paired difference) above basal value (0.609 0.55 ng/ml; n= 15 in both cases). These patients were examined together with 15 control subjects, ten females and five males, 24– 43 years old (mean 9 S.D.: 329 6 years), with a body mass index of 22.7 9 2.3, all negative for ICA, insulin auto-antibodies (IAA) or glutamate decarboxylase (GAD) antibodies. Twelve patients with autoimmune thyroiditis were also included in this study (Table 2). Subjects 3 and 11 in Table 2 were also suffering from IDDM. For the detection of auto-antibodies against mGDH by immunoblotting, 250 ng of either recombinant mGDH-GST or GST (glutathione-Stransferase) protein, prepared as described elsewhere [20], were resolved by electrophoresis on a 8% SDS-polyacrylamide gel and transferred onto nitrocellulose membranes (Bio-Rad, Milan, Italy). The membranes were blocked in Tris– buffered saline (pH 7.4, 20 mM Tris–HCl, 150 mM NaCl) containing 5% non-fat dry milk and 0.2% Nonidet P-40 for 20 min at 20°C. The

118

M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121

membranes were then incubated with the serum from either diabetic patients or control subjects (1/1000 dilution). Detection of bound antibodies was performed using peroxidase-conjugated antihuman IgG (Sigma, St Louis, MO) and the ECL chemiluminescence system (Amersham, Buckinghamshire, UK). In order to investigate whether the antibodies in the serum of IDDM patients reacting with recombinant mGDH also recognize native mGDH, rat testis were homogenized at 4°C in 4 volumes of lysis buffer (Tris–HCl 50 mM, pH 8) containing 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.02% sodium azide and protease inhibitors (1 mg/ml aprotinin, 100 mg/ml phenylmethylsulfonyl fluoride, and 1 mM EDTA). Protein concentration was measured by Lowry method (Bio-Rad). Aliquots (100 mg) of rat testis protein or recombinant mGDH (250 ng) were resolved on a 8% SDS-PAGE gel followed by electroblotting onto a cellulose membrane. The membranes were blocked (see above) and then incubated with either human sera or anti-mGDH rabbit IgG antibodies (5 mg/ml), that were raised against recombinant mGDH and allow immunodetection of native mGDH in tissue extracts [20]. Peroxidase-conjugated goat anti-rabbit antibodies (Sigma) were used to detect bound rabbit antibodies. Such antibodies were also used to enhance [21–24] the binding of human antibodies to blotted recombinant and native mGDH. For this purpose, blotted membranes were preincubated overnight at 4°C with an excess (500 mg/ml) of either anti-mGDH rabbit IgG antibodies or IgG from nonimmunized rabbits. Human sera (10 mg IgG/ml) was then added, and the membrane incubated for 18 additional h at 4°C. After washing, bound human antibodies were detected by peroxidase-conjugated anti-human IgG rabbit antibodies, as indicated above. In all immunoblots, human sera and antibodies were diluted in the blocking solution. The recognition of native mGDH by the sera of IDDM patients was also explored by immunoprecipitation. For this purpose, aliquots (containing each 1.5 mg protein) of the post-nuclear supernatant (15 min, 600 g) from a rat testis homogenate were first incubated for 4 h at 4°C with human

serum negative to recombinant mGDH and exposed for 1 h to protein A-sepharose beads (Pharmacia, Uppsala, Sweden) to decrease non-specific binding. The precleared material was then incubated with 160 ml of the patient sera for 12 h at 4°C and again exposed for 1 h to the protein A-sepharose beads. After repeated washing of the bead pellets with a Tris–HCl buffer (50 mM, pH 8) containing 150 mM NaCl, 0.1% Nonidet P-40 and 100 mg/ml phenylmethylsulfonyl fluoride, the immune complexes were eluded by boiling in loading buffer (50 mM Tris–HCl, pH 6.8, containing 100 mM dithiothreitol; 2% SDS, 0.1% bromophenol blue and 10% glycerol), separated on a 8% SDS-polyacrylamide gel, electroblotted onto a cellulose membrane, and probed overnight at 4°C with rabbit anti-mGDH serum at 5 mg IgG/ml. Visualization was eventually achieved as described above. In parallel experiments, a comparable procedure was conducted in the presence of 25 ml non-immune rabbit serum to decrease non-specific binding, and then 25 ml rabbit antimGDH serum to provoke immunoprecipitation.

3. Results Four out of 15 diabetic patients were found positive for antibodies to recombinant mGDHGST (Table 1). The serum from these four subjects failed, however, to recognize GST (Fig. 1). No antibody to either mGDH-GST or GST was detected in the serum from 15 control subjects. It was then investigated whether the antibodies found in some diabetic subjects also recognize native mGDH. A rabbit antiserum raised against recombinant mGDH-GST detected a protein of approximately 72 kDa in rat testis (Fig. 2, blot A), which corresponds to the molecular weight of the native mGDH protein. The serum from a diabetic patient, which was positive for mGDHGST, apparently failed, however, to recognize the same protein in rat testis (Fig. 2, blot B). Negative results were also obtained with the sera of other diabetic patients positive for mGDH-GST, even when the concentration of IgG was raised from 10 to 50 mg/ml.

M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121

Fig. 1. Representative detection of autoantibodies reacting with recombinant mGDH-GST in sera from a control subject (left) and an IDDM patient (right). Equal amounts (250 ng) of recombinant mGDH-GST (lane 1) or GST (lane 2), with molecular weight of 65 and 26 kDa respectively, were resolved by electrophoresis on a 8% SDS-polyacrylamide gel prior to incubation with the human sera.

Since the binding of antibodies to blotted proteins may provoke slight conformational changes and unmask protein epitopes [21 – 24], the blots were preincubated with the anti-mGDH rabbit serum before probing with the human serum.

Fig. 2. Recognition of native mGDH protein in rat testis by the serum from an IDDM patient positive towards recombinant mGDH-GST. Immunoblots of either the recombinant protein (r-mGDH) or a rat testis extract were probed with (A) rabbit anti-mGDH serum (5 mg IgG/ml), (B) anti-r-mGDH IDDM serum (10 mg IgG/ml), (C) rabbit anti-mGDH serum (500 mg IgG/ml) at 4°C for 18 h followed by the IDDM serum (10 mg IgG/ml), and (D) a non-immune rabbit serum (500 mg IgG/ml) followed by the IDDM serum. The antigen–antibodies complexes were visualized by peroxidase-conjugated antirabbit IgG (blot A) or anti-human IgG (blots B to D).

119

Under these conditions, the serum of diabetic patients reactive with mGDH-GST now clearly revealed in rat testis a band of the same molecular weight as that detected by anti-mGDH rabbit IgG (Fig. 2, blot C). Moreover, the binding of antibodies from diabetic patients to recombinant mGDH was also strongly enhanced. Such was not the case when the preincubation was performed with a non-immune rabbit serum (Fig. 2, blot D). It should be underlined that, in the experiments illustrated in blots B to D of Fig. 2, anti-human, rather than anti-rabbit, horse radish peroxidaseconjugated IgG were used to detect the antibodies bound to native mGDH. Under comparable experimental conditions, the sera of IDDM patients, whether exhibiting or not antibody activity against mGDH-GST, did not show any binding to either IgG from non-immunized rabbits dotblotted onto nitrocellulose membranes in amount ranging from 10 to 5000 ng or rabbit IgG antibodies specific for human IgM chains and bound to dot-blotted human IgM (up to 500 ng/dot, i.e. more than twice the amount of recombinant mGDH-GST used in Western blots) lacking rheumatoid factor activity. At variance with sera from control subjects, but alike rabbit anti-mGDH serum, the sera of patients positive for mGDH-GST also provoked the immunoprecipitation of native mGDH in rat testis homogenates (data not shown). Three out of the four diabetic subjects that were positive for mGDH failed to display GAD antibodies, and two of them were also negative for IAA. Four out of 12 subjects with autoimmune thyroiditis were also found positive for mGDH antibodies (Fig. 3).

4. Discussion This study reveals that patients with IDDM, examined at the onset of the disease, may display autoantibodies against mGDH, the key enzyme of the glycerol phosphate shuttle. This enzyme plays an essential role in the glucose-sensing device of the pancreatic B-cell [25]. Indeed, in glucose-stimulated islets, its activation by cytosolic Ca2 + ac-

120

M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121

autoimmune aggression of insulin-producing cells in this disease involves the production of antibodies not solely against membrane-associated or cytosolic proteins but also towards proteins that are normally located in intracellular organelles, such as either insulin stored in the B-cell secretory granules or mGDH inserted in the inner membrane of mitochondria.

Acknowledgements

Fig. 3. Detection of autoantibodies reacting with recombinant mGDH-GST in sera from 12 patients with autoimmune thyroiditis. Same presentation as in Fig. 1. The patients in whom antibodies were identified correspond to subjects 3, 5, 6 and 8 in Table 2.

counts for a preferential stimulation of oxidative relative to total glycolysis and, hence, optimalizes the yield of ATP resulting from the catabolism of the hexose in the insulin-producing cells [26,27]. Antibodies against mGDH were also found in patients with autoimmune thyroiditis, indicating that their presence is not restricted to IDDM. More work is obviously required to assess whether the identification and quantification of mGDH autoantibodies in IDDM offers any advantage over other antibodies for prediction of the disease and characterization of the autoimmune process eventually leading to the destruction of insulin-producing cells. Meanwhile, the present findings provide, to our knowledge, the second indication that a mitochondrial protein may act as an antigenic determinant in IDDM. Arden et al. [28] recently reported on a novel 38-kD islet mitochondrial autoantigen (imogen 38) recognized by T cells from a newly diagnosed type 1 diabetic patient. They concluded that imogen 38 may be a target for bystander autoimmune attack in diabetes rather than a primary autoantigen. A comparable situation may be operative in NIDDM patients with antibodies against mGDH. Thus, it is now obvious that the

This work was supported by the Islet Research European Network Concerted Action of the Commission of the European Communities, the Comissionat per la Universitat i Recerca de la Generalitat de Catalunyia, and a grant from the Fondo de Investigaciones Sanitarias (FIS 96/ 0164). We are grateful to C. Benito for GAD antibodies determination, R. Casamitjana for Cpeptide measurements, and C. Demesmaeker for secretarial help.

References [1] G.F. Bottazzo, A. Florin-Christensen, S. Doniach, Islet cell antibodies in diabetes mellitus with polyendocrine deficiencies, Lancet 2 (1974) 1279 – 1283. [2] A.C. Mac Cuish, J. Jordan, C.J. Campbell, J.P. Duncan, W.J. Irvine, Antibodies to pancreatic islet cells in insulindependent diabetes with coexistent autoimmune disease, Lancet 2 (1974) 1529 – 1533. [3] W.J. Irvine, R.S. Gray, C.J. McCallum, Pancreatic isletcell antibody as a marker for asymptomatic and latent diabetes and prediabetes, Lancet 2 (1976) 1097 – 1102. [4] A.N. Gorsuch, K.M. Spencer, J. Lister, J.M. McNally, B.M. Dean, G.F. Bottazzo, A.G. Cudworth, Evidence for a long prediabetic period in type I (insulin-dependent) diabetes mellitus, Lancet 2 (1981) 1363 – 1365. [5] J.P. Palmer, C.M. Asplin, P. Clemons, K. Lyon, O. Iatpate, P. Raghu, T.L. Paquette, Insulin antibodies in insulin-dependent diabetes before insulin treatment, Science 222 (1983) 1337 – 1339. [6] T. Wilkin, M. Armitage, C. Casey, D.A. Pyke, P.J. Hoskins, M. Rodier, J. Diaz, R.D.G. Leslie, Value of insulin autoantibodies as serum markers for insulin-dependent diabetes mellitus, Lancet 1 (1985) 480 – 481. [7] S. Baekkeskov, J.H. Nielsen, B. Marner, T. Bilde, J. Ludvigsson, A. Lernmark, Autoantibodies in newly diagnosed diabetic children immunoprecipitate specific human islet cell proteins, Nature 298 (1982) 167 – 169.

M.E. Fabregat et al. / Diabetes Research and Clinical Practice 38 (1997) 115–121 [8] S. Baekkeskov, M. Landin, J.K. Kristensen, S. Srikanta, G.J. Bruining, T. Mandrup-Poulsen, C. de Beaufort, J.S. Soeldner, G. Eisenbarth, F. Lindgren, G. Sundquist, A. Lernmark, Antibodies to a 64 000 Mr-human islet cell antigen precede the clinical onset of insulin-dependent diabetes, J. Clin. Invest. 79 (1987) 926–934. [9] S. Baekkeskov, H.-J. Aanstoot, S. Christgau, A. Reetz, M. Solimena, M. Cascalho, F. Folli, H. Richter-Olesen, P. De Camili, Identification of the 64 K autoantigen 1 in insulin-dependent diabetes as the GABA-synthesizing enzyme glutamic acid decarboxylase, Nature 347 (1990) 151 – 156. [10] B.K. Gillard, J.W. Thomas, L.J. Nell, D.M. Marcus, Antibodies against ganglioside GT3 in the sera of patients with type I diabetes mellitus, J. Immunol. 142 (1989) 3826 – 3832. [11] D.G. Karounos, J.W. Thomas, Recognition of common islet antigen by autoantibodies from NOD mice and humans with IDDM, Diabetes 39 (1990) 1085–1090. [12] A.C. Powers, S. Bowen, S. West, Carboxypeptidase H is an autoantigen of the ICA and is expressed on the cell surface of islet cells, Diabetes 40 (1991) (1A) (Abstract). [13] M. Pietropaolo, L. Castano, E. Russo, A. Powers, O. Barnea, G.S. Eisenbarth, Utilization of a human lgt 11 islet library to identify novel autoantigens associated with type I diabetes, Diabetes 40 (1991) (1A) (Abstract). [14] L.R. Inman, C.T. McAllister, L. Chen, S. Hugues, C.B. Newgard, J.R. Kettman, R.H. Unger, J.H. Johnson, Autoantibodies to the GLUT-2 glucose transporter of b-cells in insulin-dependent diabetes mellitus of recent onset, Proc. Natl. Acad. Sci. USA 90 (1993) 1281–1284. [15] M.A. Payton, C.J. Hawkes, H.R. Christie, Relationship of the 37 000- and 40 000-Mr tryptic fragments of islet antigens in insulin-dependent diabetes to the protein tyrosine phosphatase-like molecule IA-2 (ICA 512), J. Clin. Invest. 96 (1995) 1506–1511. [16] D.U. Rabin, S. Pleasic, R. Palmer-Crocker, J. Shapiro, Cloning and expression of IDDM-specific human autoantigens, Diabetes 41 (1992) 183–186. [17] W.J. Malaisse, Is type 2 diabetes due to a deficiency of FAD-linked glycerophosphate dehydrogenase in pancreatic islets?, Acta Diabetol. 30 (1993) 1–5. [18] W.J. Malaisse, Pancreatic B-cell FAD-glycerophosphate dehydrogenase deficiency in non-insulin-dependent diabetes: an updated review, in: P.R. Flatt, S. Lenzen (Eds.),

.

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

121

Insulin Secretion and Pancreatic B-Cell Research, Smith – Gordon, London, 1994, pp. 445 – 452. J. Fernandez-Alvarez, I. Conget, J. Rasschaert, A. Sener, R. Gomis, W.J. Malaisse, Enzymatic, metabolic and secretory patterns in human islets of type 2 (non-insulindependent) diabetic patients, Diabetologia 37 (1994) 177 – 181. A. Novials, M.E. Fabregat, C. Benito, J. Fernandez-Alvarez, T. Gallart, W.J. Malaisse, R. Gomis, Immunodetection of mitochondrial glycerophosphate dehydrogenase (mGDH) by a polyclonal antibody raised against a recombinant mGDH fragment product, Biochem. Mol. Med. 59 (1996) 187 – 191. R. Tosi, N. Tanigaki, R. Sorrentino, R. Accolla, G. Corte, Binding of one monoclonal antibody in human Ia molecules can be enhanced by a second monoclonal antibody, Eur. J. Immunol. 11 (1981) 721 – 726. H. Lemke, G.J. Ha¨mmerling, Clustering of antigenic determinants on H-2 molecules, J. Immunol. 128 (1982) 2465 – 2469. F.X. Heinz, C. Mandl, R. Berger, W. Tuma, C. Kunz, Antibody-induced conformational changes result in enhanced activity of antibodies to different antigenic sites in the tick-borne encephalitis virus glycoprotein, Virology 133 (1984) 25 – 34. B.S. Parekh, M.J. Buchmeier, Proteins of lymphocytic choriomeningitis virus: antigenic topography of the viral glycoproteins, Virology 153 (1986) 168 – 178. W.J. Malaisse, Glucose-sensing by the pancreatic B-cell: the mitochondrial part, Int. J. Biochem. 24 (1992) 693 – 701. J. Rasschaert, W.J. Malaisse, Hexose metabolism in pancreatic islets. Glucose-induced and Ca2 + -dependent activation of FAD-glycerophosphate dehydrogenase, Biochem. J. 278 (1991) 335 – 340. A. Sener, W.J. Malaisse, Hexose metabolism in pancreatic islets. Ca2 + -dependent activation of the glycerol phosphate shuttle by nutrient secretagogues, J. Biol. Chem. 267 (1992) 13251 – 13256. S.D. Arden, B.O. Roep, P.I. Neophytou, E.F. Usac, G. Duinkerken, R.R.P. de Vries, J.C. Hutton, Imogen 38: a novel 38-kD islet mitochondrial autoantigen recognized by T cells from a newly diagnosed type 1 diabetic patient, J. Clin. Invest. 97 (1996) 551 – 561.