fax: (49) 40-48051221
Liver Failure and Liver Disease
Small ubiquitin-related modifiers: A novel and independent class of autoantigens in primary biliary cirrhosis†
Article first published online: 22 FEB 2005
Copyright © 2005 American Association for the Study of Liver Diseases
Volume 41, Issue 3, pages 609–616, March 2005
How to Cite
Janka, C., Selmi, C., Gershwin, M. E., Will, H. and Sternsdorf, T. (2005), Small ubiquitin-related modifiers: A novel and independent class of autoantigens in primary biliary cirrhosis. Hepatology, 41: 609–616. doi: 10.1002/hep.20619
Conflict of interest: Nothing to report.
- Issue published online: 22 FEB 2005
- Article first published online: 22 FEB 2005
- Manuscript Accepted: 23 DEC 2004
- Manuscript Received: 15 NOV 2004
- Deutsche Krebshilfe
- Deutsche Forschungsgemeinschaft
- National Institutes of Health. Grant Numbers: DK39588, DK92310
- Freie und Hansestadt Hamburg
- Bundesministerium für Gesundheit und soziale Sicherung
Serum autoantibodies against components of nuclear dots (anti-NDs), namely PML and Sp100, are specifically detected in 20% to 30% of patients with primary biliary cirrhosis (PBC). Although anti-ND antibodies are nonpathogenic, the mechanisms that lead to this unique reactivity are critical to understanding the loss of immune tolerance in PBC. Importantly, Sp100 and PML are both covalently linked to small ubiquitin-related modifiers (SUMOs). Therefore, we investigated whether SUMO proteins are independent autoantigens in PBC and studied 99 PBC sera samples for reactivity against NDs, PML, and Sp100, as well as against SUMO-2 and SUMO-1 recombinant proteins. Autoantibodies against SUMO-2 and SUMO-1 were found in 42% and 15% of anti-ND–positive PBC sera, respectively. Anti-SUMO reactivity was not observed in anti-ND–negative sera. Anti–SUMO-2 autoantibodies were found in 58% of sera containing autoantibodies against both PML and Sp100 and were detected exclusively in sera containing anti-Sp100 autoantibodies. In conclusion, SUMO proteins constitute a novel and independent class of autoantigens in PBC. Furthermore, we believe our data emphasize the post-translational modification of lysine by either lipoylation in the case of AMA or SUMOylation in the case of specific anti-ND autoantibodies as the pivotal site for autoantibody generation in PBC. (HEPATOLOGY 2005 2005;41:609–616.)
Primary biliary cirrhosis (PBC) is an autoimmune disease of unknown etiology characterized by inflammatory destruction of the intrahepatic bile ducts and the presence of serum autoantibodies to mitochondrial antigens (AMAs) detected in approximately 90% to 95% of cases.1 In contrast to AMAs, approximately 50% of patients with PBC have serum antinuclear autoantibodies (ANAs),2 with the nuclear dot (ND) pattern found in 20% to 30% of cases and considered specific for PBC.3 The pathogenic role of ANA in PBC is not clear, but understanding the process of development of these disease-specific autoantibodies appears to be important in the dissection of the mechanisms leading to autoimmunity in PBC. Autoantibodies to NDs react with large nuclear multiprotein complexes sharing several biological features,4 including Sp1005 and PML6, 7 proteins, as well as small ubiquitin-related modifiers (SUMOs).8 Sp100 and PML constitute the main antigens for anti-ND ANA in PBC, with both proteins being targeted in the majority of patients with ANA-positive PBC. Interestingly, SUMO-1 and the related SUMO-2 and SUMO-3 proteins are unique among other ND proteins because they can be covalently attached to other cellular proteins by a pathway similar to ubiquitin conjugation. Unlike ubiquitination, however, binding of SUMO to a protein does not lead to its degradation, but rather stabilizes the target protein and regulates its cellular distribution and function9 and in fact, covalent conjugation of Sp100 and PML has been demonstrated for SUMO-14 and also for SUMO-2/3.10 We hypothesized that anti-SUMO autoantibodies will be found in PBC. We report herein that SUMO-2 and, less frequently, SUMO-1 are recognized by PBC sera and that such reactivity often co-occurs with anti-Sp100 autoantibodies. We submit that SUMO-1 and SUMO-2, once covalently and stably attached to unrelated target proteins, represent novel autoantigens in PBC.
Patients and Methods
Sera from clinically documented patients with PBC (n = 99) were generously donated by Drs. M.C. Jung, U. Spengler, and G.R. Pape (Institut für Immunologie, Universität München, Germany); R. Klein and P.A. Berg (Medizinische Klinik und Poliklinik, Universität Tübingen, Germany); H.-J. Lakomek (Klinik für Rheumatologie und physikalische Medizin, Minden, Germany); M. Manns (Medizinische Hochschule, Hannover, Germany); R. Mierau and E. Genth (Rheumaforschungsinstitut der Rheumaklinik Aachen, Germany); E. J. Heathcote (University Hospital, Toronto, Canada); M. Podda (Ospedale San Paolo, University of Milan, Milan, Italy); F. Rosina (Ospedale Molinette, Turin, Italy); M. Sterneck (UKE, Hamburg, Germany); and E. Penner (Universitätskrankenhaus Wien, Austria). In all cases, the diagnosis was made based on internationally accepted criteria including compatible histology obtained at diagnosis.1 The study protocol followed the ethical guidelines of the most recent Declaration of Helsinki (Edinburgh, 2000), and all patients enrolled in the study provided written informed consent at the local institution. Our series included 37 AMA-negative and 62 AMA-positive sera, as defined by indirect immunofluorescence (IIF) analysis using rat kidney cells. This series was intentionally biased towards the larger number of AMA-negative sera so that adequate representation from both groups was available for analysis. There was no selection bias in obtaining the AMA-positive cases. Serum samples from healthy subjects were used as controls.
All sera were first tested for anti-ND autoantibodies by IIF analysis using HeLa cells treated with interferon beta.11 To confirm the ND staining, cells were double labeled using rabbit antiserum against recombinant Sp100.11 Subsequently, all human sera were tested for anti-Sp100 using a commercial ELISA kit (IMTEC, Zepernick, Germany)12; the cutoff for a positive reaction was 20 U/mL. Anti-PML autoantibodies were analyzed by IIF staining of cells overexpressing PML.11 Titers higher than 1:40 were considered positive because such titers were greater than 2 standard deviations above the mean among healthy controls.
Plasmids pSG5 and p-link-EGFP, a derivative of pSG5 containing sequences encoding for the SV40 nuclear localization signal (NLS) followed by the FLAG epitope tag and the enhanced green fluorescent protein (EGFP), were used for transient expression in eukaryotic cells and for in vitro transcription/translation. The cDNAs of SUMO-2 and SUMO-1 were cloned into the corresponding polylinker region of p-link-EGFP or pSG5. The pSG5-SUMO-1 construct contains the SUMO-1 cDNA as a 1081-bp SmaI/SspI fragment inserted blunt end into pSG5. Fragments containing the complete sequences encoding SUMO-2 were excised from plasmids pET11d-SUMO-213 using restriction enzymes NcoI and BamHI. The corresponding 321-bp fragment underwent blunt-end ligation into plasmid pSG5 linearized by BamHI digestion. For expression of SUMO-2 and SUMO-1 in Escherichia coli, the corresponding cDNAs were cloned in-frame into the prokaryotic expression plasmid pEx42 a, b, or c14 using BamHI, BglII or SmaI, and HindIII, respectively.
Expression of Recombinant MS2-Conjugated SUMO and Generation of Protein Extracts.
Expression of recombinant SUMO-1 and SUMO-2 proteins fused to MS2 was carried out as previously described.14 All bacteria were centrifuged and the pellet lysed in SDS lysis buffer including 0.5-1 U/μL benzonase (Merck, Munchen, Germany). Samples were incubated at room temperature for 5 minutes, followed by 5 minutes at 100°C. MS2-SUMO-2 expression was significantly lower than MS2-SUMO-1 expression; therefore, the concentration of fusion proteins and of total bacterial proteins were then normalized using lysate from bacteria expressing only MS2. The latter extracts were also used as negative controls.
In Vitro Translation of EGFP-SUMO-2.
Using the Promega Coupled Transcend non-radioactive Translation Detection System (Promega, Madison, WI), in vitro transcription/translation of plasmids p-link-EGFP-SUMO-2 was performed as a one-tube reaction in wheat germ extract without using biotin-labeled tRNAs.
Electrophoretic Separation of Proteins and Immunoblotting.
Protein separation on 20% SDS-PAGE gels and immunoblotting were performed,15 including 5% nonfat dry milk (in water) for blocking. Patient sera (diluted 1:2000) or monoclonal antibodies against MS2 (diluted 1:50)14 were incubated in for 1 hour at 4°C. Peroxidase-labeled goat anti-human or anti-mouse immunoglobulin G antibodies (Dianova, Hamburg, Germany) were used at 1:30000 dilution to detect primary antibodies with enhanced chemiluminescence reagents (Amersham International, Little Chalfont, United Kingdom).
LMH Cell Culture and Transient Transfection Assay.
Chicken LMH cells were maintained in Dulbecco's Modified Essential Medium/Nutrient Mix F12 (Gibco Life Technologies Ltd., Paisley, Scotland), supplemented with 10% fetal calf serum. Plasmids were introduced into cells by the calcium phosphate procedure.15
Indirect Immunofluorescence Assay on LMH Cells.
Cells were fixed in methanol at −20°C for 5 minutes, followed by acetone at −20°C for 20 seconds. Double IIF staining was performed using patient sera and a mouse monoclonal anti-MS2 antibody (Sigma, Deisenhofen, Germany), both at a 1:100 dilution in phosphate-buffered saline. Fluorescein- or rhodamin-conjugated goat anti-human and anti-mouse secondary antibodies (Dianova, Hamburg, Germany) were used at a 1:200 dilution.
The prevalence of anti-SUMO autoantibodies in different serum populations was compared using the chi-square test. All analyses were two-tailed, and P values less than 0.05 were considered statistically significant.
Table 1 illustrates the prevalence of specific serum reactivities in sera from patients with PBC. Sera were first tested by indirect immunofluorescence in parallel with anti-Sp100 antisera for colocalization and anti-ND determination to discriminate the anti-ND specificity from other speckled-type ANAs. Figure 1 illustrates the punctuate nuclear staining pattern of a typical anti-ND–positive PBC serum (Fig. 1A) and colocalization with the anti-Sp100 antibodies pattern (Fig. 1B). The anti-ND pattern was observed in 53 (54%) of 99 PBC sera. These anti-ND–positive sera were then tested for Sp100 and PML autoantibodies by enzyme-linked immunosorbent assay and immunofluorescence, respectively. All anti-ND–positive sera at IIF (53 of 53) were found either positive for anti-Sp100 (9 of 53), anti-PML (11 of 53), or both autoantibodies (33 of 53). No autoantibodies against PML or Sp100 were detected in the anti-ND–negative PBC group (n = 46) or in the control sera. When sera were analyzed based on their AMA status, 13 (35%) of 37 AMA-negative and 40 (65%) of 62 AMA-positive sera demonstrated an anti-ND ANA pattern (P = .007). Among these, 23 AMA-positive and 10 AMA-negative sera recognized both Sp100 and PML proteins.
|Anti-ND-pos||Anti-Sp100-pos Anti-PML-pos||Anti-Sp100-pos Anti-PML-neg||Anti-Sp100-neg Anti-PML-pos||Anti-ND-neg|
|Anti-SUMO-2-pos (%)||22/53 (42)*||19/33 (58)*||3/9 (33)||0/11||0/46|
|Anti-SUMO-1-pos (%)||8/53 (15)**||7/33 (21)‡||0/9||1/11 (9)||0/46|
Autoantibodies Against SUMO Proteins
Immunoblotting With MS-SUMO-2 and MS-SUMO-1.
MS2-SUMO-2 and MS2-SUMO-1 were identified as bands migrating at 27 kd and 30 kd, respectively, whereas MS2 was detected as a double band around 20 kd (Fig. 2). All proteins were recognized by the anti-MS2 monoclonal antibody (Fig. 2, blot labeled anti-MS2). PBC sera were found to react against SUMO-2 (represented by serum D20, blot A), SUMO-1 (R7, blot C), both SUMO-1 and SUMO-2 (P21, blot B), or neither protein (D8, blot D). Twenty-two (42%) of the 53 anti-ND–positive sera reacted against SUMO-2, and 8 (15%) reacted against SUMO-1. Among these, 7 sera (13%) recognized both SUMO-2 and SUMO-1. Reactivities observed against SUMO-2 were stronger than against SUMO-1, thus indicating higher anti–SUMO-2 antibody titers or higher affinity of such autoantibodies. No PBC sera reacted with the MS2 protein alone. None of the 8 normal human sera or the 46 anti-ND–negative PBC sera were positive against SUMO or MS2 proteins. Nineteen (58%) of 33 sera that were positive for Sp100 and PML reacted against SUMO-2, and 7 (21%) of 33 reacted against SUMO-1. Three (33%) of 9 anti-Sp100–positive sera reacted against SUMO-2, and none reacted against SUMO-1. No anti-PML–positive sera (n = 11) recognized SUMO-2, whereas 1 (9%) recognized SUMO-1. Reactivity against SUMO proteins did not reveal any differences in prevalence or titer in AMA-positive and –negative sera (data not shown). Table 1 summarizes the autoantibody frequencies.
Immunoblotting With EGFP-SUMO-2.
EGFP-SUMO-2 migrated as a 48-kd band and EGFP migrated at 32 kd (Fig. 3). Thirteen PBC sera (10 anti-Sp100/anti-PML–positive, 3 anti-Sp100–negative/anti-PML–positive) were tested against EGFP-SUMO-2, and 6 (46%) were found positive. Figure 3 shows a representative PBC serum (panel B) positive for EGFP-SUMO-2 fusion protein but not recognizing EGFP alone, and a representative PBC serum (panel C) negative against anti–SUMO-2.
Immunofluorescence on Chicken LMH Cells.
Sera were tested by IIF staining of chicken LMH cells expressing FLAG-tagged EGFP-SUMO-2 after transient transfection was performed (Fig. 4). Human EGFP-SUMO-2 when expressed in these cells exhibits a diffuse nuclear staining pattern with additional ND staining (Fig. 4, anti-FLAG column). We tested 11 anti-Sp100–positive PBC sera and found that 8 (73%) stained EGFP-SUMO-2 using this method. One serum previously negative against SUMO-2 by immunoblot was clearly anti–SUMO-2 positive by IIF. Another serum that was proven positive at immunoblotting demonstrated a weak staining of transfected LMH cells (data not shown).
We report herein that two ubiquitin-related proteins (SUMO-2 and SUMO-1) constitute independent autoimmune targets in a subgroup (42% and 15%, respectively) of anti-ND–positive PBC sera. Reactivity against SUMO proteins was observed most frequently among anti-Sp100–positive sera and was tested using three independent methods. These data are important because PBC-specific ANAs2 are unique in the diagnosis of AMA-negative PBC and, in some cases, for the definition of a subgroup of patients with more advanced disease.16 Two ANA IIF patterns, in particular, are specifically found in PBC and include the rim-like membranous pattern (secondary to autoantibodies reacting with gp210 or p6217) and the anti-ND pattern.3 In the latter case, detected in 20% to 30% of PBC sera, the major nuclear antigens have been identified as Sp100 and PML. Interestingly, both proteins are known targets for SUMOylation,5 where lysine residues are targeted for the covalent attachment of SUMO proteins, a process similar to ubiquitination.18 These covalent modifications can influence several cellular processes, including nuclear transport and signal transduction. Thus, based on these observations we investigated whether SUMO-2 and SUMO-1 proteins were independent autoantigens in PBC. We report herein that SUMO proteins not conjugated with Sp100 or PML are recognized by 15% to 42% of anti-ND–positive PBC sera but not by control sera or PBC sera lacking anti-ND autoantibodies. In particular, reactivity against SUMO-2 was observed in 58% of sera containing both anti-Sp100 and anti-PML autoantibodies but not in anti-PML– positive/anti-Sp100–negative sera. Interestingly, anti–SUMO-1 reactivity was observed in sera containing both anti–SUMO-2 and anti-Sp100 autoantibodies, but also in a serum containing autoantibodies against PML, but not Sp100 or SUMO-2 (Fig. 2). This latter serum suggests that anti–SUMO-1 reactivity is not a mere accompanying observation to the presence of anti–SUMO-2 autoantibodies.
The sequence homology between SUMO-2 and SUMO-1 is limited to ∼50% (49 of 94) amino acid identity. In sera with autoantibodies against both SUMO proteins, the reactivity against SUMO-1 was weaker, and we hypothesize that reactivity against SUMO-1 may be in part due to crossreactivity of autoantibodies against SUMO-2. We also note that the mature SUMO-2 and SUMO-3 proteins differ only in 5 amino acid residues, thus leading us to assume that autoantibodies against SUMO-2 might also recognize SUMO-3 and vice versa, although we have not investigated this aspect in the present work. We also submit that the sequence identity between SUMO-2 and ubiquitin is only 18%. This is interesting because a role for altered ubiquitination has been proposed in the formation of Mallory bodies in PBC19 and in the modulation of autoreactive T cells in autoimmunity.20 Finally, we also note that in our patient cohort, anti-ND positivity was observed with significantly higher frequency among AMA-positive PBC sera, thus militating against a possible masking effect of high-titer AMA.
We cannot determine at the present status whether autoantibodies against ND proteins (Sp100 and PML) precede or follow the appearance of anti-SUMO reactivity. Given the observed prevalence of serum antibodies, we suggest that the induction of anti–SUMO-2 autoantibodies more likely depends on a prior autoimmune response against Sp100. Analysis of a larger number of PBC sera, possibly in a longitudinal prospective fashion, will provide an answer to this question and will also reveal whether the anti-SUMO status might be associated with a different clinical progression of PBC. Our focus in this study was to confirm our hypothesis that anti-SUMO autoantibodies would be found in PBC. In particular, it will be interesting to analyze the occurrence of anti–SUMO-2/SUMO-1 autoantibodies in patients receiving ursodeoxycholic acid, which has been shown to have immunomodulatory effects12 and to influence ubiquitin expression.21 There are several key questions that are currently under study. First, it will be important to determine whether PBC sera react with SUMOylated forms of Sp100, PML, or other known SUMO targets. Second and based on our data, it would also be interesting to monitor differences in antibody affinity between modified and unmodified antigens. Importantly, we emphasize that SUMO targets such as PML can be detected in several splicing variants, each one possibly binding SUMO molecules up to three sites, therefore presenting a high variability on electrophoresis that only the use of very specific monoclonal antibodies (not currently available) could help define. In fact, it would be of interest to perform coinfection experiments in which cells are induced to express SUMO and either PML or Sp100. Such cell lines and the definition of the reactivity to expressed recombinant proteins by PBC sera, including those anti-SUMO–positive but anti-PML–negative, would not only address the issue of specificity but may allow dissection of the AMA response in clinical subgroups of patients with PBC.
The data on SUMO proteins are important in PBC because these are different from other autoantigens, as these ubiquitin-related proteins can be covalently bound to several other proteins. Apoptotic or necrotic processes may lead to the release of multiprotein SUMOylated complexes containing Sp100 or other proteins, their presentation to the immune system, the production of autoantibodies, and ultimately the onset of the autoimmune disease.22 The release of these complexes is supported by the observation that specific nuclear matrix proteins are released from apoptotic cells in culture,23 during viral infection,24 or in cold stored blood units.25 We also note that apoptosis in PBC shares unique characteristics26 and that the major AMA epitopes are released intact from apoptotic cells.27 The presentation of the SUMOylated autoantigens and the production of autoantibodies may follow the same mechanisms described for autoreactivity against other posttranslationally modified antigens. Previously, autoantibodies to ubiquitin were detected in sera of patients suffering from systemic lupus erythematosus and scleroderma.28, 29 However, unlike other processes such as glycosylation or citrullination, the SUMO modifier can be independently processed by antigen-presenting cells and therefore generate an autoimmune response directed against its epitopes. Interestingly, 7% of systemic lupus erythematosus sera were previously shown to contain autoantibodies directed against SUMO-1, whereas no healthy serum was positive,30 consistent with our data. Moreover, unlike ubiquitination, SUMOylation is not associated with degradation of the target protein but forms a stable conjugate,31 thus possibly implying different consequences of the immune responses to SUMO and ubiquitin. Several cell proteins have been shown as targets of SUMOylation,18 similar to PML and Sp100. Such proteins also include nuclear pore-associated components,32 transcription factors,33, 34 protein kinases,35 tumor suppressor genes,36 androgen receptor,34 and possibly also cytokine receptors.37 Based on our findings in ND proteins, the conjugated forms of all these proteins may share the potential of becoming autoimmune targets in PBC. We are aware, however, that additional unknown mechanisms need to be involved in this process, since our data as well as observations on indirect immunofluorescence for autoantibody testing do not show cross-reactivity of sera with other SUMOylated proteins that are expected to be found in such cells. More specifically, when lower dilutions of anti-SUMO–positive PBC sera are used, a diffuse nuclear staining is observed, with the anti-ND pattern being prominent (data not shown). The resulting theory is that a functional or spatial connection between SUMO and SUMOylated proteins is necessary for the spreading of reactivities. Moreover, we could have expected some degree of cross reactivity between anti-ND autoantibodies of PBC sera and ND antigens in chicken cells; interestingly, however, the limited similarity of Sp100 and PML across species militates against these observations. Further, SUMO modification has been demonstrated to modulate IgM gene expression by targeting nuclear matrix attachment regions.38 As the majority of patients with PBC present elevated serum levels of nonspecific IgM, these findings suggest additional implications for anti-SUMO reactivity in PBC, possibly in combination with an aberrant response of B cells to bacterial motifs.39
Our proposed mechanism also raises the possibility that the initial breakdown of tolerance in PBC might take place anywhere, and not necessarily in the site of the autoimmune injury, the small and medium-size bile ducts. Interestingly, other SUMOylated proteins such as p53 and topoisomerase 1 have been previously described as autoantigens.40, 41 More specifically, we also note that the nuclear pore complex containing the autoantigens leading to the rim-like ANA pattern specific for PBC16 includes known SUMOylated proteins such as RanGAP-1 and RanBP2.32 Interestingly, a recent study demonstrated that a significant amount of SUMO conjugates can be found in mitochondria.42 Although SUMOylation of the mitochondrial complexes that constitute the AMA autoantigens has not been reported, the presence of lysine residues on the E2 complexes makes this a possibility. Furthermore, the lysine residue that represents the target for SUMOylation is the same amino acid required for the lipoylation which was found to be crucial for the AMA recognition.43 Finally, we also note that the immunogenicity of mitochondrial antigens in AMA-positive PBC sera can be affected by the derivitization of the lysine residue as the recognition of lipoylated epitopes by autoantibodies is enhanced compared with octanoylated PDC-E2.44 We then believe that further study of the molecular recognition sites of the autoepitopes will be critical in defining the disregulated immune response in PBC. We further suggest that the lysine residue, by virtue of its potential to be lipoylated (in the case of AMA) or SUMOylated (in the case of anti-ND autoantibodies) is the pivotal site. We cannot, however, extend our hypothesis to all PBC-specific ANAs since gp210 or p62, the major autoantigens in the rim-like pattern,2 are not known to be SUMOylated. Although AMAs are the predominant autoantibodies studied in PBC, the data herein emphasize that dissection of the ANA response may be equally important in understanding the mechanisms of tolerance breakdown.
In conclusion, the implications of our findings for the pathogenesis of PBC can be summarized in a two-step hypothesis. First, anti–SUMO-2 autoantibodies are always detected in the presence of anti-Sp100 reactivity, but not vice versa, thus suggesting that the autoimmune response may first target Sp100 and then contiguously spread to the covalently bound SUMO-2. Second, once anti-SUMO autoantibodies have been produced, the autoimmune response could also recognize other SUMOylated proteins. Based on the prevalence of anti-SUMO autoantibodies in anti-Sp100/anti-PML positive sera, this could happen between Sp100 and PML in vivo. The small covalent modifier-mediated spreading hypothesized herein differs from classical epitope spreading mechanisms,45 as it involves covalent attachment of SUMO to otherwise unrelated proteins with limited or no homology. Accordingly, a longitudinal study should investigate the prevalence of these reactivities in prospectively collected PBC sera to determine whether anti-ND–positive patients who did not react with SUMO proteins will eventually become seropositive for SUMOylated proteins.
The authors are grateful to Kirsten Jensen, Anya E. Huggins, and Dr. Gritta Janka for the critical reading of the manuscript and for fruitful discussions.
- 15Molecular cloning: a laboratory manual. 2nd ed. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press, 1989., , .
- 25Nuclear matrix protein is released from apoptotic white cells during cold (1-6 degrees C) storage of concentrated red cell units and might induce antibody response in multiply transfused patients. Transfusion 2000; 40: 169–177., , , , , , et al.