|
|
| |||||||||||||||||||||||||||||||||
Background Graft-versus-host disease (GVHD) caused by poorly defined minor (i.e., other than HLA) histocompatibility antigens remains a serious problem in recipients of bone marrow transplants. We sought to determine whether the CD31 adhesion molecule is a minor alloantigen.
Methods We directly sequenced samples of complementary DNA (cDNA) encoding CD31 molecules from 21 unrelated normal subjects. Sequence-specific primers were then designed to amplify alleles by the polymerase chain reaction, thereby permitting CD31 typing of genomic DNA from additional normal subjects. To assess the relevance of CD31 matching to bone marrow transplantation, we performed CD31 typing of 46 recipients of bone marrow (32 without GVHD and 14 with severe [grade III or IV] acute GVHD) and their HLA-identical sibling donors. The immunoreactivity of CD31 phenotypes with anti-CD31 monoclonal antibodies was compared by flow cytometry.
Results Direct sequencing of cDNA for CD31 from the 21 normal subjects identified a single polymorphism, CTG
Conclusions The adhesion molecule CD31 is polymorphic. When donor and recipient genotypes are not identical, the risk of GVHD increases. Prospective CD31 typing may reduce the risk of acute GVHD.
GTG (LeuVal), at codon 125; we designated the resulting alleles CD31.L and CD31.V, respectively. The CD31 genotypes of these and 142 other unrelated subjects were of the expected frequencies. Among the transplant recipients, 71 percent of those with acute GVHD had CD31 genotypes that were not identical to the donor's genotype, as compared with 22 percent of the recipients without GVHD (P = 0.004). The binding of anti-CD31 monoclonal antibodies as measured by fluorescence-activated cell sorting correlated with the CD31 types of homozygous cell lines.
We postulated that a likely site for minor-locus alloantigens would be the cell surface of vascular endothelial cells, a point of first contact between transplanted tissue and its host. However, alloantigens on vascular endothelial cells have not been well characterized by biochemical techniques or the use of alloantibodies.12-19 We therefore sought polymorphisms in genes encoding molecules on the surface of such cells, as a way of identifying new alloantigenic systems. The molecule selected for study was the platelet–endothelial-cell adhesion molecule 1 (PECAM-1, or CD31), a 120-to-130-kd membrane glycoprotein that is constitutively expressed on vascular endothelial cells, bone marrow stem cells, platelets, and most circulating leukocytes.20,21 A member of the immunoglobulin superfamily, the CD31 molecule has eight protein domains — six extracellular, one transmembrane, and one cytoplasmic. Four CD31 sequences, obtained from independently cloned samples of complementary DNA (cDNA), have been published.20,22-24 They show eight nonrecurring discrepancies in amino acids scattered throughout the extracellular domains, suggesting polymorphism of CD31. To test this possibility, we studied the two parts of the extracellular domains of CD31 that have the most numerous reported discrepancies in the sequence (codons 97 through 158 and 347 through 416). We found a polymorphism of CD31 that correlated with the results of phenotyping using anti-CD31 monoclonal antibodies and with the occurrence of acute GVHD in patients who had received bone marrow transplants.
Methods
Molecular Biologic Techniques
CD31 polymorphism was first identified by direct automated sequencing of the targeted regions of cDNA from randomly selected, unrelated subjects. This method avoids artifacts of cloning and permits the identification of heterozygous bases.25 We first extracted total RNA using Trizol (GIBCO BRL, Gaithersburg, Md.) from cultured umbilical-vein endothelial cells from 12 normal infants and peripheral-blood mononuclear cells from 9 randomly selected normal blood donors. These RNA samples were reverse-transcribed, and the cDNA was amplified by the polymerase chain reaction (PCR) with Gene Amp RNA PCR kits (Perkin-Elmer Cetus, Branchburg, N.J.) and primer pairs 1, 2, and 3 (Table 1), designed to amplify the regions of interest in the CD31 molecule. One member of each pair of primers was biotinylated. The other member of the pair was used as the sequencing primer and was fluoresceinated directly, or the sequencing reaction was labeled with fluorescein-labeled deoxyadenosine triphosphate. Using as template the single-strand biotinylated PCR product, which was immobilized on streptavidin-coated magnetic beads (Dynal, Oslo, Norway), we sequenced cDNA directly. The reactions were performed with T7 sequencing kits (Pharmacia, Uppsala, Sweden), and the results were interpreted with an Automated Laser Fluorescent DNA sequencer (Pharmacia). The positions of the codons in the cDNA are cited according to the nucleotide-sequence data base of the European Molecular Biology Laboratory (accession number, M 28526), as submitted by Newman et al.,20 and the positions of the introns are given as in the sequence described by Kirschbaum et al.26 The codons we expected to be polymorphic corresponded to discrepancies identified by comparing published sequences of CD31.20,22-24,26
|
Fluorescence-Activated Cell Sorting
We used three murine antihuman CD31 monoclonal antibodies: L133 (Becton Dickinson, Mountain View, Calif.), 7E8,27 and EF70 (provided by E. Engleman, Stanford University). Umbilical-vein endothelial cells were incubated at 4°C for 30 minutes with each antibody individually, washed with phosphate-buffered saline, and then incubated at 4°C for 30 minutes with fluoresceinated goat antimouse immunoglobulin (Tago, Burlingame, Calif.) before analysis by fluorescence-activated cell sorting with a flow cytometer (Epics, Coulter, Hialeah, Fla.). Values for the mean channel number associated with each peak were calculated after conversion from a log scale to a linear scale and subtraction of the background values. Differences of fewer than 30 channels were considered to be not significant.
Selection of Patients
All protocols were approved by the Institutional Review Board at Stanford University. Informed consent was obtained from each patient (and each normal subject or parent of a normal infant). Each patient had received bone marrow from an HLA-identical sibling. GVHD was diagnosed and graded as a consensus of the opinions of the attending transplant physicians, according to previously published, standard clinical criteria,28,29 summarized as follows. The hallmarks of grade III acute GVHD are a maculopapular eruption involving 25 to 50 percent of the body-surface area or generalized erythroderma, moderately elevated serum bilirubin concentrations (3.0 to 14.9 mg per deciliter [51 to 255 µmol per liter]), moderate diarrhea (more than 60 ml of stool per kilogram of body weight per day or more than 1500 ml per day), and marked general clinical impairment. The most important criteria for grade IV acute GVHD are generalized erythroderma with the formation of bullae and often with desquamation, very high serum bilirubin concentrations (>15 mg per deciliter [256 µmol per liter]), severe diarrhea (more than 90 ml of stool per kilogram per day or more than 2000 ml per day, with abdominal pain), and extreme clinical impairment. To minimize uncertainty in the grading of acute GVHD, we compared only the two most clear-cut diagnostic groups: patients with no GVHD for at least 100 days after transplantation and patients with severe (i.e., grade III or IV) acute GVHD. This design excluded patients with chronic GVHD or mild acute GVHD (groups that together constitute approximately half of our patients undergoing bone marrow transplantation), but it also eliminated diagnostic ambiguity and avoided the problem of determining whether chronic and acute GVHD are different diseases.
DNA was selected for CD31 testing solely on the basis of the availability of samples obtained before transplantation that had been kept frozen in the clinical laboratory after routine pretransplantation testing for HLA loci and DNA microsatellites. There was DNA from 32 patients without GVHD and 14 patients with acute GVHD, a yield that approximated the ratio of patients in those two groups at our institution2 (and unpublished data). The personnel performing the CD31 typing were kept unaware of the classification of each patient's GVHD.
We compared the CD31 alleles in the recipients of bone marrow with the alleles in their HLA-matched sibling donors and determined whether the CD31 genotypes were identical, compatible (i.e., the recipient had no CD31 alleles that were foreign to the donor), or mismatched (i.e., the recipient had a CD31 allele that was foreign to the donor) for GVHD.
Statistical Analysis
The goodness of fit to the Hardy–Weinberg equilibrium30 and the statistical significance of differences between groups were calculated by Fisher's exact test (two-tailed) for all two-by-two tables, by the rank-sum test (two-tailed) for differences in ages, and by the chi-square test for all other tables.31 Odds ratios and 95 percent confidence intervals were computed by standard methods.32
Results
Molecular Polymorphism
The sequencing of cDNA from cultured umbilical-vein endothelial cells from 12 infants and peripheral-blood mononuclear cells from 9 blood donors revealed a polymorphism in the targeted regions of the CD31 gene and distinguished homozygotes bearing different genotypes from each other and from heterozygotes (Figure 1). A frequent CTGGTG substitution was found in codon 125 of domain 1, corresponding to the presence of a Leu-125 or a Val-125 allele, respectively, which we provisionally designated CD31.L and CD31.V. No other polymorphism was seen. In particular, the polymorphisms suggested by other investigators to be present at codons 115, 151, 375, and 391 were not found in any of the 21 samples (i.e., the 42 haplotypes) we sequenced (Table 1).
|
Immunoreactivity
We sought evidence of an immunologic difference between the two CD31 alleles by flow-cytometric analysis of cultured umbilical-vein endothelial cells homozygous for CD31, using three anti-CD31 monoclonal antibodies. Figure 2 shows that antibody L133 bound equally to cells of both homozygous types, whereas antibody 7E8 bound to cells of the CD31.L type two to four times more than to those of the CD31.V type. Antibody EF70 bound 1.6 to 1.7 times more to cells of the CD31.L type than to those of the CD31.V type. This differential immunoreactivity was also seen in pairwise comparisons among the seven other umbilical-vein endothelial cell lines homozygous for either CD31.L or CD31.V that were tested with the three monoclonal antibodies (data not shown).
|
We sought to determine the clinical relevance of the CD31 LeuVal polymorphism in patients who had received bone marrow from sibling donors with HLA-identical genotypes. Because current protocols of bone marrow transplantation are rarely complicated by graft rejection, the more frequent problem of GVHD was chosen as an outcome measure of non-HLA histoincompatibility.1-4 Patients with no GVHD were compared with those who had grade III or IV acute GVHD. These two well-defined groups had similar clinical characteristics (Table 2) before transplantation and a similar distribution of HLA and ABO antigens (data not shown). Among the characteristics generally agreed to affect the risk of acute GVHD (age, disease status, GVHD prophylaxis, and whether or not the donor and the recipient are of the same sex),2,4 only the median ages of the two groups differed significantly (the patients with acute GVHD were older).
|
|
Discussion
The HLA system of major histocompatibility antigens is well understood in humans, but little is known about other clinically important alloantigens. We studied the CD31 adhesion molecule as a possible minor alloantigen because of its tissue distribution and suspected heterogeneity in the population. We identified an allelic LeuVal polymorphism at codon 125 of the CD31 gene. Because we identified no other replacements at codon 125 in the 42 haplotypes we sequenced and because the frequencies of the various genotypes among the patients we tested fit the Hardy–Weinberg equilibrium, the Leu and Val alleles accounted for the great majority (approximately 98 percent) of the polymorphisms at that codon. We could not verify the amino acid replacements reported by others at codons 115, 151, 375, and 391. These apparent variations in findings may be due to errors in cloning or sequencing or to the presence of rare, as yet unvalidated alleles (with gene frequencies of 0.02 or less).
The distribution of the alleles at codon 125 correlated with the ability of homozygous cell lines to react with particular anti-CD31 monoclonal antibodies. It is known that a single LeuVal substitution in immunoglobulin light chains gives rise to allotypes that can be distinguished with the use of alloantibodies33 and that a difference of a single amino acid between allelic forms of cell-surface molecules can cause the production of alloantibodies. Examples are the exchange of proline for leucine at codon 33 of the platelet-specific HPA-1 antigen34 and the exchange of aspartic acid for leucine at codon 156 of the HLA B44 allele. The rejection of bone marrow allografts has been traced to the difference of a single amino acid between the HLA B44 alleles.35
The immunogenicity of minor histocompatibility antigens can be determined definitively in vivo on the basis of graft rejection or the occurrence of GVHD among recipients of HLA-identical grafts. We therefore looked for a relation between CD31 matching and GVHD in patients who had received bone marrow from siblings with the same HLA genotype. Recipients of marrow from CD31-nonidentical donors had a substantially higher risk of acute GVHD than recipients of CD31-identical marrow. As expected,2,4 the median age of the patients without GVHD was lower than that of those with acute GVHD.
We used CD31 identity between the graft and the host rather than compatibility (in which the recipient had no CD31 alleles that were foreign to the donor) to assess the relation between CD31 and acute GVHD. This approach required no preconceived notions about mechanisms or directions of compatibility and averted the need to address poorly understood phenomena of transplantation, such as hybrid resistance.36,37 Nevertheless, when we analyzed the CD31 compatibility (rather than identity) of the host with the graft, the results were similar. CD31 thus appears to be a clinically important minor histocompatibility antigen in humans.
The occurrence of GVHD in some recipients of CD31-identical marrow indicates that CD31 alloantigens are not the exclusive determinants of transplantation compatibility at minor loci, a finding consistent with reports about other minor histocompatibility markers.5-11 The clear involvement of the CD31 alleles in GVHD has potentially important clinical implications. Prospective matching for CD31 alleles may reduce the risk of GVHD in patients undergoing bone marrow transplantation by affording the opportunity to select among suitable related donors with HLA-identical genotypes or among unrelated, HLA-matched donors when more than one such donor is available.
Other, more complex, explanations for the association between CD31 and GVHD are possible. For example, in the CD31 segments we did not sequence there could be amino acid substitutions that contribute to alloantigenicity and are found predominantly with specific alleles at codon 125. Alternatively, CD31 alleles may serve only as markers for other minor histocompatibility loci that are linked to CD31 on chromosome 17.38 However, the immunologic difference between the two CD31 phenotypes is consistent with the hypothesis that CD31 itself has alloantigenicity.
CD31 is believed to have a role in attracting leukocytes to inflammatory sites, the activation of lymphocytes, interactions between leukocytes and endothelial cells, and perhaps angiogenesis.21,39 It would be interesting to know whether the CD31 polymorphism we found affects these physiologic functions differentially and whether the functions have a role in the relation between CD31 polymorphism and GVHD.
Supported in part by grants from the National Institutes of Health (HL33811) and the American Cancer Society (JFRA-446).
We are indebted to Dr. P. Newman, Milwaukee, for providing early access to the CD31 genomic sequence; to Drs. M. Metzelaar and K. Nieuwenhuis, Utrecht, the Netherlands, for the 7E8 antibody; to Dr. E. Engleman, Stanford, Calif., for the EF70 antibody; to Dr. E. Gazit, Tel-Hashomer, Israel, and Dr. M. Amylon, Stanford, for DNA samples from their patients; to Dr. P. Bryant, Portsmouth, Va., for human umbilical-vein endothelial cells; to Dr. A. Ginat, Stanford, for technical suggestions; to Drs. G. Long, K. Blume, and H. Sussman, Stanford, and Dr. L. Rosenbaum, Santa Cruz, Calif., for critical review of the manuscript; to the staff of the Stanford Histocompatibility and Bone Marrow Transplant Laboratories for their assistance; and to Ms. M. Guzman for her secretarial perseverance.
Source Information
From the Departments of Pathology (E.B., D.D.H., S.K., J.L.Z., F.C.G.), Medicine (N.J.C., J.L.Z.), and Health Research and Policy (B.W.B.), Stanford University, Stanford, Calif.
Address reprint requests to Dr. Grumet at the Stanford University Blood Center, 800 Welch Rd., Palo Alto, CA 94304.
References
| |||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
