Characterization of HLA Class I SpecificAntibodies by ELISA Using SolubilizedAntigen Targets: II. Clinical Relevance

Andrea A. Zachary, Lloyd E. Ratner, Julie A. Graziani,Donna P. Lucas, Nancy L. Delaney, and Mary S. Leffell

ABSTRACT: Until recently, the nature of humoral sen-sitization to HLA has been characterized by data fromlymphocyte-based assays, predominantly cytotoxicitytests. We have examined the characteristics, determinedby enzyme-linked immunosorbent assay (ELISA), of serafrom 191 subjects known to have produced HLA-specificantibody. We found that ELISA detected higher frequen-cies compared with cytotoxicity of many (74.5%), but notall, HLA-specific antibodies; in many cases (42.6%) thefrequencies of these antibodies were higher than predictedfrom population frequencies whereas some antibodies(23.4%) occurred with lower than expected frequencies.Some of the increase in frequencies can be accounted forby crossreactivity, i.e., sensitization to epitopes sharedamong two or more allelic products. The presence of

epitopes shared between a recipient’s antigen and a mis-matched antigen in a donor also tended to narrow thespecificity of antibody produced. However the data alsoindicate differences in immunogenicity among differentantigens suggesting that crossreactive group matchingwould be beneficial in some but not all cases. Finally, wepresent case studies to illustrate the value of ELISA inpredicting humoral rejection episodes and in monitoringthe efficacy of rejection therapies. Human Immunology 62,236–246 (2001). © American Society for Histocompat-ibility and Immunogenetics, 2001. Published by ElsevierScience Inc.

KEYWORDS: HLA antibody; ELISA; immunogenicity;CREG

ABBREVIATIONSCREG crossreactive groupCYT cytotoxicityELISA enzyme-linked immunosorbent assay

QID GTI QuikID assayQS GTI QuikScreen assay

INTRODUCTIONThe detrimental effects of antibodies specific for donorhuman leukocyte antigens (HLA) are well documented.These effects include hyperacute rejection (in the mostsevere case), increased frequencies of acute rejection ep-isodes, and probably certain aspects of chronic rejection(reviewed in [1]). Factors determining the effect of an-tidonor HLA antibodies on a transplant include the typeof organ being transplanted and the titer and immuno-

globulin class of the antibody. Because transplanted kid-neys are highly susceptible to antibody-mediated injury,crossmatches that are positive by techniques of lowersensitivity are considered an absolute contraindication bymost, and those positive by highly sensitive method areconsidered a contraindication in high risk situations suchas regrafts. However, a false-positive reaction due tonon-HLA specific antibodies may prevent a transplantfrom occurring, inappropriately, and using a techniquethat does not detect low levels of antibody specific fordonor HLA may result in injury to the graft. Therefore,it is important to detect and characterize donor-specificantibodies through the appropriate testing of serum frompatients both before and after transplantation.

HLA molecules bear multiple antigenic epitopes,many of which are the so-called “public” epitopes thatare shared among the products of several different HLA

From the Departments of Medicine (A.A.Z., J.A.G., D.P.L., N.L.D.,M.S.L.) and Surgery (L.E.R.), Johns Hopkins University School of Medi-cine, Baltimore, MD.

Address reprint requests to: Andrea A. Zachary, Ph.D., ImmunogeneticLaboratory, Johns Hopkins University School of Medicine, 2041 E. Monu-ment Street, Baltimore, MD 21205; Tel: (410) 614-8978; Fax: (410)955-0431; E-mail: aaz@jhmi.edu.

Received October 30, 2000; revised November 29, 2000; accepted De-cember 6, 2000.

Human Immunology 62, 236–246 (2001)0198-8859/01/$–see front matter© American Society for Histocompatibility and Immunogenetics, 2001

Published by Elsevier Science Inc. PII S0198-8859(00)00253-6

alleles, resulting in the apparent crossreactive groups ofantigens (CREGs) [2–4]. These shared epitopes providethe opportunity to sensitize a patient to multiple HLAantigens with only a single antigen mismatch [4]. Sen-sitization to HLA antigens is a major factor reducingaccess to transplantation and accounts for most, if not all,of the disparity between Caucasians and African-Ameri-cans in the rate of renal transplantation [5, 6]. Therefore,it is of considerable interest to know if all HLA epitopesare equally immunogenic and if any given epitope isequally immunogenic in individuals with different HLAphenotypes.

Historically, investigations of HLA-specific antibod-ies were conducted using the methods of cytotoxicity(CYT) and/or flow cytometry on lymphocyte targets.Effective complement activation in the cytotoxicity assayrequires the presence of two IgG antibodies in closeproximity [7], either antibodies specific for two differentHLA epitopes on the same molecule, the so-called syn-ergistic lysis [8, 9], or one HLA-specific antibody and anantiglobulin reagent [10]. Thus, unless an antiglobulinreagent is used, the complement-dependent cytotoxicityassay will fail to detect the presence of a single alloanti-body. While flow cytometry performed on lymphocytetargets overcomes this shortcoming, neither assay readilydifferentiates HLA- specific antibodies from other lym-phocyte-reactive antibodies. Solid phase immunoassaysutilizing solubilized HLA molecules as targets have beenused recently to detect and characterize HLA-specificantibodies. We and others have shown that these assayshave a high degree of sensitivity that exceeds the anti-globulin-augmented cytotoxicty assay and may equalthat of lymphocyte-based flow cytometry [11–13]. Fur-ther, these assays more readily differentiate betweenHLA-specific antibodies and those specific for othermarkers. These attributes enable a more accurate assess-ment of humoral alloreactivity. In turn, this will permita more accurate assessment of the immunogenicity ofvarious epitopes. We present our findings using an en-zyme-linked immunosorbent assay (ELISA) and the GTIQuikID (QID; GTI, Brookfield, WI, USA), which pro-vides evidence of a qualitative difference in the antibod-ies detected by cytotoxicity and those detected by solidphase immunoassays. Notably, we have evidence of pos-sible differences in immunogenicity among groups ofantigens. We also present examples of the utility of thistype of assay in monitoring patients with low levels ofantibody specific for mismatched donor HLA.

MATERIALS AND METHODSThe study population consisted of 191 subjects known toproduce HLA-specific antibodies. Demographics of thestudy group are given in Table 1. One hundred eighty-

two were patients awaiting renal transplantation, 66 hadreceived a transplant previously and 9 were parous fe-males. Data from the nine parous females were excludedfrom certain analyses because their HLA phenotypes hadnot been determined. For this analysis we assessed thecharacteristics of 431 serum specimens obtained from the191 patients. Sera were tested by ELISA using the GTIQuikID kit and by complement-dependent microlym-phocytotoxicity according to methods previously de-scribed [14]. Validation of the efficacy of this assay ispresented in a companion article (see this issue of HumanImmunology) [14]. Antibody specificity was determinedby computer analysis with manual verification for cyto-toxicity test data and by manual analysis by two indi-viduals for ELISA data, as described previously [14].

We determined the frequency of antibodies of differ-ent specificity as follows:

frequency of anti-X5number subjects with anti-X

number subjects with antibody

where X is an HLA antigen and the denominator is thenumber of subjects with any antibody of HLA specificitydefined by the assay being examined.

Crossreactive groups considered in the analyses were asfollows: A1-A3-A11-36; A1-A9; A1-A10-A11; A10-A11;A2-A9-A28; A2-B17; A25-A32; A29-A30-A31-A32-A33; A28-A33-A34; B5-B18-B35-B53-B70; B7-B8;B7-B22-B42; B7-B27-B47; B7-B41-B48-B60-B61-B81;B8-B14; and B15-B17.

The expected frequencies of antibodies were deter-mined by calculating the probability that a person wholacks an antigen would encounter that antigen, as fol-lows. Expected frequency of antibody to antigen x 5(1 2 frequency of study patients possessing antigen x) z(frequency of antigen x in the general population). Thequantity, 1 2 frequency of patients lacking antigen x,was determined from the population of antibody produc-ers. The frequency of antigen x in the population wastaken from the frequencies in the United States organdonor population [15] and two values were determinedfor the expected frequency of each antibody using anti-gen frequency data from both Caucasians and African-Americans.

TABLE 1 Study population

Race Number Gender Number

African-American 98 Male 69Caucasoid 85 Female 122Other 8 Total 191

237Properties of HLA Antibodies Defined by ELISA

RESULTSNature of ELISA-Defined AntibodiesWe have shown a locus-related disparity between resultsobtained by ELISA versus those obtained by cytotoxicity,with ELISA finding a larger proportion of patients hav-ing antibodies to antigens of both the A and B loci ratherthan to antigens of one locus only [14]. To determine if

these differences were related to particular antigens, weexamined the frequencies of individual antibody speci-ficities. The frequencies of the antibodies defined by QIDand CYT for the A and B locus antigens are given inFigures 1 and 2, respectively. The designation A9 wasused when both A23 and A24 were present, while thefrequencies given for the designations of A23 and A24were for instances where only one of those two antibodieswas identified. Similarly, the A10 designation was usedfor those sera giving reactivity with A25, A26, A34, andA66. B40 represents the combination of B60 and B61.These frequencies are not intended to represent individ-ual antibodies; that is, a serum reacting with B7, B22,and B42, which could result from the presence of one tofive different antibodies, was scored for each of those

FIGURE 1 Frequencies of antibodies specific for A locusantigens detected by ELISA versus cytotoxicity. The designa-tions 9T and 10T are for the sums of the frequencies of A9,A23, A24, and A10, A25, A26, A34, A66, respectively.

FIGURE 2 Frequencies of antibodies specific for B locusantigens detected by ELISA versus cytotoxicity.

238 A. A. Zachary et al.

specificities. There are several points to note from thesecharts. There is an overall increased frequency of anti-bodies detected by QID versus CYT. However, the dif-ference in sensitivity varies among the antigens. Appre-ciable differences for A locus antigens occurred withBw4-related antigens, A9, A25, and A32, as well as with

A31 and A36, whereas much smaller differences wereseen with A1, A2, A3, and A28 (Figure 1). In contrast,substantial differences were seen for most B locus anti-gens occurring with sufficient frequency to detect differ-ences. These antigens included B7, B8, B14, B15, B17,B18, B22, B27, B35, B40, B60, B41, B42, B47, B48,B70, and B81 (Figure 2).

Because differences in the frequencies of antibodiesdetected by the two assays varied according to specificity,we sought to determine if these differences reflecteddifferences in the capabilities of the assays, differences inthe immunogenicity of the antigens, or both. In order toassess these possibilities, we first compared the antibodyfrequencies observed for the two assays with those pre-dicted from population frequencies. These data areshown in Figures 3 and 4. Antibodies detected by cyto-

FIGURE 3 Expected and observed frequencies: antibodiesspecific for A locus antigens. Expected frequencies were deter-mined using antigen frequencies for two populations: African-American and Caucasian.

FIGURE 4 Expected and observed frequencies: antibodiesspecific for B locus antigens. Expected frequencies were deter-mined using antigen frequencies for two populations: African-American and Caucasian.

239Properties of HLA Antibodies Defined by ELISA

toxicity were within or below the expected frequencyranges in all but three cases: A10, A11, and A28. Thefrequencies of different antibodies detected by ELISAexceeded expected frequencies for A9, A10, A11, A28,A29, A31, A32, A36, B7, B17, B22, B27, B40, and B41and were appreciably below expected frequencies for B5,B12, B53, B70, and Bw6 and, to a lesser extent, Bw4.These data suggest that some antibodies occur at agreater than expected frequency because of sharedepitopes and the ability of an ELISA to detect a singleantibody.

Crossreactivity as a cause for greater than expectedantibody frequencies is supported by the data provided inTable 2. Here we have provided the numbers of cases inwhich reactivity with a particular antigen occurred aspart of CREG reactivity, i.e., along with reactivity toother, crossreactive antigens, or by itself (i.e., as a private

specificity). In only 48 of 238 (0.20) instances did reac-tivity occur as a private specificity. We further examinedto what extent the occurrence of antibodies to apparentprivate epitopes was due to CREG blocking, which wedefine here as the failure to make antibodies to some orall of the antigens crossreactive with those in one’s ownphenotype, i.e., to one’s own epitopes. This occurred in32 of 48 (67%) of the cases in which a patient made anantibody to a private specificity. The total number ofinstances of CREG antibody was determined by countingany instance in which the antibody reacted with two ormore members of a CREG as one instance, and countingantibodies that reacted with two or more members of twodifferent CREGS as two or more. For example, serumreacting with B7, B22, and B42 would be counted as oneinstance whereas one reacting with A2, A9, A28, B7,B22, and B42 would be counted as two instances. Casesin which the nature of the antibody could not be dis-cerned were not included in this analysis. For example,for a serum reacting with A1, A2, A9, and A28 it is notpossible to determine if the reactivity with A1 is as aprivate specificity or as part of the A1-A9 CREG.

It is not apparent from these data either why certainantibodies were less frequent than expected or whetherantigens differ in their immunogenicity. We examinedthe occurrence of antibodies to antigens mismatched inthe 60 transplanted patients to investigate those issues.The data are presented in Table 3. Auto-CREG mis-matches (defined here as mismatches of antigens cross-reactive with one or more of the patient’s antigens) arethought to be less immunogenic than mismatches out-side the patient’s CREGs (referred to here as allo-CREG)[16, 17]. In addition, our own data presented in Table 2suggests that there is a diminished response, overall, toauto-CREG mismatches. Therefore we also adjusted forthose occurrences. The numbers for individual antigensare small. However, when the most common mismatchesare examined, there is a uniformly high degree of re-sponse to mismatches, with some exceptions. The ad-justed frequency of response to A1, A2, A3, A9, A25,A26, and B7, collectively, was 62 of 73 (85%). Incontrast, the adjusted frequency of response to B5 andB35, taken together, was 1 of 8 and the unadjustedfrequency was 1 of 18. Interestingly, although there wasa high degree of responsiveness to B7 mismatches (11/11) the response to other members of the various B7CREGs was mediocre (B8, B27, B40 5 10/21) eventhough observed frequencies of antibodies to most mem-bers of B7 CREGS were greater than expected. We alsodetermined the frequency of occurrence of auto-CREGantibody, independent of transplantation, in order toassess the extent to which CREG blocking might occur.The occurrence of auto-CREG antibodies, according tothe different antigens possessed by the patient, approx-

TABLE 2 Occurrence of private vs. CREG antibodies

Antibodyspecificity

Number of occurrences asCREG

blockingaPart of CREG Private specificity

A1 30 2 2A3 20 2 2A11 23 1 0A36 16 0 0A9 22 2 1A2 34 7 4A28 26 0 0A10 41 8 8A29 6 1 1A30 7 0 0A31 14 5 2A32 21 0 0A33 8 0 0B5b 5 0 0B18 7 0 0B35 12 3 1B53 5 0 0B70 12 0 0B7 34 5 5B8 5 7 5B22 23 0 0B40 31 0 0B41 28 0 0B42 23 0 0B47 19 0 0b48 28 0 0B81 31 0 0B15 5 3 0B17 19 2 1Total 190c 48 32 (67%)

a Number of times patient making antibody with “private” specificity pos-sessed an antigen in the CREG.b Includes B51 and B52.c The total reactivity cannot be obtained from summing the column becauseinstances of antibody reactive with multiple antigens in a CREG are listed foreach of the antigens.

240 A. A. Zachary et al.

imates the antibody frequencies expected from antigenpopulation frequencies with a few exceptions, as shownin Table 4. When categorized by previous graft, auto-CREG antibody was found in 61% of previously trans-planted patients and 55% of those not previously trans-planted (x1

2 5 0.55, p 5 0.47) (Table 5).

Application of ELISA to Patient MonitoringAs we have demonstrated here and in other publications,ELISA is capable of rapid, highly sensitive, specific de-tection of HLA-specific antibodies [11, 14]. This pro-vides an invaluable tool for monitoring alloreactivity inpatients. We have performed ELISAs in parallel with allflow cytometric crossmatches. The QID test is run in allcases in which there is a positive T-cell crossmatch and apositive GTI QuikScreen assay (QS) test. In eight ofeight cases of positive T-cell crossmatch, the QID testwas able to identify antibody specific for mismatcheddonor HLA. We have used QID to monitor patients withlow levels of HLA class I specific antibody and, inparticular, patients undergoing our plasmapheresis-hy-perimmune globulin treatment protocol [18]. We pro-vide two examples representing the utility of ELISA.

Patient ST. Patient ST received a second, cadaveric renaltransplant mismatched for A23, 29, B72, DR4, andDR7. Despite a negative antiglobulin crossmatch, thepatient experienced an episode of biopsy proven, acutehumoral rejection at 72 h following transplantation. Thepatient demonstrated antibody detectable by cytotoxicitytesting for 6 years prior to the transplant. The antibodyspecificity included A1, A3, A11, and periodically A9with the last occurrence of A9-specific antibody being at18 months pretransplant. For a period of 3 months, 4years prior to the transplant, the patient also demon-strated antibody to B35 and B53. Cytotoxicity screenswere negative at the time of and for 5 months prior to thetransplant. However ELISAs of pretransplant sera re-vealed the presence of antibodies specific for A3, B5,B18, B35, B53, and B70, which were the most likelycause of the humoral rejection.

Patient CR (Figure 5).Patient CR had been highly sensitized. Over time,

antibody specific for a mismatched antigen of the poten-tial donor (anti-A1) diminished below levels detectableby cytotoxicity testing whereas other HLA-specific anti-

TABLE 3 Antibodies against mismatched antigens

Mismatchedantigen

Number ofoccurrences

Number withantibody

CREGblockinga

Adjustedantibody occurrenceb

A1 12 10 (83)c 1/2 10/11A2 20 17 (85) 1/3 17/19A3 11 5 (45) 3/6 5/8A9 17 7 (41) 7/10 7/10A25 6 5 (75) 1/1 5/5A26 10 7 (70) 1/3 7/9A11 7 3 (43) 0/4 3/7A28 4 1 (25) 2/3 1/2A29 3 2 (67) 0/1 2/3A30 6 2 (33) 1/4 2/5A33 4 1 (25) 2/3 1/2B51 5 0 1/5 0/4B7 11 11 (100) — 11/11B8 7 2 (28) 1/5 2/6B44 6 4 (67) 0 4/6B45 1 0 1/1 —B14 3 0 0 0/3B62 6 0 1/6 0/5B16 5 0 0 0/5B17 3 1 (33) 0 1/3B21 5 1 (20) 0 1/5B27 7 4 (57) 0 4/7B35 13 1 (8) 9/12 1/4B40 10 4 (40) 2/6 4/8Bw4 9 4 (44) 5/5 4/4

a Ratio is the number having an antigen crossreactive with the mismatch over the number lacking antigen to the mismatch.b Ratio is number with antibody to mismatch over number of occurrences of mismatch minus number with CREG blocking(numerator from ratioa).c Numbers in parentheses are percentages.

241Properties of HLA Antibodies Defined by ELISA

bodies remained. ELISA revealed the continued presenceof antibody specific for HLA-A1 and the patient wastreated with plasmapheresis, hyperimmune globulin, andimmunosuppression prior to transplant. The presence ofA1-specific antibody was confirmed by absorption withcells selected to remove antibodies of other specificities.Although present at very low levels at the time of trans-plant, donor reactivity disappeared soon after transplantwith continued treatment. At 14 days following trans-plant the patient experienced acute humoral rejectionand was again treated with plasmapheresis and hyperim-mune globulin. ELISA revealed the recurrence of anti-body specific for HLA-A1 and its subsequent disappear-ance following treatment.

DISCUSSIONWe have previously established that the QS and QIDELISAs for HLA-specific antibody are rapid, reliable,specific, and more sensitive than cytotoxicity assays [11,14]. In our companion article data from the ELISAsuggest that the disparity between ELISA and cytotox-icity is not uniform, with ELISA detecting more anti-bodies specific for B locus antigens than does cytotoxicity

[14]. Here we have set out to further evaluate character-istics of antibodies detected by ELISA. Comparisons ofantibodies detected by the two assays revealed that dif-ferences were antigen specific and, for the most part,appeared to correlate with known crossreactive groups ofantigens, i.e., antigens that share epitopes. This suggeststhat the basis of the disparity is the ability of ELISA todetect the presence of a single HLA-specific antibody.

Disparities in the frequencies of antibodies detectedby the two methods were seen for some but not allcrossreactive groups. This could be explained by differ-ences in immunogenicity, in the numbers and distribu-tion of shared epitopes, or in exposure as a result ofchance differences in the frequency of exposure to variousantigens. We examined data from two sources in anattempt to evaluate these different possibilities (expectedversus observed antibody frequencies), which maximizedthe available data and response to mismatched antigens.The results obtained with cytotoxicity tests exceeded therange of expected frequencies for antibodies specific forA10, A11, A28, and A32 and were below expectedfrequencies for antibodies specific for A3, B5 (51), B8,B12, B44, B15, B16, B17, B57, B18, B21, B25, Bw4,and Bw6. In contrast, the majority of antibodies detected

TABLE 4 Occurrence of auto-CREG antibodies

Patient’santigen Number

Number withauto-CREG antibody Antibody specificities and occurrencea

A1 31 14 A3 (2) [exp: 5–8]b, A9 (7), A10 (5), A11 (1)A2 60 22 A9 (14), A28 (4) [exp: 5–11], B17 (4) [exp: 5–16]A3 33 6 A1 (4), A11 (2)A23 17 13 A1 (6), A2 (9)A24 13 6 A1 (2), A2 (4),A25 4 3 A1 (1), A11 (1), A32 (1)A26 13 5 A1 (3), A25 (2)A34 10 7 A1 (1), A11 (2), A25 (2), A28 (3)A66 5 1 A1A11 14 9 A1 (2), A3 (3), A10 (6) [exp: 1–2]A28 32 13 A2 (12), A34 (1)A30 35 1 A31A32 9 1 A25A33 15 3 A10 (1), A28 (2)A36 9 2 A1 (1), A3 (1)B7 31 3 B8 (1) [exp: 2–6], B27 (1), B47 (1), B48 (1), B40 (1), B81 (1)B8 31 12 B7 (11) [exp: 6–8], B14 (1)B62 12 5 B17 [exp: 1–3]B17 35 14 A2B18 19 3 B35 (3), B70 (2)B22 13 3 B7B53 23 2 B18 (1), B35 (2), B70 (2)B60 15 4 B7 (4), B41 (1), B81 (1)There was no occurrence of auto-CREG antibodies for the following antigens: A31 (7)c, B51/52 (25) [exp: 6–15], B14 (12), B35 (19) [exp: 3–11],

B41 (5), B47 (1), B63/75/78 (5), B70 (12) [exp: 3]

a Numbers in parentheses are occurrences of the antibody.b Expected numbers are shown, in brackets, for those cases where the observed numbers differ from the expected.c Numbers in parentheses are occurrences of the antigen.

242 A. A. Zachary et al.

FIGURE 5 Antibody screen results for patient CR. Anti-gens given in graphs (top and bottom) indicate antibodyspecificities. Antibodies specific for donor antigen are given in

bold. AHG-XM 5 antiglobulin crossmatch; Tx 5 transplant;REJ 5 rejection. Top panel: cytotoxicity testing; bottompanel: ELISA testing.

243Properties of HLA Antibodies Defined by ELISA

by the QID exceeded the expected range. Many of thesecould be explained by reactivity with shared epitopesthat was detected by ELISA but not by cytotoxicity.However, the greater than expected frequencies of anti-A10, -A11, -A28, and -A32 and the reduced frequenciesof anti-B5, -B12, -B44, -B18, -B35, -Bw4, and - Bw6occurring with both assays suggests the possibility ofdifferences in immunogenicity. Alternatively, it is pos-sible that for those antibodies seen less frequently thanexpected there were, by chance, fewer sensitizing events.However, it is curious that three of the specificities, B5,B18, and B35, are members of the same CREG. Toevaluate possible differences in immunogenicity we ex-amined the occurrence of ELISA-detected antibodies pro-duced in response to mismatched antigens. In makingthis assessment we allowed for the possibility of CREG-blocking, i.e., the reduced potential to make antibodiescrossreactive with one’s own antigens. The data unad-justed for CREG blocking indicate that the antigens canbe roughly divided into three groups according to thefrequency of response elicited: strong mismatches (thosethat evoked antibody in more than two-thirds of thecases: A1, A2, A25, A26, B7), intermediate mismatches(those evoking antibody in one-third to two-thirds of thecases: A3, A9, A11, A29, A30, B44, B17, B27, B40, andBw4), and weak mismatches (those evoking antibody inless than one-third of the cases: A28, A33, B51, B8, B21,and B35). These data suggest several things. First, thereduced response to mismatches of certain antigens is incontrast to the higher than expected frequencies of anti-bodies that were observed for those antigens (e.g., A28)provides further evidence that the antibodies were mostlikely generated in response to other antigens bearingshared epitopes (i.e., A2). Second, there are differences inimmunogenicity of the antigens. It has been reportedthat individuals are less likely to respond to antigenscrossreactive with their own antigens (auto-CREG) thanto antigens in other CREGS (allo-CREG), a phenomenonwe refer to as CREG blocking [16, 17]. We examinedwhether this could account for the reduced response tosome antigens and attempted to account for this possi-bility by excluding apparent cases of CREG blockingfrom the analysis (Table 3). The adjusted occurrences ofantibody to mismatched antigens indicate that CREGblocking could account for a substantial portion of the

lack of response to mismatches of A9, B35, and Bw4 butmade an appreciable difference in responsiveness to onlyA9 and Bw4. B35 remained in the weak mismatch groupeven after accounting for CREG blocking. Unlike B35,very little of the lack of responsiveness to B51 could beattributed to CREG blocking, suggesting that some orall the members of this crossreactive group are less im-munogenic.

We have reported elsewhere [19] that mismatching ofthe public epitopes Bw4 and Bw6 does not contribute toan increased frequency of rejection or of sensitization.The data here suggest that the responsiveness to a Bw4mismatch can be strong but that it is often tempered bythe presence of A9, A25, or A32 in the recipient and,therefore, Bw4 mismatches would only be possible inindividuals who are both homozygous for Bw6 and lack-ing A9, A25, and A32.

The possibility of CREG blocking is the basis fororgan allocation schemes based on matching for CREGsrather than individual antigens. Justification for theseschema comes from data indicating that CREG blockingdoes occur [20, 21]. We also present data here showingthat in two-thirds of the cases in which a patient makesantibody against only one member of a CREG, thatpatient possesses another member of the CREG as part oftheir own antigen phenotype. The ELISA data presentedhere demonstrated, that among patients producingHLA-specific antibody, antibody specific for auto-CREGantigens occurred, in most cases, with a frequency ex-pected from population frequencies. However many ofthe HLA-specific antibodies occurred with higher thanexpected frequencies overall (Figures 3 and 4). Thereforethe occurrence of auto-CREG antibodies only at expectedfrequencies supports the concept of CREG-blocking. Au-to-CREG antibodies occurring with a lower than ex-pected frequency included anti-A3 in patients with theA1 antigen and anti-B8 in those having the B7 antigen,whereas higher frequencies of anti-A10, anti-B7, andanti-B17 were observed in patients possessing A11, B8,and B15 antigens, respectively. HLA-A1, A10, and B7were among those mismatches associated with the high-est occurrence of antibody. Thus, it appears that thepatient is less likely to make auto-CREG antibody whenone or more of these antigens occurs in the patient’sphenotype, but more likely to make auto-CREG anti-body when they are a mismatched donor antigen. Per-haps some antigens are more effective at CREG blockingas well as at overcoming CREG blocking. Unlike reportsof others [20], we did not find a significant difference inthe occurrence of auto-CREG antibody among previouslytransplanted versus nontransplanted patients. Our obser-vations were not because of lack of exposure to auto-CREG antigens among transplanted patients because therate of auto-CREG-specific antibody was comparable

TABLE 5 Production of auto-CREG antibodies

Previouskidney

transplant

Production of auto-CREG antibody

Yes No

Yes 40 (60.6) 26 x12 5 0.55, p 5 0.47

No 62 (54.9) 51

244 A. A. Zachary et al.

among patients known to have been mismatched for anauto-CREG antigen (27 of 71, 61%). We believe thedifferences between our observations and those of othersmay be due to lower levels or more restricted clonality ofantibodies in nontransplanted patients, the detection ofwhich was improved by the increased sensitivity of theELISA used here compared with cytotoxicity assays usedelsewhere. Thus, although the potential benefit of CREGmatching is supported by tests of low levels of antibody,CREG blocking is not absolute and further studiesshould be conducted to evaluate qualitative differences inCREG matches.

The relevance of antibody specific for mismatcheddonor HLA antigens is well established (reviewed in [1]).Low levels of such antibodies do not produce the imme-diate and dramatic effects resulting from high titeredantibodies but lead to increased and repeated insult tothe graft that, in turn, results in reduced graft functionand survival. The ability to detect low levels of antibod-ies provides the opportunity to prevent or overcomeacute humoral rejection. Like others [13, 22–24], wehave found solid phase immunoassay to be superior tocytotoxicity for detecting low levels of antibodies specificfor mismatched HLA antigens and in correlating withnonhyperacute types of rejection. We have providedsome examples of utilization of ELISA in monitoringpatients. Because it is rapid, unaffected by antibodiesspecific for antigens other than HLA, highly sensitive,and not dependent on an adequate supply of viable donorcells it is invaluable in monitoring patients’ humoralresponse to mismatched HLA antigens. By performingroutine tests of pretransplant sera, we can better assessthe risk for acute humoral rejection without the need forperforming flow cytometric crossmatch tests prospec-tively on cadaver donors. Similarly, these tests can be useto monitor the development and course of donor HLA-specific antibodies following transplantation and, inturn, identify developing problems before they are de-tected clinically.

We believe that solid phase immunoassays have thepotential to provide new information about immunoge-nicity that is relevant to matching for transplantationand about HLA epitopes. We propose that a large scalenational or international study be conducted to permitaccumulation of sufficient data to address these issues.

REFERENCES

1. Zachary AA, Hart JM: Relevance of antibody screeningand crossmatching in solid organ transplantation. In Lef-fell MS, Donnenberg A, Rose NC (eds.): Handbook ofHuman Immunology. Boca Raton, FL: CRC Press, 1997.

2. Legrand F, Dausset J: Serological evidence of the existenceof several antigenic determinants (or factors) on the HL-A

gene products. In Dausset J, Colombani J (eds.): Histo-compatibility Testing 1972. Copenhagen: Munksgaard,1973.

3. Rodey GE, Fuller TC: Public epitopes and the antigenicstructure of HLA molecules. In Atassi MZ (ed.): CriticalReviews in Immunology. Boca Raton, FL: CRC Press,1987.

4. Schwartz BD, Luehrman LK, Rodey GE: Public antigenicdeterminant on a family of HLA-B molecules: basis forcrossreactivity and a possible link with disease predispo-sition. J Clin Invest 64:938, 1979.

5. Kallich JD, Adams JL, Barton PL, Spritzer KL: Access toCadaveric Kidney Transplantation. Santa Monica, CA:RAND, 1993.

6. Thompson JS: American Society for Histocompatibilityand Immunogenetics crossmatch study. Transplantation59:1636, 1995.

7. Humphrey JH, Dourmashkin RR: Electron microscopestudies of immune cell lysis. In Wolstenholme GEW,Knight J (eds.): Ciba Foundation Symposium. London:J&A Churchill Ltd., 1965.

8. Elliott EV, Pindar A, Stevenson FK, Stevenson GT: Syn-ergistic cytotoxic effects of antibodies directed againstdifferent cell surface determinants. Immunology 34:405,1978.

9. Howard JC, Butcher GW, Galfre G, Milstein C, MilsteinCP: Monoclonal antibodies as tools to analyse the serolog-ical and genetic complexities of major transplantationantigens. Immunol Rev 47:139, 1979.

10. Fuller TC, Fuller AA, Golden M, Rodey GE: HLA alloan-tibodies and the mechanism of the antiglobulin-aug-mented lymphocytotoxicity procedure. Hum Immunol56:94, 1997.

11. Lucas DP, Paparounis ML, Myers L, Hart JM, ZacharyAA: Detection of HLA class I-specific antibodies by theQuikScreen enzyme-linked immunosorbent assay. ClinDiagn Lab Immunol 4:252, 1997.

12. Moore SB, Ploeger NA, DeGoey SR: HLA antibodyscreening: comparison of a solid phase enzyme-linkedimmunoassay with antiglobulin-augmented lymphocyto-toxicity. Transplantation 64:1617, 1997.

13. Zaer F, Metz S, Scornik JC: Antibody screening by en-zyme-linked immunosorbent assay using pooled solubleHLA in renal transplant candidates. Transplantation 63:48, 1997.

14. Zachary AA, Delaney NL, Lucas DP, Leffell MS: Charac-terization of HLA class I specific antibodies by ELISAusing solubilized antigen targets: I. Evaluation of the GTIQuikID assay and analysis of antibody patterns. HumImmunol 62:228, 2001.

15. Leffell MS, Steinberg AG, Bias WB, Machan CH, ZacharyAA: The distribution of HLA antigens and phenotypesamong donors and patients in the UNOS Registry. Trans-plantation 58:1119, 1994.

16. Starzl TE, Eliasziw M, Gjertson D: HLA and cross-reac-

245Properties of HLA Antibodies Defined by ELISA

tive group matching for cadaver kidney allocation. Trans-plantation 64:983, 1997.

17. Laundy GJ, Bradley BA: The predictive value of epitopeanalysis in highly sensitized patients awaiting renal trans-plantation. Transplantation 59:1207, 1995.

18. Montgomery RA, Zachary AA, Racusen LC, Leffell MS,King KE, Burdick J, Maley WR, Ratner LE: Plasma-pheresis and intravenous immune globulin provide ef-fective rescue therapy for refractory humoral rejectionand allow kidneys to be successfully transplanted intocross-match positive recipients. Transplantation 70:887,2000.

19. Leffell MS, Kraus E, Racusen LC, Ratner LE, Charney D,Zachary AA: The effect of Bw4 and Bw6 epitope mis-matches on antibody production, acute and chronic rejec-tion, and graft survival in renal allografts. Transplanta-tion, submitted.

20. Fernandez-Fresnedo G, Pastor JM, Ruiz JC, CotorrueloJG, Setien MA, Lopez-Hoyos M, Arias M: Differences inanti-CREG antibody formation between transplanted and

nontransplanted renal patients. Transplantation 67:1188,1999.

21. Papassavas, AC, Iniotaki-Theodoraki A, Boletis J, Ko-stakis A, Stavropoulos-Giokas C: Development of anti-HLA-antibodies against intra-CREG-mismatches in renaltransplant recipients. Transplant Proc 31:757, 1999.

22. Jaramillo A, Smith MA, Phelan D, Sundaresan S, TrulockEP, Lynch JP, Cooper JD, Patterson GA, MohanakumarT: Development of WLISA-detected anti-HLA antibodiesprecedes the development of bronchiolitis obliterans syn-drome and correlates with progressive decline in pulmo-nary function after lung transplantation. Transplantation67:1155, 1999.

23. Schonemann C, Groth J, Leverenz S, May G: HLA class Iand class II antibodies: monitoring before and after kidneytransplantation and their clinical relevance. Transplanta-tion 65:1519, 1998.

24. Sumitran-Karuppan S: The clinical importance of choos-ing the right assay for detection of HLA-specific donor-reactive antibodies. Transplantation 68:502, 1999.

246 A. A. Zachary et al.