Assessing Antibody Strength: Comparison of MFI, C1q, and Titer Information

Transcription

1 American Journal of Transplantation 215; 15: Wiley Periodicals Inc. C Copyright 215 The American Society of Transplantation and the American Society of Transplant Surgeons doi: /ajt Assessing Antibody Strength: Comparison of MFI, C1q, and Titer Information A. R. Tambur 1,5, *, N. D. Herrera 1,5, K. M. K. Haarberg 1,5, M. F. Cusick 1,5, R. A. Gordon 2,5, J. R. Leventhal 3,5, J. J. Friedewald 4,5 and D. Glotz 6 1 Transplant Immunology Laboratory, Northwestern University Feinberg School of Medicine, Chicago, IL 2 Division of Cardiology, Northwestern University Feinberg School of Medicine, Chicago, IL 3 Department of Surgery, Northwestern University Feinberg School of Medicine, Chicago, IL 4 Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL 5 Comprehensive Transplant Center, Northwestern University Feinberg School of Medicine, Chicago, IL 6 Nephrology and Transplantation Service, H^opital Saint- Louis, Paris, France Corresponding author: Anat R. Tambur, a-tambur@northwestern.edu The presence of donor-specific HLA antibodies before or after transplantation may have different implications based on the antibody strength. Yet, current approaches do not provide information regarding the true antibody strength as defined by antigen antibody dissociation rate. To assess currently available methods, we compared between neat mean fluorescence intensity (MFI) values, C1q MFI values, ethylenediaminetetraacetic acid (EDTA)-treated samples, as well as titration studies and peak MFI values of over 7 Luminex-based single-antigen HLA antibody data points. Our results indicate that neat MFI values do not always accurately depict antibody strength. We further showed that EDTA treatment (6%) does not always remove all inhibitory factors compared with C1q or titration studies. In this study of patients presenting with multiple antibody specificities, a prozone effect was observed in 71% of the cohort (usually not affecting all antibody specificities within a single serum sample, though). Similar to titration studies, the C1q assay was able to address the issue of potential inhibition; however, its limitation is its low sensitivity and inability to detect the presence of weak antibodies. Titration studies are the only method among the approaches used in this study to provide information suggesting antigen antibody dissociation rates and are, therefore, likely to provide better indication of true antibody strength. Abbreviations: AMR, antibody-mediated rejection; DSA, donor-specific antibody; EDTA, ethylenediaminetetraacetic acid; MFI, mean fluorescence intensity; SAB, single antigen beads; SPA, solid phase assays Received 28 November 214, revised 9 February 215 and accepted for publication 28 February 215 Introduction Solid phase assays (SPA) revolutionized our ability to predict virtual crossmatching prior to transplant or to perform posttransplant monitoring even in the absence of donor cells. A significant limitation of this assay, however, is its inability to quantify antibody levels. SPA readout is provided in mean fluorescence intensity (MFI) units, which in a perfect world could have been correlated with the number of antibodies attached to the HLA antigens on the microparticle beads; however, this is not the case and even the manufacturers claim to report at best semiquantitative data (1). Yet, many studies have been attempting to correlate between antibody strength as determined by MFI value and transplant-related outcomes (2 5). The significance of being able to report quantitative levels of HLA antibodies comes to play as we evaluate donor recipient compatibility, when considering desensitization protocols and even more so, posttransplantation when determining the best therapeutic approach to remove donor-specific antibodies (DSA) in the face of an antibodymediated rejection (AMR). Biologically speaking, antibody strength refers to the intensity of affinity and avidity for a particular antigen antibody complex. In other words, antibody strength provides a measure for the kinetics of antigen antibody binding, or more accurately, antigen antibody dissociation. This information is not provided by MFI values. At Northwestern University, titration studies are being used to assess antibody strength for the past few years. As a result, we identified a significant proportion of patients exhibiting some inhibitory effects in their serum, causing the appearance of a false negative or a falsely reduced MFI levels (prozone effect). We further recognized the gap between neat MFI values and actual antibody strength as measured by antigen antibody dissociation rates (loosely assigned as antibody titer). Other recently published 2421

2 Tambur et al approaches suggest methods to eliminate prozone effect by pretreatment of the serum sample with ethylenediaminetetraacetic acid (EDTA) (6), or the use of the complement binding SPA to determine the more detrimental DSA (7,8). In this report, we provide comparison between the different approaches and highlight the added layer of quantitative information obtained by titration studies. Materials and Methods Serum samples included in this study were selected based on the following criteria: (1) A clinical indication warranted titration studies (i.e. a patient with a potential living donor but with HLA-DSA; patients in whom prozone was suspected based on antibody response patterns on the neat assay; patients tested in conjunction with for-cause biopsies when DSA were detected); (2) At least one strong antibody (MFI >1 ) was detected. All consecutive samples that answer these criteria were included. Fifty-five patients from the Northwestern Comprehensive Transplant Center were tested using Luminex SPA for the presence of HLA class I (N ¼ 27) and HLA class II (N ¼ 49) antibodies, using LabScreen W Single Antigen Antibody Detection Tests and the C1qScreen TM (both One Lambda, Inc., Canoga Park, CA) following manufacturer s recommendations to obtain neat MFI and C1q MFI values, respectively. The same serum samples were tested using a series of doubling dilutions (with PBS) to determine antibody Titer (the dilution in which a positive antibody becomes negative). Serum samples from 2 patients were further tested using EDTA pretreatment as described by Schnaidt et al (6). Briefly, 95 ml of serum was incubated with 5 ml of EDTA (Cellgro/Mediatech, Inc., Manassas, VA, Cat# cl) for a final concentration of 25 mm, and was then added to SA beads. Peak MFI was determined as the highest MFI value observed during titration to represent elimination of all inhibitory effects. All serum samples were tested using a single lot of SAB and C1q reagents. We, therefore, had a total of 2619 data points for class I and 4459 for class II, providing 778 MFI data points for each of the four analyses performed. EDTA pretreatment of serum was aborted following analysis of initial samples. The patient population tested for this study included 3 males and 25 females; mean age 45 17; 7 Hispanic, 15 Black, 14 Caucasian, three Asian, 16 patients with unreported ethnicity. The patients mean PRA was 38 36% for class I and 57 3% for class II. Statistical analysis was performed using Pearson Correlation, GraphPad, Prism 6 software (San Diego, CA). Results Antibody signatures of serum samples from 55 patients (76 assays) were run using three different variations of the Luminex single antigen (SA) solid phase immunoassay for both HLA class I and class II: (1) Conventional IgG SA assay; (2) serial dilution/titration studies, as described in methods; and (3) the C1q complement binding SPA. In addition, several serum samples were pretreated with EDTA prior to running the conventional SA assay, following published protocol (6), in an attempt to remove inhibitory factors. Figure 1 provides comparison between the three assaypermutations, as well as the EDTA pretreatment, showing results obtained from a single serum sample. MFI values of only a few beads (1 HLA-DQ alleles) are provided for ease of visualization. Titration studies were performed up to a 1:124 dilution. HLA-DQ allele specificity is provided in legends, but has no significance for data interpretation. The different antibodies are further labeled 1 1 based on antibody strength the strongest antibody being the one that has the highest MFI at the highest dilution. A prozone (inhibitory) effect is clearly visible for a few specificities within this serum sample, namely alleles 1 8. In these examples, the neat MFI value is lower than the MFI value in the subsequent dilution(s). Alleles 9 and 1 show the expected dilution pattern in which the MFI values decrease as the serum sample is diluted. It is important to note that a prozone effect does not necessarily affect all antibody specificities within this single serum sample, and does not have the same trajectory for all affected specificities. Therefore, the presence/absence of prozone affecting one antibody-specificity cannot be implied by the presence/ absence of prozone for other specificities within that same serum sample. In fact, in this study, all serum samples showed this phenomenon. EDTA pretreatment of serum samples was proposed as a mean to remove inhibitory factors, thus, uncovering prozone effect, and supposedly providing a more accurate estimate of antibody strength. Figure 1 clearly shows that while EDTA indeed removed some of the inhibitory effects, it did so only for some of the antibody specificities. In this particular example, the prozone effect was at least partially removed for specificities 2 and 4 8 but was not removed for antibody specificities 1 and 3. The failure to fully unmask prozone is not unique to this example and was observed in other EDTA treated serum samples as well. The C1q assay showed better correlation with antibody strength overall, although the hierarchy of the MFI values obtained for this assay did not match the one showed by titration studies. For example, if we compare beads 2 and 8, both showing prozone effect: bead 2 starts at about 3 MFI, peaks at 21 (1:64 dilution) and remains very strong with 18 MFI at a 1:124 dilution. Bead 8 starts with somewhat high neat MFI of about, peaks at 16 5 MFI (1:16 dilution) but drops to an MFI value of 5 at a 1:124 dilution, thus, showing that bead 8 is significantly weaker than bead 2. Yet, the corresponding MFI values for the C1q assay are about 7 for bead 2 and 14 for bead 8, suggesting that bead 8 is stronger than bead 2. A total of 2619 and 4459 data points for HLA class I and HLA class II antibody specificities, respectively, were analyzed and the results are presented in Figure 2. Two different comparisons are shown for each class of antibodies. MFI values obtained using the C1q assays were compared, per bead, to neat MFI values (two left panels) and to the highest MFI value obtained in titration studies peak MFI value unmasking any prozone effect (two right panels). The correlation between either class I or class II neat MFI values and their respective C1q values is quite poor, as 2422 American Journal of Transplantation 215; 15:

3 Determining Antibody Strength Figure 1: Luminex SA analysis was performed on a single serum sample (all performed in one tube). MFI values of only a few beads (1 HLA-DQ alleles; specificities are provided in legends at bottom of figure; however, these have no significance for data interpretation) are provided for ease of visualization. The results from three assay permutations (neat with titration studies up to a titer of 1:124, EDTA pretreatment to attempt elimination of inhibitory factors [prozone effect], and C1q MFI results) are presented. The different antibodies are also labeled 1 1 based on antibody strength the strongest antibody being the one that has the highest MFI at the highest dilution. A prozone (inhibitory) effect is clearly visible for a few specificities within this serum sample, namely alleles 1 8. In these examples, the neat MFI value is lower than the MFI value in the following dilution(s). Alleles 9 and 1 show the expected dilution pattern, in which the MFI values decrease as the serum sample is diluted. Of note, the sequence of what is considered the strongest antibody by the different methods is not concordant. EDTA, ethylenediaminetetraacetic acid; MFI, mean fluorescence intensity. indicated by the low r values (.461 and.482 for class I and class II, respectively). This is in contrast with a better correlation coefficient between the peak MFI values and C1q MFI results (.786 and.869, for class I and class II, respectively). The correlations between C1q and either neat MFI or peak MFI values for the individual HLA loci are presented in Figures S1 and S2. It is interesting that class I loci HLA-A, -B, and -C behave quite similarly; however, while results for HLA-DRB1 show similar correlation pattern between C1q and either neat or peak MFI values (r ¼.788 and r ¼.856, respectively), the results obtained for HLA-DQ and HLA-DP loci are quite different. Evidently, much of the reduced correlation (r ¼.344) between neat MFI and C1q for the HLA-DQ antibodies is due to prozone effect, given the high number of data points shifting to the left in the peak MFI data. Yet, even with peak MFI, the correlation coefficient reaches a value of only.66, suggesting additional factors that may be associated with the discrepancy observed. It is more difficult to assess the discrepancy observed for the DP locus as only a few beads show positive results (r ¼.197 for neat MFI and r ¼.689 for peak MFI). American Journal of Transplantation 215; 15: Prozone effect was determined as at least 1% increase in MFI value between the neat and the peak MFI values. Prozone effect, as revealed by titration studies, was observed for at least one antibody specificity in 11/27 class I assays (4%); 32/49 class II assays (65%) and overall in 29/55 patients (71%). Many antibody specificities reached highest MFI values at a 1:16 dilution, but some, as shown in Figure 1, reach their peak MFI values at much higher dilutions (1:256 in this example). The number of individual beads that showed prozone effect is only about.5% of the total beads tested. This observation may explain the lack of appreciation of the true prevalence of prozone effect among the highly sensitized patients, especially those presenting with multiple antibody specificities, when no titration studies are performed. Thus, using an approach in which only a single value is obtained per antibody specificity (no titration), if prozone affects a bead that does not carry a DSA, the likelihood of this phenomenon to go undetected is very high. Figure 3 depicts the correlation between C1q MFI value and Titers assignment for both HLA class I and HLA class II 2423

4 Tambur et al Class I C1q MFI Class II Neat MFI r=.461 C1q MFI r= Peak MFI C1q MFI Neat MFI r=.486 C1q MFI r= Peak MFI Figure 2: MFI values of C1q assay were plotted against the same beads results obtained by either neat/conventional IgG assay (left) or against the highest MFI value observed in titration studies peak (right) for either class I (N ¼ 2619) or class II (M ¼ 4459 data points). A much higher concordance was observed for the C1q versus peak MFI values (r ¼.786,.869) compared with the concordance observed for the C1q versus neat MFI values (r ¼.461 and.486, respectively). Specific information regarding the individual HLA loci is provided in supplemental figures. The big shift of positive C1q data points from low neat MFI values to high peak MFI values represent prozone effect. MFI, mean fluorescence intensity. antigens. The overall correlation is similar to that observed between C1q and peak MFI values. However, differences are observed for the individual loci (Figures S3, S4 and Table 1), where HLA-DQ, HLA-DP and HLA-C loci correlations show substantial improvement. These data demonstrate that C1q results correlate best with IgG titration studies. Most of the outliers are observed in higher titer antibodies especially for class II. Upon examination of the data, the vast majority of the outliers are HLA-DQ antibodies. Figure 3 also demonstrates that the C1q assay is likely to miss the lower titer (weaker) antibodies. In general, the class-i threshold for C1q binding is at a titer of around 1:16 1:32 and about a log higher for class II titers of 1:32-1:64, although the threshold changes for the individual loci as seen in the Supporting Information. Recently, C1q assays were suggested as a diagnostic tool to assess antibody strength in relation to transplant outcome. Figure 4 depicts three different ranges of C1q MFI values plotted against titer information for some of the corresponding beads. The lower range was chosen as 3 5 MFI as most labs choose it to determine positive C1q assay. Our data, left inset, clearly show that while most of the beads in this example indeed have a relatively low titer (1:16 1:64), a few antibodies have higher titers in which even a dilution of 1:124 shows an MFI values >1. The second inset looks at a range of C1q MFI values of 1 11, representing weakly positive antibodies. In this example, most antibodies have a titer of up to 1:256, but there are still a few antibody specificities remaining positive at higher titer. Lastly, using a C1q MFI range of 7 based on the cut-off value proposed to distinguish between worse and better clinical outcomes in a recent study (9), the variability of responses ranged from fairly weak to moderate strength, and a group of very high strength antibodies. These clinically significant nuances in antibody strength, observed by titration studies, are completely invisible by either of the other methods. The following examples are provided to illustrate the superiority of the titration approach especially when antibody strength determination is critical: 2424 American Journal of Transplantation 215; 15:

5 Determining Antibody Strength C1q :248 1:124 1:512 1:256 Class I 1:128 1:64 1:32 1:16 Titer 1:8 1:4 r=.784 1:2 1:1 Neat C1q 4 2 1:248 1:124 1:512 1:256 Class II 1:128 1:64 1:32 1:16 Titer 1:8 1:4 r=.881 1:2 1:1 Neat Figure 3: C1q MFI values were plotted against the same beads assigned titer information for either class I (N ¼ 2619) or class II (M ¼ 4459 data points). A vertical line is placed to emphasis the points from which C1q values become positive. For class I that point correlates with a titer of 1:16 1:32; for class II it is a little higher with a titer of 1:32 1:64. Thus, statistically speaking, C1q assays will be negative for titers lower than those mentioned above. Note that there are many more outlier data points for class II compared with class I. Specific information regarding the individual HLA loci is provided in the Supporting Information. MFI, mean fluorescence intensity. BJ, a 43YO African-American male, 18 months posttransplant was tested for the presence of DSA due to clinical presentation of reduced kidney function. Antibodies to both donor HLA-DQ specificities were detected in this serum sample, both with MFI values of around 1 11 MFI. The results of the top 1 highest MFI beads are shown in Table 2 together with dilution studies. DSAs are represented by beads 6 and 8 (highlighted; all specificities are from the same serum sample, run in a single well simultaneously). The results of the dilution studies clearly show that the MFI value of top seven beads, including the DSA represented by bead 6 became negative at or around titer of 1:124. The bottom three beads, including the DSA represented by bead 8 became negative between the titers of 1:16 and 1:64 (therefore assumed to be at a titer of 1:32). Our experience with desensitization or treatment of AMR suggests that the antibody recognizing bead 6 (DSA with a titer of 1:124) will likely require more intervention to be removed compared with the antibody that recognizes bead 8. In this case, both antibodies represent DSAs; thus, both needed to be treated. However, if these DSA were present in two different patients, the treatment design and the prognosis would have been different despite the fact that they both show very similar neat MFI values. The second example is of a highly sensitized 37YO female patient awaiting heart transplantation. Given the breadth and depth of sensitization, the decision was to use bortezomib-based desensitization (1,11). Table 3 provides the follow-up data on 12 representative antibodies, from the pretreatment time point up to 6 months of follow-up. The table provides the C1q MFI information on the left and the corresponding titer information on the right. Five representative groups of beads are shown. In the top group, the three antibodies show C1q MFI values of about 17 with varying decrease in MFI values over the period, but still remaining with fairly strong C1q values. This corresponds to baseline titers of 1:496, reduced to titers of 1:128 at the end of the period. The lower group shows three additional antibodies with similar C1q results (in the 2 MFI range), reduced to what is likely to be considered transplantable range following the second treatment cycle. These results correspond nicely with the titer information as well. The three pairs of beads in the middle show baseline C1q MFI values that are 2 3 MFI units apart (15 38/14 912, /13 953, and /13 192) but varying antibody strength using titration studies (1:248/ 1:512, 1:248/1:256, 1:496/1:256, respectively). The decrease in antibody strength observed for beads 5, 7, Table 1: Correlation between different approaches currently used to assign antibody strength, for different HLA loci Correlation HLA-A HLA-B HLA-C HLA-DR HLA-DQ HLA-DP C1q vs. Neat C1q vs. Peak C1q vs. Titers American Journal of Transplantation 215; 15:

6 Tambur et al Figure 4: Three different ranges of C1q MFI values were chosen to illustrate the range of different neat and peak MFI values associated with them as well as the actual titration plots. The lowest C1q MFI range shown is 3 5 fluorescent units with a large range of IgG MFI values, peak values, and titers up to >1:124; a moderate range shows C1q MFI values between 1 11 fluorescent units with even a wider range of the IgG studies; and lastly a fairly high range of C1q MFI values showing some antibodies with treatable titer information. MFI, mean fluorescence intensity. and 9 could have been predicted by the lower titers observed at baseline (1:512, 1:256, 1:256, respectively), but not by the high C1q values. On the other hand, the counterpart beads in each pair, with similar C1q values, indeed show much lower responses to treatment, as could have been predicted by the higher titer information. The response to treatment could have been predicted a priori based on the lower titer information associated with these antibody specificities but not by the C1q MFI values. Discussion The ability to assess antibody strength is critical for performing virtual crossmatching, risk stratification pretransplantation, choosing more appropriate therapeutic interventions posttransplantation and monitoring treatment efficacy (12 16). SPA assays provided our field with sensitive and specific means to detect the presence of HLA antibodies, but offer semi-quantitation of antibody strength at best (17). In an attempt to express antibody strength, many researchers and clinicians resorted to describing high MFI value antibodies as high titer antibodies. This practice is probably a carry-over from ABO antibody language, but is, of course, an erroneous representation of true antibody strength since titers cannot be assigned without actually performing dilution/ titration studies. The authors caution the readers to critically verify the actual method used for antibody strength assignment in different publications while they form their conclusions. Recently, another approach was introduced to stratify the more pathogenic antibodies utilizing their ability to bind C1q, assuming that this assay identified only those antibodies that have the capacity to bind complement and, therefore, those are the more harmful DSA (8). The present report provides side-by-side comparisons of over 7 data points in which the conventional singleantigen assay was run (neat MFI), the matching dilution studies (titer) were conducted and C1q assay was performed (C1q MFI). To address the inhibition/prozone phenomenon that results in false negative or falsely low 2426 American Journal of Transplantation 215; 15:

7 Determining Antibody Strength Table 2: Comparison between neat MFI values of several HLA-DQ antibody specificities with the corresponding MFI values as the serum was diluted, in a patient presenting with AMR and in need for antibody removal therapy Neat MFI Bead ID Neat 1:4 1:16 1:64 1:256 1: DQA1*3:1/DQB1*2:1 6 DQA1*3:2/DQB1*3:3 2 DQA1*3:1/DQB1*3:1 7 DQA1*3:2/DQB1*3:2 3 DQA1*3:3/DQB1*4:1 8 DQA1*1:3/DQB1*6:1 4 DQA1*3:1/DQB1*3:3 9 DQA1*1:3/DQB1*6:3 5 DQA1*3:1/DQB1*3:2 1 DQA1*1:2/DQB1*6:9 AMR, antibody-mediated rejection; MFI, mean fluorescence intensity. MFI values, we treated some of the serum samples with EDTA as suggested previously (6,18). Our results demonstrated that EDTA treatment (6%) is not always sufficient to remove all inhibitory factors, leaving some antibody specificities present with lower MFI values relative to their accurate strength, as presented in Figure 1. We, therefore, chose to follow all our samples using peak MFI values, where we can get confirmation that no further increase in antibody strength occurs following additional dilutions. This approached uncovered the very high prevalence of the inhibition phenomenon, especially in patients with multiple antibody specificities. In the current study, the vast majority of the patients (71%) exhibited prozone that affected at least one antibody specificity. As we were writing this manuscript, two additional publications using EDTA to uncover prozone were reported (19,2). Those studies used higher concentrations of EDTA and claim to uncover all prozone present. It is possible that a higher concentration of EDTA may have unmasked prozone better in our study as well; however, the main message of our work is to show the added value of titration in obtaining clinically relevant nuances of antibody strength, regardless of prozone. Our experience, as well as reports by others (21), suggested that the capacity to bind C1q is associated with antibody strength. Yet, a simple correlation between neat MFI values and C1q values show fairly poor r-values (Figure 2 and Figures S2 and S3). In fact, previous reports indeed indicated that some patients might have IgG negative but C1q positive antibodies (22), using this observation to support the superiority of the C1q assay. We argue that the main cause for this poor relationship is likely due to prozone effect, as evaluating those same samples using peak MFI values (no inhibition present) increased the American Journal of Transplantation 215; 15: r-values significantly. This is nicely illustrated by the shift from left to right for most of the strongly positive C1q data points when comparing the neat and peak MFI charts (summarized in Table 1). In our study, the locus that showed highest susceptibility to prozone effects is HLA-DQ. Interestingly, this is also the locus in which more patients exhibit de novo donor-specific antibodies (23 25). Moreover, even with the use of peak (rather than neat) MFI values, the correlation with C1q results was only.66. On the other hand, correlating C1q data to titer assignment showed improved r-values, up to.87 (Figure S4). This observation suggests that additional factors associated with antibody strength are likely unique to the HLA-DQ locus. HLA-DQ antibodies also showed the highest rate of outliers, even in the titer to C1q correlation, especially for the higher titer antibodies (1:128 and higher). The implications of this observation are not clear and will need further assessment, potentially by correlating antibody strength to responsiveness to treatment via desensitization or elimination of AMR. Figure 3 clearly shows a titer threshold below which C1q results are likely to be negative. This threshold is around titers of 1:16 1:32 for class I antibodies and a log higher for class II antibodies (1:32 1:64). This observation can be explained by the fact that a certain antibody density needs to be available for the complement to bind (26). Lower titer antibodies are not likely to reach this critical density and, thus, may appear as non-c1q binding. Our concern is that although these antibodies should not prevent transplantation, they harbor a definite risk of rejection. In a transplant scenario, these non-c1q-binding antibodies may be underappreciated for their potential pathogenicity. Not receiving 2427

8 Tambur et al Table 3: Comparison between C1q MFI values and corresponding titers of HLA-DQ antibodies in a patient undergoing Bortezomib desensitization Titer C1q MFI Bead 6M 3M 1M Post-2X Pre-2X Post-1X 6M Nmb Pre- 3M 1M Post-2X Pre-2X Post-1X Pre- DQA1*1:3/DQB1*6: DQA1*1:2/DQB1*6: DQA1*1:1/DQB1*5: DQA1*3:3/DQB1*4: DQA1*2:1/DQB1*3: DQA1*1:1/DQB1*6: DQA1*3:2/DQB1*3: DQA1*1:2/DQB1*6: DQA1*3:2/DQB1*3: DQA1*2:1/DQB1*3: DQA1*3:1/DQB1*3: DQA1*3:1/DQB1*3: proper monitoring may end up causing an unexpected rejection. This may explain the data reported by Loupy et al (9) in which the worst hazard ratio was observed in patients that were IgG positive/c1q negative before transplant but switched to IgG positive/c1q positive (whether these are C1q negative that became positive or de novo C1q-positive antibodies). Our findings also suggest that the association found between C1q positivity and worse graft outcome may be due mainly to the higher level of DSA per se and not necessarily associated with complement binding capabilities. Over the past few years, several modalities attempting to remove or counter the effects of HLA antibodies were developed. Those include, among others, the use of different doses of IVIg, plasmapheresis, rituximab, bortezomib and its derivatives, and inhibitors of the complement cascade. Each of these modalities has different side effects, and varying range of costs associated with the treatment. Being able to determine a priori what is the most effective, yet least toxic, combination of therapies to remove an antibody seems to be a desirable approach. We believe that the most granular data can be obtained by performing actual titration studies. While the C1q assay has provided thus far the most reliable data in stratifying de novo DSA production and transplant outcome, Figure 4 illustrates the inability of the C1q assay approach to capture quite a few nuances in antibody affinity and avidity, or antibody strength that is shown by the titration studies. The two cases presented further demonstrate this point. A criticism raised against the dilution approach is the potential increase in cost. Indeed, if one chooses to run a complete set of serial dilutions, this is a valid point. However, a trending of antibody responses can be obtained by running the neat sera plus 2 additional dilutions, for example 1:16 and 1:128 (or other permutations based on the comfort level of each center in treating different levels of antibodies). The added cost of this approach includes some additional labor time and reagents for two additional tests. This added cost is, in fact, similar to the cost associated with running the C1q assay (in which one also needs to run the two sets of beads, plus the added cost of the quite expensive C1q reagent). Thus, the difference in cost is diminished, while the granularity of data is increased significantly. Needless to say, only a portion of the patients will require the use of the titration approach. In our experience, patients that benefit from this approach are patients with multiple antibody specificities, patients requiring the use of antibody removal strategies and patients in whom the efficacy of these approaches need to be monitored. Consequently, the minimal increase in expense for antibody titration is more than offset by the ability to avoid the financial and human costs of a catastrophic AMR and graft loss. In summary, we provided data indicating that neat MFI values do not always accurately depict antibody strength American Journal of Transplantation 215; 15:

9 Determining Antibody Strength We further showed that EDTA treatment (6%) does not always remove all inhibitory factors. The prevalence of prozone effect is quite high among patients with multiple antibody specificities (at least one specificity exhibited prozone effect in the vast majority of the patients enrolled in this study). Similar to titration studies, the C1q assay was able to address the issue of potential inhibition; however, its limitation is its low sensitivity and inability to detect the presence of weak antibodies. The reported correlation between C1q binding ability and worse graft outcome may be an indication of antibody strength and not necessarily reflection on complement binding abilities per se. To the best of our knowledge, titration studies are currently the only method to provide more granular information about antigen antibody dissociation rates and, therefore, are likely to provide better indication of true antibody strength. The clinical relevance of this finding still remains to be validated in future studies. Disclosure The authors of this manuscript have conflicts of interest to disclose as described by the American Journal of Transplantation. Denis Glotz has received research funding and honoraria for presentation from ThermoFisher. References 1. Tait BD, Susal C, Gebel HM, et al. Consensus guidelines on the testing and clinical management issues associated with HLA and non-hla antibodies in transplantation. Transplantation 213; 95: Marfo K, Ajaimy M, Colovai A, et al. Pretransplant immunologic risk assessment of kidney transplant recipients with donor-specific anti-human leukocyte antigen antibodies. Transplantation 214; 98: Alachkar N, Lonze BE, Zachary AA, et al. Infusion of high-dose intravenous immunoglobulin fails to lower the strength of human leukocyte antigen antibodies in highly sensitized patients. Transplantation 212; 94: Zachary A, Reinsmoen NL. Quantifying HLA-specific antibodies in patients undergoing desensitization. Curr Opin Organ Transplant 211; 16: Lefaucheur C, Loupy A, Hill GS, et al. Preexisting donor-specific HLA antibodies predict outcome in kidney transplantation. J Am Soc Nephrol 21; 21: Schnaidt M, Weinstock C, Jurisic M, Schmid-Horch B, Ender A, Wernet D. HLA antibody specification using single-antigen beads-a technical solution for the prozone effect. Transplantation 211; 92: Sutherland SM, Chen G, Sequeira FA, Lou CD, Alexander SR, Tyan DB. Complement-fixing donor-specific antibodies identified by a novel C1q assay are associated with allograft loss. Pediatr Transplant 212; 16: Tyan DB. New approaches for detecting complement-fixing antibodies. Curr Opin Organ Transplant 212; 17: Loupy A, Lefaucheur C, Vernerey D, et al. Complement-binding anti-hla antibodies and kidney-allograft survival. New Engl J Med 213; 369: Morrow WR, Frazier EA, Mahle WT, et al. Rapid reduction in donorspecific anti-human leukocyte antigen antibodies and reversal of antibody-mediated rejection with bortezomib in pediatric heart transplant patients. Transplantation 212; 93: Patel J, Everly M, Chang D, Kittleson M, Reed E, Kobashigawa J. Reduction of alloantibodies via proteasome inhibition in cardiac transplantation. J Heart Lung Transplant 211; 3: Campbell P Clinical relevance of human leukocyte antigen antibodies in liver, heart, lung and intestine transplantation. Curr Opin Organ Transplant 213; 18: Susal C, Roelen DL, Fischer G, et al. Algorithms for the determination of unacceptable HLA antigen mismatches in kidney transplant recipients. Tissue Antigens 213; 82: Chang D, Kobashigawa J. The use of the calculated panel-reactive antibody and virtual crossmatch in heart transplantation. Curr Opin Organ Transplant 212; 17: Zachary AA, Sholander JT, Houp JA, Leffell MS. Using real data for a virtual crossmatch. Hum Immunol 29; 7: Tambur AR, Ramon DS, Kaufman DB, et al. Perception versus reality?: Virtual crossmatch-how to overcome some of the technical and logistic limitations. Am J Transplant 29; 9: Gebel HM, Bray RA. HLA antibody detection with solid phase assays: Great expectations or expectations too great? Am J Transplant 214; 14: Guidicelli G, Anies G, Bachelet T, et al. The complement interference phenomenon as a cause for sharp fluctuations of serum anti-hla antibody strength in kidney transplant patients. Transplant Immunol 213; 29: Visentin J, Vigata M, Daburon S, et al. Deciphering complement interference in anti-human leukocyte antigen antibody detection with flow beads assays. Transplantation 214; 98: Schwaiger E, Wahrmann M, Bond G, Eskandary F, Bohmig GA. Complement component C3 activation: The leading cause of the prozone phenomenon affecting HLA antibody detection on singleantigen beads. Transplantation 214; 97: Zeevi A, Lunz J, Feingold B, et al. Persistent strong anti-hla antibody at high titer is complement binding and associated with increased risk of antibody-mediated rejection in heart transplant recipients. J Heart Lung Transplant 213; 32: Chen G, Sequeira F, Tyan DB. Novel C1q assay reveals a clinically relevant subset of human leukocyte antigen antibodies independent of immunoglobulin G strength on single antigen beads. Hum Immunol 211; 72: Everly MJ, Rebellato LM, Haisch CE, et al. Incidence and impact of de novo donor-specific alloantibody in primary renal allografts. Transplantation 213; 95: Kim JJ, Balasubramanian R, Michaelides G, et al. The clinical spectrum of de novo donor-specific antibodies in pediatric renal transplant recipients. Am J Transplant 214; 14: Walsh RC, Brailey P, Girnita A, et al. Early and late acute antibodymediated rejection differ immunologically and in response to proteasome inhibition. Transplantation 211; 91: Diebolder CA, Beurskens FJ, de Jong RN, et al. Complement is activated by IgG hexamers assembled at the cell surface. Science 214; 343: American Journal of Transplantation 215; 15:

10 Tambur et al Supporting Information Additional Supporting Information may be found in the online version of this article. Figure S1: MFI values of C1q assay were plotted against the same beads results obtained by either neat / conventional IgG assay (left) or against the highest MFI value observed in titration studies peak (right) for HLA-A, HLA-B, and HLA-C antibodies. The correlation values are higher for the C1q versus peak comparison than those calculated for the C1q versus neat MFI values. Figure S2: MFI values of C1q assay were plotted against the same beads results obtained by either Neat / conventional IgG assay (left) or against the highest MFI value observed in titration studies Peak (right) for HLA-DRB1, HLA-DQ, and HLA-DP antibodies. The correlation values are higher for the C1q versus peak comparison than those calculated for the C1q versus neat MFI values. Figure S3: C1q MFI values were plotted against the same beads assigned titer information for HLA-A, HLA- B, and HLA-C antibodies. Very few positive data points were available for HLA-C antibodies. Figure S4: C1q MFI values were plotted against the same beads assigned titer information for HLA_DRB1, HLA-DQ, and HLA-DP. Very few positive data points were available for HLA-DP antibodies. Note the high level of outlier results obtained specifically for HLA-DQ antibodies. 243 American Journal of Transplantation 215; 15: