The Role of Immunoglobulin-G Subclasses and C1q in De Novo HLA-DQ Donor-Specific Antibody Kidney Transplantation Outcomes

Abstract

Background 

Anti–HLA-DQ antibodies are the predominant HLA class II donor-specific antibodies (DSAs) after transplantation. Recently, de novo DQ DSA has been associated with worse allograft outcomes. The aim of this study was to determine the further complement-binding characteristics of the most harmful DQ DSA.

Methods 

Single-antigen bead technology was used to screen 284 primary kidney transplant recipients for the presence of posttransplantation DQ DSA. Peak DSA sera of 34 recipients with only de novo DQ DSA and of 20 recipients with de novo DQ plus other DSAs were further analyzed by a modified single-antigen bead assay using immunoglobulin (Ig)-G subclass-specific reporter antibodies and a C1q-binding assay.

Results 

Compared with recipients who did not have DSA, those with de novo persistent DQ-only DSA and with de novo DQ plus other DSAs had more acute rejection (AR) episodes (22%, P=0.005; and 36%, P=0.0009), increased risk of allograft loss (hazards ratio, 3.7, P=0.03; and hazards ratio, 11.4, P=0.001), and a lower 5-year allograft survival. De novo DQ-only recipients with AR had more IgG1/IgG3 combination and C1q-binding antibodies (51%, P=0.01; and 63%, P=0.001) than patients with no AR. Furthermore, the presence of C1q-binding de novo DQ DSA was associated with a 30% lower 5-year allograft survival (P=0.003).

Conclusions 

The presence of de novo persistent, complement-binding DQ DSA negatively impacts kidney allograft outcomes. Therefore, early posttransplantation detection, monitoring, and removal of complement-binding DQ might be crucial for improving long-term kidney transplantation outcomes.

In Brief

Supplemental digital content is available in the article.

The presence of de novo, posttransplantation, anti–donor specific antibodies (DSAs) has been frequently associated with acute rejection (AR) and allograft loss in different solid-organ transplantations (^1–3). Posttransplantation anti–human leukocyte antigen (HLA) class II antibodies have been shown to be a risk factor for late kidney allograft failure (⁴). Morales-Buenrostro et al. (⁵) have recently described a high prevalence of anti–HLA-DQ antibodies in nonalloimmunized Mexican blood donors. They suggested that those HLA antibodies were produced to cross-reactive epitopes found in microorganisms, ingested proteins, and allergens. Antibodies against HLA-DQ were not only common in the nonalloimmunized population but also highly prevalent in kidney transplant recipients (^6–8). In addition to its evident prevalence in transplant recipients, HLA-DQ appears to be more resistant to treatment with proteasome inhibitors (⁹), reinforcing the importance of early detection. The United Network for Organ Sharing has recently amended its policy, requiring deceased-donor HLA typing to include identification of the HLA-DQ antigens before making any kidney, kidney-pancreas, pancreas, or pancreas islet offers (¹⁰). Unfortunately, changes in allocation still do not include HLA-DQ matching in the algorithms for organ allocation programs, despite recent reports by different transplantation centers independently citing high prevalence of HLA-DQ DSA, which were associated with a significant risk of AR and allograft loss (^{11, 12}).

Solid-phase assays, especially single-antigen bead (SAB) technology, have detected DSA in much lower concentrations and have undoubtedly changed researchers’ and clinicians’ perspectives on the importance of anti-HLA antibodies in transplantation. However, there are many other factors that can affect the analysis and interpretation of this very sensitive assay, and we might need to find further ways to identify the clinically relevant DSA. The need for standardization of protocols for solid-phase assays remains urgent; meanwhile, although the use of solid-phase assays for quantification has not yet been cleared by the Food and Drug Administration, and their use might be confounded by issues affecting assay reproducibility (¹³), analysis of antibody strength to define a more appropriate positive mean fluorescence intensity (MFI) threshold may help define the clinically relevant DSA. In fact, Mizutani et al. (¹⁴) have shown that anti-HLA strength is an important factor to consider. Analyzing kidney transplant recipients’ sera, they found that the strength of antibodies in the sera from 39 patients who subsequently had allograft failure was markedly higher than that in the sera of 26 patients with continued good allograft function. The correlation of antibody strength with crossmatch positivity has also been discussed, and higher MFI thresholds have been shown to better predict donor-specific crossmatches (¹⁵).

Complement activation results in the production of several biologically active molecules that contribute to inflammation. The ability to bind and activate the classical complement pathway is another factor that can differentiate clinically relevant DSA. Although still controversial, the association of pretransplantation and posttransplantation alloantibody of the strong complement-binder subclasses (IgG1/IgG3) was shown to influence worse allograft outcomes (^{16, 17}). C1q is the first step in the classic complement cascade activated by antibody, and it precedes C4d deposition. Using the modified solid phase Luminex assay (One Lambda, Inc., Canoga Park, CA) that detects the complement-binding HLA antibody C1q assay, Tyan et al. were able to correlate complement binding with antibody-mediated rejection (¹⁸) and allograft loss (¹⁹).

Our study’s purpose was not only to emphasize the clinical relevance of HLA-DQ DSA in kidney transplantation. Most importantly, we wanted to further analyze the different properties, IgG isotype subclasses, C1q binding, and antibody strength that characterize the DSA that is ultimately deleterious to the kidney allograft.

RESULTS

Demographics

This analysis focused on three groups of first kidney transplantation recipients: those with only de novo DQ DSA (DQ-only) (n=34), with de novo DQ plus other DSAs (DQ+other) (n=20), and with no DSA (n=149) (see Figure S1, SDC, https://links.lww.com/TP/A792). The groups did not differ in gender, age at transplantation, donor age, or panel reactive antibody. African American (AA) ethnicity predominated in each group; however, the no-DSA group included 19% fewer AAs than the DQ-only group (P=0.04) and 36% fewer AAs than the DQ+other group (P=0.002). The DQ-only and DQ+other groups received their grafts mostly from deceased donors (65%, P=0.007; and 70%, P=0.01); only 39% of the no-DSA recipients had deceased-donor transplantations. The mean±SD number of HLA-A/B and HLA-DR mismatches was significantly higher in the DQ+other group than in the DQ-only and no-DSA groups: A/B mismatches, 3.35 (0.7) (DQ+other) versus 2.41 (1.1), P=0.0005, and 2 (1.3), P<0.0001; DR mismatches, 1.5 (0.6) versus 1.08 (0.6), P=0.03, and 0.9 (0.7), P=0.0006. The mean±SD number of DQ mismatches was similar with DQ-only (1.3 [0.4]) and DQ+other (1.3 [0.5]), but significantly lower in recipients with no DSA (0.8 [0.7], P<0.0001). Hypertension, as a cause of end-stage renal disease, and delayed allograft function were more frequent in DQ-only recipients than in those with no DSA (53% vs. 29%, P=0.007; and 12% vs. 3%, P=0.04). The mean±SD follow-up time was shorter with DQ+other than with no DSA (64 [33] months vs. 83 [37], P=0.02) (Table 1).

TABLE 1: Demographics

De Novo DQ DSA Subgroups: AR and Allograft Loss

The DQ DSA appeared later in the posttransplantation period in both DQ DSA groups. Significantly fewer recipients in the no-DSA group had one or more episodes of biopsy-proven AR episodes than in the DQ-only (19% vs. 41%, P=0.005) and DQ+other (19% vs. 55%, P=0.0009) groups (Table 2).

TABLE 2: Acute rejection and graft loss

More recipients in the DQ+other and DQ-only groups lost their allograft than in the no-DSA group (40%, P<0.0005; and 21%, P=0.054 vs. 8%). Time to allograft loss after initial de novo DQ DSA was similar for the two groups (Table 2). Multivariate analysis by Cox proportional hazard model for risk of allograft failure (adjusted for recipient’s ethnicity, HLA-A/B/DR/DQ mismatch, type of allograft, cause of end-stage renal disease, occurrence of delayed graft function, and study group) showed an independent increased risk of allograft loss for recipients in the DQ+other and DQ-only groups (hazards ratio [HR], 11.4; confidence interval, 2.9–45.9; P=0.001) and (HR, 3.7; confidence interval, 1.1–11.8; P=0.03) (see Table S1, SDC, https://links.lww.com/TP/A792).

IgG Subclasses/C1q: AR and Allograft Loss

Among all 59 DQ DSA detected, anti-DQ7 was the most prevalent (36%), and 47.5% of the recipients with anti-DQ7 DSA had both IgG1 and IgG3 subclasses in their sera (see Figure S2, SDC, https://links.lww.com/TP/A792). IgG1/IgG3 was either in association with both IgG2 and IgG4 or with IgG2 or IgG4 individually. A large percentage of recipients with AR had IgG1/IgG3 compared with those who had no AR if they had only DQ (80% vs. 29%, P=0.01) or had DQ+other (55% vs. 36%, P=0.6) (Table 3). The association of de novo DQ DSA IgG1/IgG3 subclasses and allograft loss was not significant, although 71% of the DQ-only recipients with allograft loss had IgG1/IgG3 DQ DSA subclasses (Table 3), and univariate analysis by Cox proportional hazard model showed an increased risk of allograft loss for recipients with an IgG3 DQ DSA (HR, 3.5 [1.3–9.5]; P=0.01) when the analysis included all 54 DQ DSA recipients (data not shown).

TABLE 3: HLA DQ-only and DQ+other DSA, IgG subclasses/Clq: acute rejection and allograft loss

C1q, an indirect marker of complement binding, was also significantly associated with AR. The incidence of C1q-positive DSA in recipients with DQ-only who had AR was extremely high compared with those DQ-only recipients who did not have AR (100% vs. 37%, P=0.001). The same phenomenon was observed in DQ+other recipients who had AR compared with DQ+other recipients who had no AR (100% vs. 54%, P=0.037) (Table 3). Although the small number of patients might account for the fact that association of allograft loss with C1q binding did not reach significance, the presence of C1q-positive DSA was also higher in recipients with allograft loss in DQ-only and DQ+other groups than with those with no allograft loss (Table 3).

DSA Strength

Interestingly, the DQ DSA of recipients who had AR and allograft loss had an average higher peak MFI than did recipients without AR or allograft loss. The difference was significant in DQ-only recipients who had AR versus those who had no rejection (16,085 MFI vs.10,045 MFI, P=0.01) and in those with DQ+other who had allograft loss versus those with no allograft loss (17,392 MFI vs. 8863 MFI, P=0.0004), suggesting that DSA strength might have a great impact on allograft outcome (Table 4).

TABLE 4: Peak donor-specific antibody strength: acute rejection and allograft loss

Death-Censored Allograft Survival

The log-rank test for survival analysis showed that recipients with DQ-only and DQ+other had lower 5-year death-censored allograft survival than recipients without DSA (Fig. 1). The presence of a C1q-binding de novo DQ was also associated with lower 5-year allograft survival (Fig. 2A). Furthermore, recipients who had DQ DSA with an MFI of 12,000 or greater had a significantly worse allograft survival than recipients with an MFI of less than 12,000 (P=0.002) (Fig. 2B).

FIGURE 1: De novo DQ donor-specific antibody (DSA), death-censored: allograft survival by group. Kaplan-Meier allograft survival of kidney transplant recipients with no DSA, DQ-only, and DQ+ other DSA. 254×190 mm.

FIGURE 2: De novo DQ donor-specific antibody (DSA), death-censored: allograft survival by C1q-binding and antibody strength. A, Kaplan-Meier allograft survival of kidney transplant recipients by DQ DSA C1q-binding properties. B, Kaplan-Meier allograft survival of kidney transplant recipients by DQ DSA strength. 254×190 mm.

DISCUSSION

In accordance with previous studies, we found a high prevalence of DQ DSA (30.6%) in the East Carolina University kidney transplantation population (see Figure S1, SDC,https://links.lww.com/TP/A792). Surprisingly, only 10 DQ DSAs (11%) were preformed. The presence of de novo DQ DSA alone or with other DSA was significantly associated with an increased risk of allograft loss. Recently, two groups (^{11, 12}) found a correlation between posttransplantation DQ DSA and worse allograft outcomes in kidney transplantation. Unlike us, Devos et al. (¹¹) did not find a significantly higher prevalence of AR or worse allograft survival in recipients with DQ alone. However, our median (range) follow-up time was over twice as long than in the Devos study (68 [9–146] vs. 26 [3–47] months), perhaps providing enough time for late persistent DQ DSA to develop and trigger AR and allograft loss. Wiebe et al. (²⁰), in a long follow-up analysis of posttransplantation DSA in kidney transplantation, also demonstrated that de novo DSA (68% of them class II alone) developed later after transplantation and were associated with clinical rejection. Later appearance of DSA after transplantation could be associated with nonadherence to immunosuppressive regimens, which is more apt to occur later, after transplantation (²¹), and although we have excluded recipients with reported nonadherence, we cannot rule out the possibility of nonadherence either unnoticed or unreported in patients’ charts.

No other study of de novo DQ DSA has analyzed their characteristics in depth. So, an important aim of this study was to investigate the further characteristics that differentiate harmful from innocuous DQ DSA, especially regarding antibody strength, ability to bind C1q, and the presence of complement-binding IgG subclasses. The analysis of all DQ DSA IgG subclasses in our study showed that IgG1 was the most prevalent of the four IgG subclasses, which accords with most analyses of IgG subclasses in kidney (^{16, 22}) and liver transplantation (¹⁷). IgG1 is most efficient in complement lysis, although IgG3 binds more C1q (²³). When combined in patient’s sera, these two subclasses may have a higher complement-binding activity than if they presented separately. The presence of IgG1/IgG3 occurred 51% more often in sera of recipients with AR than in sera of recipients without AR in the DQ-only group. Moreover, recipients who lost their allograft had 34% more IgG1/IgG3 than those in the DQ-only group who did not lose their allograft. Although the result was not significant, perhaps because of the small number of patients, other groups have reported an increased risk of allograft loss associated with IgG3, alone, in liver (¹⁷) and in kidney recipients (¹⁶). Furthermore, our study’s Cox proportional hazard model showed an increased risk of allograft loss for recipients with IgG3 DQ-only (HR, mean±SD, 3.5 [1.3–9.5], P=0.01) when analysis included all 54 DQ DSA recipients.

In accordance with our IgG subclasses data, C1q binding was also significantly associated with AR. All DQ-only recipients with AR had C1q-binding IgG compared with only 37% of those with no AR who had C1q-binding IgG (P=0.001). Our analysis confirmed recent data from Sutherland et al. (²⁴), who used the same C1q solid-phase assay we used, and, in a study of 193 pediatric kidney transplant recipients, found that, of the 35 with de novo DSA, 15 (43%) were positive for C1q (11 had class II DSA alone). Patients with C1q-binding DSA were significantly more likely to have AR and allograft loss than those without C1q-binding DSA. In our study, 86% of the DQ-only recipients with allograft loss had C1q-binding DSA compared with 48% of recipients without allograft loss (P=0.1). When DQ was associated with other DSA, 100% of the DQ-only recipients with allograft loss had C1q-binding DSA compared with 58% of recipients with no allograft loss (P=0.1). Although the results did not reach statistical significance, there was a trend toward allograft loss in the groups with DQ DSA. Furthermore, recipients with C1q-binding DQ IgGs had a 5-year allograft survival rate of 63%, whereas those without C1q-binding DQ IgGs had a significantly better rate of 93% (despite the possibility that, with longer follow-up analysis, recipients with non–complement-fixing DSAs could show poorer graft survival than those who had no DSA). Although recipients with DQ had an overall poor prognosis, analysis of both the IgG subclass and C1q provided additional discriminating information for recipients with DQ, helping to further characterize the harmful DQ DSA.

It has been shown that antibody strength, per se, is a good way to identify deleterious DSA (¹⁹). In the above study (²⁴), Sutherland et al. mentioned that recipients with C1q-binding DSA had significantly higher IgG MFI than those without C1q-binding DSA (mean±SD, 7885 [5175] vs. 1798 [1284]; P<0.001). Our study found the same trend: recipients with C1q-binding DQ-only DSA also had higher IgG DSA MFI than those without C1q-binding DQ-only DSA (mean±SD, 16,247 [5166] vs. 6215 [4530]; P<0.0001) (data not shown). The difference is that, in the Sutherland study, recipients with allograft loss did not have significantly higher MFIs by IgG or C1q testing. In our study, recipients with DQ-only DSA who had AR had significantly higher IgG DSA MFI than those with no AR (mean±SD, 16,085 [5597] vs. 10,045 [6826]; P=0.01), and recipients whose DQ DSA was associated with other DSA, and who had allograft loss, had significantly higher IgG DSA MFI than those with no allograft loss (mean±SD,17,392 [3768] vs. 8863 [4951]; P=0.0004).

By using data available from the dbMHC Anthropology database (www.ncbi.nlm.nih.gov/gv/mhc) to estimate HLA-DQ allele frequencies for each recipient’s allograft based on the donor’s ethnicity and HLA-DRB1, and on DQ alpha and DQ beta epitope tables by Deng et al. (²⁵) and N. El-Awar (unpublished data), we were able to look for antibody reactivity patterns in the HLA-DQ SAB results, which would suggest antibodies targeting epitopes on both the DQ alpha and the DQ beta chains, and we could correlate them with MFI strength. Interestingly, our analysis showed a surprising correlation between the presence of DSA against both DQ chains (alpha and beta) and MFI strength: 28/33 (85%) recipients with MFIs 10,000 or greater had DSA patterns that could suggest simultaneous reactions against both DQ chains, whereas only 2/21 (9.5%) of recipients with MFIs less than 10,000 exhibited those patterns. Tambur et al. (²⁶) have recently, in an elegant study, shown that both alpha and beta polymorphic extracellular domains of the HLA-DQ molecule may contribute to the antibody recognition site. In the current study, we have not just confirmed their findings but have also shown that the presence of DSA concomitantly targeting both DQ alpha and beta chains could add a deleterious impact to the allograft.

In conclusion, de novo DQ DSA can be extremely harmful for kidney transplant recipients. In our study, de novo DQ DSA increased the prevalence of allograft loss and reduced allograft survival. However, our further analysis showed that there are several relevant factors that need to be considered when interpreting de novo DQ DSA data: antibody strength, reactivity to the DQ alpha chain, and the ability to bind and activate complement—as measured by IgG subclasses and C1q assay—are highly important for identifying the DQ DSA ultimately harmful to the kidney allograft. Therefore, it is important not only to diagnose and monitor the presence of the DQ DSA after transplantation but also to define such relevant characteristics of those DSA to improve kidney allograft outcome.

MATERIALS AND METHODS

Patients and Sera Screening

We retrospectively screened 284 first kidney transplant recipients (performed between April 1999 and December 2007) at the Brody School of Medicine, East Carolina University, Greenville, NC, for the presence of anti-HLA antibodies. All patients were negative for lymphocyte crossmatch before transplantation, either complement-dependent cytotoxicity crossmatch (CDC/AHG-XM) or flow crossmatch. Sera samples were collected after transplantation at regular intervals. All samples were analyzed for the presence of anti-HLA using LABScreen mixed beads (One Lambda, Inc., Canoga Park, CA). When tested positive, sera were further tested using LABScreen class I and class II SAB (One Lambda, Inc.). Specificities with an MFI of 1000 or greater were considered positive. Recipients with preformed (pretransplantation) DQ DSA (n=10), or transient DQ DSA (DSA in only one serum sample after transplantation) (n=20) or were nonadherent (stated in the recipient’s chart) (n=7), were excluded from the study. Six recipients lost to follow-up (transferred to a different center or their sera was no longer collected) were also excluded (see Figure S1, SDC,https://links.lww.com/TP/A792). Of the 38 recipients with HLA-A, HLA-B, or HLA-DR, only 21 had persistent DSA, and were excluded, as were 11 recipients from living related donors with zero HLA mismatches. All serum samples had previously been obtained as a part of local standard care, and patients had given written informed consent to have sera drawn on a regular basis after transplantation. This study was approved by the institutional review board.

DQ DSA IgG Subclass and C1q Testing

De novo HLA-DQ DSA was defined as the presence of DSA only in posttransplantation sera. All recipients with HLA-DQ DSA in more than one posttransplantation serum sample had their peak DSA sera tested for IgG isotype subclasses (IgG1, IgG2, IgG3, and IgG4) and C1q. We used a modified assay in which the conventional phycoerythrin-conjugated, antihuman IgG was replaced with a phycoerythrin-conjugated, IgG subclass–specific, antihuman IgG (IgG1 clone 4E3, IgG2 clone HP6002, IgG3 clone HP6050, and IgG4 clone HP6025; Southern Biotech, Birmingham, AL). A single test for each IgG subclass was performed separately with LABScreen SABs with the same protocol used for the detection of anti-HLA IgG antibodies. A normalized trimmed MFI greater than 500 was defined as positive on the basis of binding patterns after validation and dilution experiments as previously reported (¹⁷). C1q binding was tested using C1qScreen (One Lambda, Inc.) according to the manufacturer’s directions. Specificities with MFI of 500 or greater were considered positive; the 500-MFI cutoff was based on previous publications (²⁷).

Biopsy Data and Allograft Loss

All biopsies were performed in the context of allograft dysfunction. AR episodes were categorized according to the Banff ’97 criteria (update 2003) (^{28, 29}). Allograft loss was defined as return to dialysis. Follow-up time was measured from the time of transplantation to the time of patients’ last sera collection/clinical visit.

HLA Typing

All recipients and donors included in this study were typed either by PCR sequence-specific primer methods, using A/B/DR/DQ trays, or by standard monoclonal (microcytotoxicity method) HLA typing trays (One Lambda, Inc.) for HLA-A, HLA-B, HLA-DR, and HLA-DQ. Recipient and donor typing were not performed for DQ alpha, DP, or C antigens.

Immunosuppression

All recipients received induction therapy, which consisted of monoclonal humanized anti-CD25 antibody (daclizumab) or polyclonal antibody (rabbit antithymocyte globulin). Maintenance immunosuppression consisted of a calcineurin inhibitor (cyclosporine or tacrolimus), prednisone, and mycophenolate mofetil.

Study Groups

• 1: De novo DQ-only DSA: posttransplantation DQ DSA, without other DSA (A/B/DR) (n=34)
• 2: De novo DQ+other DSA: posttransplantation DQ DSA, associated with other DSAs (A/B/DR) (n=20)
• 3: No DSA: no DSA (A, B, DR, or DQ) in any posttransplantation sera (n=149)

Data Analysis

Baseline characteristics were compared between groups by the t test; Pearson chi-square test or Fisher exact test was used for numerical or categorical variables. Descriptive statistics are presented as percentages, means, and standard deviations. All P values 0.05 or less were considered statistically significant. Allograft survival rates were estimated using the Kaplan-Meier method; statistical comparisons of survival curves were made by log-rank test. Multivariate analysis was performed by Cox proportional hazards regression. Statistical analysis used STATA/MP v. 10.0 (StataCorp, College Station, TX).

ACKNOWLEDGMENTS

The authors thank Rene Castro for assistance with the STATA statistical software and Nadim El-Awar for helpful discussions on HLA-DQ alpha and DQ beta epitopes.