This retrospective study included all patients who underwent a second kidney transplantation between 2010 and 2021 with preformed DSA against the second donor, at two French transplantation centers: Necker-Enfants Malades and Foch Hospitals. We considered DSA against HLA-A, -B, -C, -DRB1, -DRB3/4/5, -DQA1 and -DQB1 loci, but not DPB1 or DPA1, with a Mean Fluorescence Intensity (MFI) ≥ 500. We excluded the patients whose follow-up was less than 6 months, those without complete HLA typing of the first donor and the patients whose analysis of DSA eplet specificity could not identify any target. The study was done in accordance with the Declaration of Helsinki. The clinical and research activities being reported are consistent with the Principles of the Declaration of Istanbul as outlined in the “Declaration of Istanbul on Organ Trafficking and Transplant Tourism.” All patients provided written informed consent at the time of transplantation for anonymous collection of data for clinical research purposes.

Clinical data were retrospectively retrieved from patients’ medical records, the DIVAT database, ¹ and from the national database of kidney transplant recipients and donors (Cristal). ² The results of all graft biopsies, screening and for-cause biopsies, have been collected for all patients. Each biopsy was assessed by an experienced nephropathologist, according to the Banff classification available at the time of biopsy. The conclusions were then reanalyzed according to the 2022 Banff classification for the diagnosis of ABMR [30]. All patients were followed-up from the day of the transplantation until the occurrence of graft loss, death, or the date of the final data extraction (29 December 2023).

The method of HLA typing depended on the transplantation period. For recipients before 2017 and all deceased donors, HLA typing was performed using polymerase chain reaction sequence-specific primers (Olerup SSP) and the results were provided with two-digits for HLA-A, -B, -DRB1, and -DQB1. For recipients since 2017 and living donors, HLA typing was done with sequence-specific oligonucleotide technology (PCR-SSO One Lambda) or next-generation sequencing for the same loci, and HLA-Cw and -DQA1. In the case where the typing provided a list of ambiguities, the most frequent allele was retained based on the most frequent haplotype. To complete loci without specific typing or for some ancient donors with no DNA remaining, we imputed the most likely high-resolution allelic typing (four-digit) from the two-digit resolution typing using the HaploSFHI tool. This algorithm was developed from data from a majority of the France histocompatibility laboratories and validated on European cohorts [31].

Patients on the waiting list were systematically screened for anti-HLA antibodies before transplantation every 3 months. All sera were analyzed in the same laboratory (Immunology Laboratory of Saint-Louis Hospital, Paris) using the Luminex Labscreen screening assay, and if positive completed with the Single Antigen assay (both from One Lambda ThermoFisher Scientific, West Hills, CA). Loci tested were HLA-A, -B, -Cw, -DRB1, -DRB3/4/5, -DQA1, and -DQB1.

Preformed DSA were defined with an MFI above 500 after subtracting the background noise and the autoreactivity. The immunodominant DSA was defined as the DSA with the highest MFI value over all sera collected during the pretransplant period. De novo DSA were detected during the post-transplant course but not before transplantation.

HLA compatibility between the recipient and the second donor was assessed by the number of mismatches at the antigenic (assimilated to serological group or first-field DNA typing), allelic (assimilated to second-field DNA typing), and eplet level for HLA-A, -B, -Cw and HLA-DRB1, -DRB3/4/5, DQB1 and DQA1 (only at the allelic and eplet level). For the latter, we determined the mismatched eplets expressed by the first and the second donor but not by the recipient, from the eplet database of the HLAMatchmaker Antibody Analysis software (version 3.1). ³ Mismatched eplets shared by the two donors were identified as eplet repeated mismatches. We focused on antibody-verified eplets, i.e., eplets verified by analyzing reactivity patterns of either polyclonal sera or monoclonal antibodies and updated in the HLA Eplet Registry. ⁴

We used the HLAMatchmaker Antibody Analysis software to determine the targeted eplets by preformed DSA. We entered donor and recipient allelic typing into the software, the MFI values of all Luminex beads from the pre-transplant sera. The cut-off for MFI positivity was set at a minimum of 500 and sometimes higher, based on the average value of self-antigen beads.

Eplets expressed by the HLA molecules of the recipient and of the negative beads were excluded as potential candidates for targeted eplets. We determined which eplets expressed by the donor antigens targeted by DSA could explain the antibody reactivity pattern. When these eplets were repeated between the first and the second donor, the DSA was referred as “preformed DSA targeting a Repeated antibody-verified Eplet MisMatch (DREMM).”

Patient data were summarized using frequencies and percentages for categorical variables, and medians with interquartile ranges or means with standard deviations for continuous variables. We used the Wilcoxon rank-sum test to compare numeric data, and chi-square or Fisher’s exact tests for categorical data. The cumulative incidence of ABMR and non-death-censored graft loss considered as competitive risks were determined with the Aalen-Johansen estimator and compared in patients with and without DREMM (Package Survival).

Graft and patient survival analyses were performed using Kaplan‒Meier estimates and were compared between patients with or without DREMM using the log-rank test. Univariate and multivariate Cox proportional hazards models were applied using hazard ratios to identify immunological factors associated with the risk of ABMR and graft loss from the day of transplant. Multicollinearity between covariates was tested when necessary, with the package “Car”. All statistical analyses were performed using R software version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided p-value of <0.05 was considered statistically significant.

We screened 69 recipients who underwent a second kidney transplantation between 2010 and 2021 with preformed DSA and after applying exclusion criteria, we included 45 sensitized patients with a median follow-up of 5.1 years (Supplementary Figure S1). The first transplantation had been performed between 1982 and 2015, and patients received their second transplant after a median delay of 12 years. All recipients had preformed DSA explaining the use of thymoglobulin as induction therapy in 89% of patients. The patients were primarily on a triple maintenance immunosuppressive regimen including steroids, calcineurin inhibitor – mainly tacrolimus – and mycophenolic acid. Only one patient received a steroid-free regimen. Depending on local habits and characteristics of DSA, the immunosuppressive induction treatment for the second transplant also included, except for one recipient, at least one of the following: intravenous immunoglobulins, plasma exchanges, rituximab or eculizumab (Table 1).

The mean numbers of mismatches at the antigenic, allelic, and eplet (antibody-verified) level were 5.1 ± 2.3 (min-max: 1–10), 7.3 ± 3.2 [1–10] and 14.2 ± 7.2 [4–30], respectively (Supplementary Table S1). Whatever the level, the number of mismatches was lower for class II compared to class I. In total, 18 (40%) patients had no antibody-verified class II eplet mismatches, compared to only 1 (2%) patient for class I.

Subsequently, we focused on the identification of repeated mismatches (RMM) at these three levels between the first and the second donor (Table 2). RMM were more frequently evidenced at the eplet level since 10 (22%) recipients were exposed to an antigenic RMM, 13 (29%) to an allelic RMM and 39 (87%) to an eplet RMM during the second transplant. In addition, the eplet level allowed a finer stratification of mismatches with a wider range of RMM observed per patient: [0–13] at the eplet level versus [0–2] at the antigenic level and [0–4] at the allelic level. Class I antigenic and allelic RMM only concerned the HLA-Cw locus whereas eplet RMM involved the three class I loci.

Figure 1 illustrates an example of the three levels of class I compatibility for one recipient and his two consecutive donors, and the analysis of the eplet reactivity of the preformed DSA against the second donor. At the antigenic and allelic levels, there was no RMM. Antibody-verified eplet load was higher towards the second donor (6 versus 5) despite fewer antigenic mismatches, highlighting the inconstant correlation between antigenic and eplet levels. Although there was no antigenic or allelic RMM, an antibody-verified eplet RMM (73TVS) was present due to the shared expression of this eplet by HLA B46 and Cw1 (1st donor), and Cw10 (2nd donor) molecules. The recipient had a preformed anti-Cw10 DSA. Analyzing the Luminex Single Antigen bead profile with HLAMatchmaker showed that this anti-Cw10 DSA was directed against the 73TVS antibody-verified eplet RMM.

The study population was divided into two groups based on the presence or absence of preformed DSA targeting a Repeated antibody-verified Eplet MisMatch (DREMM). Twenty-seven (60%) recipients presented a DREMM whereas 18 (40%) did not. No difference was observed between the two groups regarding the class or locus of the highest preformed DSA, which was mainly class I (69%) (Table 3). The interval between the two transplants and the frequency of transplantectomy were similar in the two groups. The DREMM was isolated in 15 of the 27 patients (56%) or associated to one or more other DREMM in the 12 remaining patients (44%). The DREMM was the immunodominant DSA in 22 of the 27 cases. In the peak serum, i.e., the serum with the DSA of highest MFI between the two transplants, the MFI of the highest DSA and the sum of DSA MFI were significantly higher for DREMM compared to non-DREMM (median: 8,326 versus 1,676, P = 0.01 and 14,249 versus 2,482, P < 0.001, respectively).

The post-transplant evolution of the DSA also differed significantly depending on whether it targeted an eplet RMM or not (Table 3 and Figure 2). A significantly higher proportion of DREMM persisted at three but not at 12 months compared to non-DREMM [73% versus 39% (at month −3), P = 0.02, 54% versus 28% (at month −12), P = 0.09]. More DREMM experienced a significant increase (greater than 500 units) of their MFI between day-0 and month −3 compared to no-DREMM but this difference was not significant (31% versus 5%, P = 0.06) (Figure 2).

De novo DSA appeared in 3 (7%) patients, all in the DREMM group, within a median (IQR) time of 19 months (17.8–43.1). Only one of them experienced an ABMR episode after the occurrence of this de novo DSA.

Among the 45 patients with preformed DSA, 24 (53%) patients developed an active ABMR during their follow-up. The presence of DSA directed against an antigenic RMM was not associated with an increased risk of ABMR (57% versus 43%; P = 1). However, patients with DREMM had significantly higher rates of ABMR during follow-up compared to patients without DREMM [19/27 (70%) versus 5/18 (28%); P = 0.005] (Table 4). The increase in the risk of ABMR associated with DREMM was statistically significant when these DSA targeted a class I eplet RMM but not a class II eplet RMM. We compared the cumulative incidence of ABMR and of overall graft loss as competitive risks in patients with and without DREMM (Figure 3). These graft losses, occurring before any ABMR episode, mainly consisted in the death of patients (n = 7) and in death-censored graft loss in one patient. Patients with DREMM had a higher cumulative incidence of ABMR (HR 3.37 [1.25–9.06], P = 0.018), but not of overall graft loss (HR 2.08 [0.85–5.1], P = 0.2).

Regarding loci, all patients with DSA against a B and DQ antibody-verified eplet RMM (9 and 6 patients, respectively), experienced ABMR. In comparison, rates of ABMR with DREMM against A, Cw, and DR were 71%, 64% and 57%, respectively.

ABMR characteristics did not differ between the 2 groups, regarding the proportion of sub-clinical ABMR (29%) and the post-transplant time of ABMR diagnosis (median 2.4 months) (Table 4). Three patients with ABMR did not receive treatment for this episode of rejection. One patient was included in an Eculizumab trial and two patients were not categorized as ABMR according to the Banff classification used at that time. Noteworthily, there was no association between the risk of ABMR and any level of HLA incompatibility, antigenic, allelic or eplet, whether we considered repeated mismatches or not (Supplementary Table S2).

In univariate Cox model, factors associated with the survival without ABMR were the maximal MFI of the immunodominant DSA, the sum of the DSA MFI, and the presence of a DREMM (combined class I and II or class I but not class II alone) (Table 5). The number of mismatches and the presence of RMM, whatever the antigenic or eplet level, were not associated with an increased risk of ABMR. In the multivariate Cox model combining the MFI of DSA (either maximal value or sum of DSA MFI) and the presence of a DREMM, none of these factors was independently associated with the risk of ABMR. Given the association between DREMM and MFI of DSA (Table 3), we tested the presence of multicollinearity between these covariates, which was negative.

During the median follow-up of 61 months, 8 death-censored graft losses occurred, all in the DREMM group, attributed to rejection in all but one case. Therefore, the death-censored graft survival was lower in patients with DREMM (log-rank test: P = 0.02) (Table 4; Figure 4). No difference was observed in overall graft survival or in patient survival (Supplementary Figure S2).