Eplet mismatch analysis and allograft outcome across racially diverse groups in a pediatric transplant cohort: a single-center analysis

This is a retrospective cohort study of pediatric (≤ 21 years old) kidney transplant recipients followed at the Johns Hopkins Children Center, who received a first kidney transplant between January 2006 and April 2017. In this group, 92 patients were transplanted at Johns Hopkins Hospital and 18 patients were transplanted at other centers prior to follow-up at Johns Hopkins. Patient and donor demographics and clinical information were retrieved from the electronic patient information records (EMR) under an approved IRB protocol.

Donor selection for pediatric candidates listed for transplantation at Johns Hopkins includes an evaluation of the level of HLA mismatch between donor and recipient with the goal of limiting the incidence of antibody development. Transplant candidates are discussed during quarterly meetings between clinicians and the histocompatibility laboratory. Thresholds for acceptable HLA mismatches with a deceased or living donor are established for candidates based on each patient’s clinical characteristics, including the urgency for transplantation, sensitization against HLA antibody (calculated PRA), and the frequency of the patient’s HLA antigens compared with the donor pool. Candidates with several potential living donors who are medically suitable for donation may be paired with the best HLA-matched donor.

All patients transplanted at Johns Hopkins received induction therapy. Induction treatment typically consisted of Thymoglobulin (1.5 mg/kg/day for 5 days; n = 79); in a few cases, Daclizumab (1 mg/kg/day for 5 days; n = 7), Basiliximab (20 mg at time of transplant and 20 mg on post-operative day 4; n = 6) all with methylprednisolone 10 mg/kg delivered at the time of transplant (day 0), followed by 8 mg/kg on day 1, 6 mg/kg on day 2, 4 mg/kg on day 3, 2 mg/kg on day 4, and 1 mg/kg on day 5. Three patients transplanted at other institutions were given Alemtuzumab (standard dose 30 mg at time of transplant) as induction. Maintenance treatment consisted of either steroid free or steroid minimization (0.2 mg/kg/day or qod), mycophenolate mofetil (initially 1200 mg/m²/day, with decrease to 600 mg/m²/day when tacrolimus therapeutic), and tacrolimus (goal serum level of 7–10 ng/mL). Indications for steroid minimization versus steroid free included underlying immunologic disease and/or presence of HLA antibody.

HLA-A, HLA-B, HLA-C, HLA-DR, HLA-DQ, and HLA-DP typing were performed by reverse sequence-specific oligonucleotide probe (SSO) assay (One Lambda LABType®) and HLA antigen equivalents were reported for donors and patients who were transplanted at Johns Hopkins. HLA-typing data from transplants not performed at Johns Hopkins were retrieved from patients’ archived medical records stored in Johns Hopkins EMR or UNOS records. HLA mismatches between recipients and their donor were determined at the HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DQB1, and HLA-DPB1 loci and were assigned as 0, 1, or 2 antigen mismatches for each locus. HLA antibody specificities were identified using multianalyte bead-based assays performed on the Luminex® platform (Immucor-Lifecodes, Stamford, CT and One Lambda, Canoga Park, CA). Based on the single antigen beads, levels of antibodies against HLA-A, HLA-B, and HLA-DR were reported as cytotoxic positive for MFI values > 10,000, flow cytometric positive (FCXM+) for MFI values 4000 to < 10,000, or FCXM- Luminex+ (Lum+) for MFI values 2000 to < 4000. Antibodies against HLA-C, HLA-DQ, and HLA-DP were reported for MFI values > 20,000 as CDC+, 16,000 as FCXM+, and 4000 as Lum+. MFI below these values were evaluated based on reactivity patterns on several assays and those that lacked specificity patterns were reported as negative. Cytotoxic crossmatch tests were performed prior to transplantation, with positively selected T and B lymphocyte targets [21]. FCXM tests were performed as needed and as previously described [22] and were acquired on BD FACSCanto II using FACSDIVA software (BD Bioscience, Franklin Lakes, NJ). Removal of interfering substances in patients’ sera, including IgM autoantibodies and IgG immune complexes, was done by hypotonic dialysis [23]. Post-transplantation, HLA antibody measurements were performed in cases where pre-transplant HLA-DSA was present, if serum creatinine rose greater than 20% above the baseline, or if a biopsy indicated dysfunction. A biopsy indicating dysfunction is defined as injury on biopsy with a diagnosis of rejection (cellular mediated, antibody mediated), transplant glomerulopathy, or if a biopsy showed evidence of viral infiltration. Antibodies reported as de novo HLA-DSA were > 1000 MFI and confirmed using multiple assays.

The HLA Matchmaker software (HLA-ABC version 02, update June 2016 and HLA-DRDQDP (includes DRβ1/3/4/5, DQβ1 and DPβ1), version 2.1, update January 2017) was used to identify recipient-donor mismatches at the eplet level [24, 25]. Evaluation of eplet mismatching was done using the most common allele when an intermediate resolution typing was available for both recipient and donor. For patients and donors with only serologic level typing, the HaploStat application (https://​www.​haplostats.​org) was used to estimate the most likely allele based on the race of the patient and the donor [26, 27]. To further confirm that there was no significant difference between the first and second allele in the string, both were evaluated for eplet load. The total number of eplets for each locus was counted and included antibody verified and non-verified eplets. Data obtained with the HLA Matchmaker software was further verified using One Lambda’s HLA Fusion version 4.2 software which features integration of an HLA Matchmaker module developed in partnership with Dr. Rene Duquesnoy.

The clinical outcomes were obtained from chart review. Rejections were diagnosed by biopsy performed for creatinine rise, proteinuria, or hematuria and included a review of all clinically indicated allograft biopsies performed during the follow-up time for each patient. Biopsies were graded according to BANF 2009–2013 criteria [28‐31]. Graft loss was defined as return to dialysis maintenance or requirement for repeat transplant. Recurrence of disease was defined as any post-transplant recurrent diagnosis of a pre-transplant condition. Viral infection was defined as any diagnosis of EBV or CMV infection reported post-transplant based on clinical findings in the presence of viremia. BK infection was included only if confirmed by findings on biopsy. Medication non-adherence was considered present if concerns about adherence were included in documentation by the members of the health care team.

Analyses were performed in GraphPad Prism (version 6) and Stata/SE 14.1 for Windows (College Station, Texas). Continuous variables were compared using Student’s t test or Wilcoxon rank-sum test as appropriate and categorical variables were compared using the χ2 test and/or Fisher’s exact test. The relative risks for each outcome based on the number of eplet mismatches at different HLA loci were determined for the entire cohort and for race match status between donor and recipient (SRT vs DRT) using unadjusted and adjusted models. Because there were more than 10% for all tested outcomes, modified Poisson regressions were used to compare interactions [32]. Effect measure modification of eplet mismatch was examined by testing the interactions between eplet mismatch and recipient and donor race match status for each outcome. We adjusted for donor and recipient age (age and age-squared to account for the nonlinearity), gender, donor source, kidney allocation period, recipient diagnosis, and CPRA. A complete case analysis was done, in which observations with missing information on any covariates were excluded from the analysis and p values less than 0.05 were considered statistically significant.

Of 155 pediatric kidney transplant patients followed at the Comprehensive Transplant Center at Johns Hopkins between January 2006 and July 2017, 113 patients were first transplant recipients. Three patients for which complete donor HLA typing information was missing were excluded from the study. The characteristics of the remaining 110 first kidney transplant recipients are summarized in Table 1. The mean follow-up time was 5.8 years (0–11 years). The median age at time of transplantation was 13 years (2–21 years old). The transplanted cohort consisted of 60% male and 52% Caucasian recipients. Pre-transplant HLA antibody levels were missing for the patients transplanted at other centers. The majority of patients with available pre-transplant HLA antibody tests (79%) were negative for HLA antibody prior to transplantation and only 5% were transplanted across a Luminex + antibody directed against a donor antigen (HLA-DSA). Overall, there were slightly more living donor (55%) compared with deceased donor (45%) transplants. The number of living-related versus living-unrelated donors was 45(74%) and 16 (26%), respectively. Donors were mostly Caucasian (61%) and male (52%), ages 10 to 49 years old. Despite the reported decrease in kidney donation from living donors after the enactment of Share 35 in 2005 nationally [5], of 98 transplants performed in this cohort, between 2006 and 2014, 56% of the organs were from living donors. The number of deceased donor transplants did not increase significantly during 15 months (January 2015 and April 2017) after the implementation of the new KAS in December 2014 (44% versus 50% for pre and post KAS, respectively; p = 0.6).

We assessed antigen mismatches by donor source and recipient race based on low-resolution HLA typing. HLA-A, HLA-B, and HLA-DR typing were available for all patients. HLA-C, HLA-DQ, and HLA-DP typing were missing for 5 of 110 (4.5%), 2 of 110 (1.8%), and 28 of 110 (25%) patient/donor pairs. As shown in Table 2, Caucasian recipients had significantly fewer HLA class I mismatches with their donor compared with non-Caucasian patients (p = 0.006), but there was no significant difference in HLA-class II antigen mismatches between the racial groups (p = 0.126). Patients who received an organ from a deceased donor had significantly more mismatches with their donors at all loci compared with patients who received an organ from a live donor (p < 0.001; Table 2).

Eplet mismatch analysis could be performed for 105 of the 110 patients for HLA class I (A, B, C), 82 of the 110 patients for HLA class II (DRβ1/3/4/5, DQβ1, DPβ1), 110 patients for HLA-DRβ1/3/4/5 only, and 82 patients for HLA-DQβ1 only. Similar to the antigen level mismatches, there were significantly fewer numbers of HLA class I eplet mismatches between Caucasian recipients and their donors compared with all other racial groups (p = 0.028; Table 3). There was no significant difference in HLA-DRβ1/3/4/5 (p = 0.456) or DQα1/DQβ1 (p = 0.397) eplet load based on recipient race. The eplet mismatch load was significantly higher at all loci for recipients of a deceased donor kidney compared with those who were transplanted with a living donor (Table 3). A significant difference was also noted between recipients of a living-related versus living-unrelated donor for class I eplet load (p < 0.001), but not for class II eplet load (p = 0.052) (Table 3).

The cohort was then grouped into recipients who received a transplant from a donor with same race (SRT; n = 70) and recipients who received a transplant from a donor of a different race (DRT; n = 29) (Table 4). There were more Caucasian recipients in the SRT group compared with the DRT group (67% versus 14%; p < 0.001). There were more transplants with deceased donors in the DRT group compared with the SRT group (65% versus 30%) and fewer transplants with a living-related donor in the DRT group (21% versus 56%; p = 0.002). Consequently, the DRT group had higher eplet mismatched loads compared with the SRT group for class I (includes HLA-A,-B-C; mean difference in eplet load = 9; 95% CI 2–15; p = 0.007) and class II (includes HLA-DRβ1/DRβ345/DQβ1/DPβ; mean difference in eplet load = 8; 95% CI 1–16; p = 0.029) (Fig. 2). Interestingly, there was no difference in the mean eplet load when considering only mismatches for HLA-DRβ1 (p = − .573) (Fig. 2).

Post-transplant HLA antibody assessments were performed for 86 of the 110 (78%) transplanted patients. By end of follow-up, 59 of 86 patients (68%) were sensitized against HLA antibody. Moreover, 48 of the 86 patients (56%) had detectable antibodies directed against one or more donor antigens (de novo HLA-DSA). Of those who developed de novo HLA-DSA, 14 patients (27%) had infections and immunosuppression was reduced for 4 patients (8%) due to infection; medication non-adherence was reported for 10 patients (19%), and 2 (4%) had disease recurrences. Since sera were tested only at time of dysfunction, it was not possible to accurately determine when HLA-DSA developed; however, the time from transplantation to first detection of HLA-DSA ranged between 5 and 135 months (median time 50 months). The majority of de novo HLA-DSA was against HLA class II (47 were positive for HLA class II and 20 were positive for both HLA class I and class II antibody). There were more antibodies against HLA-DR and HLA-DQ antigens compared with all other antigens (p < 0.001). In an unadjusted model, the risk for development of de novo DSA was statistically significantly higher with greater total eplet loads for DR and DQ but not HLA-A, HLA-B, and HLA-C (Table 5). When only considering antibody verified eplets, the risk of developing de novo DSA was increased for all loci (supplemental Table 6A). After adjusting for all confounders, the risk of de novo HLA-DSA with increased eplet loads for class II remained statistically significant (RR = 1.02; 95% CI 1.00–1.03; p = 0.01). Similarly, the relative risk of developing de novo DSA was higher with increased HLA antigen mismatches (supplemental Table 6B).

The incidence of de novo DSA development was not significantly different between SRT and DRT groups (44% versus 48%; p = 0.7; Table 4). There was no significant difference in class I and class II eplet mismatch load between DRT with de novo DSA (n = 12) compared with SRT with de novo DSA (n = 35) (supplemental Table 7). After adjusting for all confounders, the association between eplet mismatch for class I and class II molecules and de novo DSA development was also not significantly different between DRT and SRT group for class I (RR = 0.99; 95% CI 0.95–1.03; p = 0.7) and for class II (RR = 0.98; 95% CI 0.96–1.01; p = 0.4).

The incidence of rejection in the entire cohort was 38% (n = 42) and included CMR and AMR. The relative risk of rejection was not significantly higher as the number of total, antibody verified eplet loads and HLA-mismatched antigens increased (Table 5 and supplemental Table 6). The risk of rejection for the DRT group became significantly higher compared with the SRT group when the class I mismatched eplet load rose to ≥ 70 (RR = 1.05; 95% CI 1.01–1.08_;p = 0.004; Fig. 3). The risk of rejection was not significantly increased for the DRT group compared with the SRT group with greater total eplet load for class II (RR = 0.94; 95% 0.88–1.01; p = 0.1).

Graft loss was reported for 25 patients (23%) in this cohort; 22 of the 25 patients had de novo DSA, 14 (56%) had rejection prior to graft loss, and 5 patients (20%) had reports of medication non-adherence. As shown in Tables 4 and 5, we found no increased risk of graft loss based on total or antibody verified eplet loads at all loci (RR = 1; 95% CI 0.99–1.01; p = 0.7). Similarly, we found no association between graft loss and total or antibody verified eplet load differences in SRT versus DRT groups for all loci. There was also no correlation between antigen mismatch and graft loss.