Clinical risk stratification of paediatric renal transplant recipients using C1q and C3d fixing of de novo donor-specific antibodies

Patients who tested positive for DSA (DSA+) were identified from our previously published single-centre cohort study [4]. In brief, all renal transplant recipients from 1 January 2006 (existing and new transplants after this date) were screened prospectively (1–3, 6, 12 months post-transplant and annually thereafter) using OneLambda assays (One Lambda, Canoga Park, CA) and pan-immunoglobulin G (IgG) secondary antibody. The cumulative frequency of the antibody tests was 60, 86 and 98% at 3, 6 and 12 months, respectively. All sera were heat inactivated in a water bath at 56 °C for 30 min to alleviate prozone effects. No threshold MFI for positive DSA was set as an a priori criteria.

In this study, the first DSA-positive serum was further tested for complement binding capabilities (Fig. 1a). Clinical characteristics were as previously described. Follow-up estimated glomerular filtration rate (eGFR, calculated using Schwartz formula) data were extended until April 2015. Immunosuppression data were obtained at the time of DSA detection. Histological classification was based on the Banff 2009 criteria.

Experiments were performed by researchers blinded to patient information at the Clinical Transplantation Laboratory, Viapath, Guy’s Hospital, London, in a single run and using assays from the same batch. Patients were defined as ‘complement positive’ if at least one DSA showed complement fixing.

C1q-binding DSA were identified using C1qScreen™ (One Lambda) according to the manufacturer’s protocol [5]. Sera were pre-treated with heat inactivation of the complement system as part of the protocol. Analysis was performed using the HLA Fusion 2.0 software (One Lambda). Complement positivity was assigned at >1000 MFI based on the negative control sera and internal negative control beads.

C3d-binding DSA were identified using Lifecodes C3d and Single Antigen assay (Immucor, London, UK) according to the manufacturer’s protocol [6]. In addition, sera were tested for pan-IgG using Lifecodes Single Antigen kits (Immucor). Results were analysed using the same manufacturer’s MatchIt software, and complement positivity was defined as per the software algorithm.

Data were presented as the mean ± standard error of the mean) and as the median with the interquartile range (IQR), as appropriate. Comparisons between groups were performed with the Mann–Whitney test, and comparisons of proportions were performed using the Fischer and chi-square tests. Correlation between IgG MFI and complement positivity was estimated using logistic regression. Event-free survival was estimated with the Kaplan–Meier method and was compared between risk groups using the log-rank test. The primary event was defined as a sustained 50% reduction (defined as two consecutive results at least 3 months apart) from baseline eGFR as per the previous study results, and patients were censored at the end of the follow-up period [3]. Statistical analysis was performed using GraphPad Prism 5.0 (GraphPad Software Inc., LaJolla, CA), with p values of <0.05 considered to be significant. Multilevel linear modelling was performed using ‘nlme’ on the R statistical platform [7]. The model used individual patient nesting and fixed effect of follow-up time as described in our previous study [4]. The associations investigated were IgG MFI α C3d MFI and eGFR α C3d MFI over time.

In our original cohort, 215 patients underwent prospective screening for HLA antibodies (at 1–3, 6, 12 months and annually thereafter), using the LABScreen Mixed screening tool followed by screening with the Single Antigen Beads (SAB) assay from OneLambda. Of these 215 patients, 75 tested positive for IgG DSA at a median time of 0.25 years post-transplant. Serum samples for 65 of these 75 patients were available for further testing, and the first positive DSA sample was tested using the C1q and C3d assays (Fig. 1a); serum samples for the remaining ten patients were unavailable due to these patients, and their sera, transferring to adult centres out of the region. This latter group of ten patients represented an older group (median age of transplant 13.7 years) with more cellular rejection [Electronic Supplementary Material (ESM) Table 1], of whom six met the primary outcome of 50% reduction from baseline eGFR. Using the C1q assay, 32 of the 65 (49%) patients included in the study tested positive for the following DSA: HLA-A (n = 5), HLA-B (n = 8), HLA-C (n = 2), HLA-DQ (n = 22) and HLA-DR (n = 4). Using the C3d assay, 23 of the 65 (35%) patients tested positive for the following DSA: HLA-A (n = 3), HLA-B (n = 5), HLA-DQ (n = 14) and HLA-DR (n = 4). The breakdown of HLA types according to the C1q and C3d assays was similar to that according to the whole DSA+ cohort assay (Table 1). Serum samples tested later post-transplant showed a tendency towards positive complement (C1q+/C3d+) results, although this trend was not statistically significant [Table 2 (median): 2.6 (C1q+) vs. 0.4 (C1q−) years, p = 0.17; 2.3 (C3d+) vs. 0.4 (C3d−) years, p = 0.28]. Higher total IgG MFI was observed for patients who had C1q+ DSA compared to those who had C1q− (mean ± SEM 4968 ± 1492 vs. 3006 ± 607, p < 0.005; Fig. 2) based on the original pan-IgG antibody identification. Nonetheless, there was a large overlap between MFI values and a poor correlation between IgG MFI and C1q results (adjusted R ² = 0.072). Similarly, higher total IgG MFI was observed for patients with C3d+ DSA than for those with C3d− DSA (9483 ± 2289 vs. 4184 ± 648, p < 0.005) on the original OneLambda pan-IgG antibody identification; however, there was a poor correlation between IgG MFI and the C3d results (adjusted R ² = 0.11). There was no difference between total IgG MFI in C1q− and C3d− patients, or in C1q+ and C3d+ patients. Therefore, it would appear that IgG MFI is not a significant predictor of DSA complement binding capabilities.

Comparison of the C1q and C3d results revealed that 19 patients were C1q+/C3d+, 13 patients were C1q+/C3d− and four patients were C3d+/C1q− (Fig. 1b). Therefore, a large proportion of patients with positive C1q binding did not show concomitant C3d binding (13/32, 41%). The breakdown of individual DSA classes are as shown in Table 1. There was a better concordance between C1q and C3d binding for Class II DSA. As the single antigen beads differ due the different manufacturers of the C1q and C3d assays, all sera were retested for DSA IgG binding using the SAB assay from the C3d manufacturer in order to confirm antibody detection and to rule out the detection of potential false positives due to the differing manufacturing methods between the kits. For C1q+/C3d− patients, 17/22 (77%) DSA IgG specificities were detected using the assays of both manufacturers. For C1q−/C3d+ patients, 3/7 (43%) DSA IgG specificities were detected using the assays of both manufacturers. Of note, there were no samples which were positive for complement binding and negative for DSA IgG detection, suggesting that no non-IgG DSA were detected.

The clinical characteristics of patients grouped according to complement binding are shown in Table 1. There were no differences in age at transplantation, donor type and number of mismatches. In terms of C1q, there were more boys and fewer congenital anomalies of the kidney and urinary tract (CAKUT) in the C1q− group than in the C1q+ group. In terms of immunosuppression at the time of DSA detection, roughly half of patients were on dual therapy consisting of prednisolone and either tacrolimus or mycophenolate mofetil (MMF). Patients were switched to MMF to minimise calcineurin inhibitor toxicity, which is in line with standard clinical practice at that time [8, 9]. Taking into consideration the small numbers in the different groups, the differences in medications were statistically significant in the patients tested for C3d. C3d+ patients were more likely to be on MMF than on tacrolimus when on dual therapy [6/8 (75%) C3d+ vs. 7/21 (33%) C3d−; p < 0.05]. The proportion of patients on azathioprine was higher in the C3d− group than in the C3d+ group, compared to MMF for patients on triple therapy [3/7 (48%) C3d+ vs. 15/20 (75%) C3d−; p < 0.05].

The primary outcome was a 50% reduction in eGFR, which was used as a surrogate marker of long-term renal allograft survival (Fig. 3). Patients who were DSA− from the previous study were used in the present study for comparative purposes. There was no difference in eGFR decline between C1q+ and C1q− patients [median time to primary event 5.9 vs. 6.4 years, respectively; p = 0.58; hazards ratio (HR) 0.74; 95% confidence interval (CI) 0.30–1.81]. On the other hand, C3d+ patients had a significantly faster eGFR decline than C3d− patients (5.6 vs. 6.5 years, p = 0.04; HR 0.38; 95% CI 0.15–0.97). Combining the C1q and C3d results did not improve the significance of the complement binding assays (C1q+/C3d+ vs. C1q−/C3d−: 5.2 vs. 6.6 years respectively, p = 0.09; HR 0.42; 95% CI 0.15–1.14) (ESM Fig. 1). Counter-intuitively, single binding of only either C1q or C3d (C1q+/C3d− and C1q−/C3d+) did not adversely affect renal allograft function.

Histology findings based on complement binding results are shown in Table 3. C1q+ patients were associated with increased episodes of tubulitis [0.75 ± 0.18 (C1q+) vs. 0.25 ± 0.08 (C1q−) episodes per patient; p < 0.05]. Patients who were positive for complement binding showed a higher proportion of C4d binding on biopsies, which reached statistical significance for the C3d results [48% (C3d+) vs. 20% (C3d−); p < 0.05). There was no difference in AMR (composite of glomerulitis, pericapillaritis and glomerular double contouring) or presence of CD20 aggregates.

Based on the better correlation with outcome obtained with the C3d assay, testing for C3d was extended to all DSA+ sera. Of the 65 patients enrolled in the study, 33 had multiple DSA+ sera available for testing (median 3, IQR 3–4 sera per patient). Ten patients remained C3d−; four patients converted from negative to positive; 12 patients were C3d+ throughout; four patients were intermittently positive; three patients were positive at the start, then became negative. The latter three patients had low C3d MFI of 1979 (HLA-B), 1750 (HLA-DR) and 1386 (HLA-DQ). Four patients received intravenous rituximab—three for chronic AMR and one for post-transplant lymphoproliferative disorder. C3d remained positive in three patients who received intravenous rituximab; the remaining patient had low C3d+ DSA (MFI 1386) which became negative after receiving an intravenous infusion of rituximab and increasing immunosuppression to prednisolone, tacrolimus and MMF. Over time, the increase in C3d MFI correlated with increasing IgG MFI (co-efficient 1.5, ±1.1 units; p < 0.0001). There was no correlation between eGFR and C3d MFI (1.0 ± 1.5 ml/min/1.73 m²; p = 0.9).