Introduction: Plasma donor-derived cell-free DNA (dd-cfDNA) has been implicated as a noninvasive marker for transplant (tx) rejection. dd-cfDNA evaluation requires differentiation of donor/recipient DNA by sequencing; recent studies have amplified hundreds of target SNPs to detect active rejection in kidney allografts with 59.3% sensitivity and 84.7% specificity. We measure thousands of informative SNPs to assess dd-cfDNA with high accuracy in a selected cohort of kidney tx patients having contemporaneous tx biopsies scored for presence and type of Banff-graded T cell-/antibody-mediated–rejection (TCMR/ABMR) and borderline rejection (BL).
Materials and Methods: 292 unique plasma samples from 187 unique patients were categorized as stable (STA; n=73), acute rejection (AR; n=52), other injury (OI; n=85), or BL (n=82), and processed by massively multiplex PCR targeting 13,392 SNPs. Cross-sectional samples were obtained from AR and OI patients (other causes of graft dysfunction were drug toxicity [n=18], acute tubular necrosis [n=2], BK nephritis [n=4], chronic allograft nephropathy [CAN; n=57], and tx glomerulpathy [n=3]). AR was scored by Banff for TCMR (n=95), ABMR (n=37), and BL. 41 patients contributed 3–4 samples each (146 samples total) over 12–24 months for longitudinal assessment. dd-cfDNA performance was evaluated by ROC with inclusion of eGFR in the prediction model. Calculations were determined using 95% confidence.
Results: dd-cfDNA circulatory burden was significantly higher in AR (3.075±2.136%) compared to STA (0.428±0.851%; p<0.0001) and OI (1.051±1.112%; p<0.0001) (Figure 1). dd-cfDNA was also higher in Banff-confirmed AR over BL rejection (0.834 ±0.765%; p<0.0001), with no difference in burden observed for TCMR and ABMR (3.003 ±2.292% and 3.185 ±1.931%, respectively; p=0.5203).
To compare dd-cfDNA to eGFR score, samples with available eGFR score were used (STA, n=7; AR, n=52). Using a cutoff of >1% dd-cfDNA, AR was detected with 91.8% sensitivity (CI 80.4–97.7) and 100% specificity (CI 59–100). Area under the curve (AUC) of 0.985 showed strong AR detection power of dd-cfDNA. Using a logistic regression integrating both dd-cfDNA and eGFR with a >50% probability cutoff, classification of samples was 100% accurate. Estimated CI for sensitivity and specificity were (92.7–100) and (59–100), respectively, with AUC of 1 (compared with AUC of 0.79 using eGFR alone).
Discussion: The novel SNP-based mmPCR assay enabled rapid detection of dd-cfDNA without need for sequencing or laborious analytics. Irrespective of rejection type, the assay observed a threshold for STA patients and an exponential increase in kidney injury burden in CAN and BL rejection with much greater burden in AR; taken together, these data suggest that combined dd-cfDNA and eGFR markers can accurately assess AR risk in kidney tx recipients.
Conclusion: This technology may provide a less invasive and more sensitive approach to monitoring the health of kidney allografts.
Figure 1. Relationship of Plasma dd-cfDNA Levels and Graft Rejection Status