Transpl. Int. Transplant International Transpl. Int. 1432-2277 Frontiers Media S.A. 15517 10.3389/ti.2026.15517 Original Research Diagnostic Potential of Urine CXCL10 and Donor-Derived cfDNA in Kidney Transplant Rejection Fantus et al. Urine Chemokines, dd-cfDNA and rejection Fantus Daniel 1 2 * Balshaw Robert 3 Saw Chee Loong 4 Belkaid Majda 2 Tangprasertchai Narin S. 5 Viard Thierry 5 Gougeon François 6 Belair Justin 2 Daniel Claude 4 Lamarche Caroline 7 Casas Sílvia 5 Cardinal Heloise 1 2 Ho Julie 8 Division of Nephrology, Department of Medicine, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada Centre de Recherche de CHUM (CRCHUM), Montreal, QC, Canada George and Fay Yee Centre for Healthcare Innovation, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada Histocompatiblility and Immunogenetics Laboratory, McGill University Health Center (MUHC), Montreal, QC, Canada CareDx Inc., Brisbane, CA, United States Department of Pathology and Cell Biology, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada Division of Nephrology, Centre de Recherche de l’Hopital Maisonneuve-Rosemont, Montreal, QC, Canada Department of Internal Medicine, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada Correspondence: Daniel Fantus, daniel.fantus.med@ssss.gouv.qc.ca 27 03 2026 2026 39 15517 01 09 2025 25 01 2026 11 03 2026 Copyright © 2026 Fantus, Balshaw, Saw, Belkaid, Tangprasertchai, Viard, Gougeon, Belair, Daniel, Lamarche, Casas, Cardinal and Ho. 2026 Fantus, Balshaw, Saw, Belkaid, Tangprasertchai, Viard, Gougeon, Belair, Daniel, Lamarche, Casas, Cardinal and Ho https://creativecommons.org/licenses/by/4.0/ This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

Data suggests donor-derived cell-free DNA (dd-cfDNA) and urine CXCL10 outperform serum creatinine as a biomarker of antibody-mediated rejection (AMR) and T cell-mediated rejection (TCMR). We hypothesized that combining these biomarkers would improve the overall detection of rejection. We performed a retrospective two-center, case-controlled study of 103 adult renal transplant recipients who had for-cause or surveillance biopsies with corresponding urine and plasma samples. Rejection was classified by Banff 2022 criteria. While log10%dd-cfDNA correlated more strongly than log10CXCL10 with glomerulitis (r = 0.55, p < 0.001 vs. r = 0.25, p = 0.01) and peritubular capillaritis (r = 0.47, p < 0.001 vs. r = 0.23, p = 0.02), log10CXCL10 was a better correlate of tubulitis (r = 0.28, p = 0.004 vs. r = 0.054, p = 0.59). Both dd-cfDNA > 0.5% (OR 21.9, 95% CI 3.74–180, p < 0.001) and de novo DSA (OR 10.4, 95% CI 1.16–157, p = 0.037) were independently associated with AMR vs. no rejection (NR), while log10 serum creatinine and log10CXCL10 were not (p > 0.05). While dd-cfDNA >0.5% (OR 5.37, 95% CI 1.04–31.5, p = 0.047) was independently associated with Banff ≥1A TCMR vs. NR, log10CXCL10 was a significant predictor of TCMR in a model without %dd-cfDNA (OR 3.12, 95% CI 1.09–10.4, p = 0.043). Biomarker-guided screening strategies based on dd-cfDNA and urine chemokines such as CXCL10 for AMR (microvascular injury) and TCMR (tubulitis) warrant further study.

 Graphical Abstract

Scientific infographic presenting a summary of a study on urine CXCL10 and donor-derived cfDNA as biomarkers for kidney transplant rejection, including cohort details, biomarker list, Banff score correlations, AMR and TCMR results, and main conclusions. Graphic icons of a test tube, urine drop, and kidney appear, with logos for ESOT and Transplant International.

CXCL10 dd-cfDNA kidney transplantation non-invasive immune monitoring rejection The author(s) declared that financial support was received for this work and/or its publication. This work was supported by a research grant from the Fondation du Centre Hospitalier de l’Université de Montréal (CHUM) (DF). The urine CXCL10 testing was supported by CIHR Project Grant #364003 (JH). Introduction

Despite the widespread use of modern immunosuppressive drugs, alloimmune injury remains the most common cause of renal allograft failure today [1–6]. While a kidney biopsy remains the gold standard for rejection diagnosis, they are invasive, resource-intensive, subject to sampling error, and unsuited for longitudinal monitoring. Despite the need for non-invasive biomarkers, serum creatinine and donor specific antibody (DSA) remain the primary non-invasive tests of alloimmune renal injury used clinically [7]. Serum creatinine is a poor diagnostic biomarker of rejection [8, 9] and often only increases after significant allograft damage has occurred. DSA can be absent despite microvascular injury [10, 11], and detection of de novo DSA is only 50% predictive of subclinical AMR [12]. A novel non-invasive biomarker should be able to rule out rejection and accurately detect rejection early [13].

Urine C-X-C motif chemokine ligand 10 (CXCL10) is an interferon-gamma induced, CXC chemokine receptor 3 (CXCR3) chemokine that is highly expressed by infiltrating leukocytes and renal tubules and plays an important role in leukocyte trafficking and T helper cell type 1 (Th1) responses. Studies have shown that allograft inflammation from a variety of causes including BK viremia and urinary tract infection can upregulate urine CXCL10 [14–16]. During acute rejection, urine CXCL10 rises prior to serum creatinine [8, 17], can detect subclinical rejection [18–20], decreases after treatment of rejection [21], and remains elevated in patients who develop allograft dysfunction [17, 21–24]. Furthermore, in an unselected kidney transplant population, CXCL10 outperformed serum creatinine-based monitoring [20, 25]. However, urine CXCL10 largely reflects inflammation in the tubulointerstitial compartment [24]. As a result, arteritis and/or glomerulitis may be missed by this monitoring strategy.

Donor-derived cell-free DNA (dd-cfDNA) is a non-invasive biomarker of allograft injury that can be measured by assaying a preselected number of bi-allelic single nucleotide polymorphisms (SNPs) to distinguish donor vs. recipient cell-free DNA (cfDNA) in a sample extracted from plasma [26]. Cell death due to apoptosis and necrosis is thought to be the principal driver of cfDNA, including dd-cfDNA, release [27]. Multiple studies have shown that dd-cfDNA strongly correlates with rejection and even more strongly with AMR [28–32]. Further, dd-cfDNA has been show to increase between 1 and 5 months before biopsy-proven AMR [33] and decrease after treatment of rejection [34]. However, sensitivity is lower for TCMR, particularly low-grade and borderline TCMR [28, 35, 36].

Due to these different properties, we hypothesized that plasma dd-cfDNA and urine CXCL10 are complementary biomarkers. While elevated plasma dd-cfDNA reflects injury in the vascular compartment, urine CXCL10 reflects tubulointerstitial injury and inflammation. By directly comparing both the individual and combined diagnostic performance of plasma dd-cfDNA and urine CXCL10 when added to standard-of-care biomarkers (serum creatinine and DSA), we sought to assess how each biomarker complements rejection phenotypes.

 Materials and Methods Study Population

We performed a two-site, case-controlled study selecting 121 biopsies with corresponding pre-biopsy urine and plasma samples in our transplant biobank. This study was approved by the research ethics board of the Centre hospitalier de l’Université de Montréal on 29 August 2022 (REB number MP-02-2023-10920).

Inclusion criteria were adults (18 or more years of age) who received a renal allograft between February 2011 and June 2021 and underwent a for-cause or protocol biopsy. We excluded recipients who were pregnant, simultaneous dual-organ transplants, or patients who were prior solid non-renal organ or bone marrow transplant recipients. We also excluded patients with high-grade BK viremia ( >1000 copies/mL) and clinical cystitis/pyelonephritis at the time of biopsy, as both are associated with high urine CXCL10 levels. Our centers routinely perform one surveillance biopsy 3–6 months post-transplant.

The usual immunosuppressive protocol at our centers includes induction immunosuppression with either anti-thymocyte globulin or basiliximab and maintenance immunosuppression with mycophenolic acid, tacrolimus, and prednisone (see Table 1).

Demographic characteristics of the overall selected cohort and by diagnostic subgroup (no rejection, borderline TCMR, Banff ≥1A TCMR, and AMR).

Characteristic OverallN = 103 a No RejectionN = 53 a BorderlineN = 18 a TCMRN = 15 a AMRN = 17 a p-value b
Recipient age at biopsy 52 [40, 61] 54 [41, 60] 54 [42,63] 53 [40,69] 42 [34,53] 0.093
 Recipient male sex 72 (70%) 39 (74%) 12 (67%) 9 (60%) 12 (71%) 0.8
Cause of ESRD ​ ​ ​ ​ ​ 0.2
Polycystic kidney disease 26 (29%) 13 (31%) 5 (31%) 3 (20%) 5 (31%) ​
Glomerulonephritis 25 (28%) 11 (26%) 2 (13%) 7 (47%) 5 (31%) ​
 Diabetes 18 (20%) 12 (29%) 2 (13%) 3 (20%) 1 (6.3%) ​
 Vascular 8 (9.0%) 2 (4.8%) 2 (13%) 1 (6.7%) 3 (19%) ​
 Genetic 5 (5.6%) 1 (2.4%) 2 (13%) 0 (0%) 2 (13%) ​
 Other 7 (7.9%) 3 (7.1%) 3 (19%) 1 (6.7%) 0 (0%) ​
 Unknown 14 (13.6%) 11 (20.8%) 2 (11.1%) 0 (0%) 1 (5.9%) ​
Dialysis duration (months) 31 [13,59] 30 [13,54] 24 [15,65] 37 [15,72] 36 [12,59] >0.9
 Repeat transplant 5 (4.9%) 1 (1.9%) 0 (0%) 1 (6.7%) 3 (17.6%) 0.048
 Donor age 51 [36, 57] 51 [40, 55] 54 [49, 64] 52 [39, 63] 35 [31, 52] 0.051
 Donor male sex 53 (51%) 27 (51%) 11 (61%) 7 (47%) 8 (47%) 0.8
Living donor 25 (24%) 16 (30%) 2 (11%) 1 (6.7%) 6 (35%) 0.093
 Delayed graft function 13 (13%) 4 (7.5%) 4 (22%) 5 (33%) 0 (0%) 0.009
 Induction therapy ​ ​ ​ ​ ​ 0.8
 Basiliximab 79 (77%) 40 (75%) 13 (72%) 13 (87%) 13 (81%) ​
 Anti-thymocyte globulin (ATG) 13 (13%) 8 (15%) 2 (11%) 2 (13%) 1 (6.3%) ​
 Other 10 (9.8%) 5 (9.4%) 3 (17%) 0 (0%) 2 (13%) ​
Maintenance immunosuppression ​ ​ ​ ​ ​ 0.023
 Triple therapy (tacrolimus, mycophenolate and prednisone) 84 (82%) 47 (89%) 14 (78%) 11 (73%) 12 (71%) ​
 Bitherapy (tacrolimus and either mycophenolate or prednisone) 9 (8.7%) 4 (7.5%) 2 (11%) 3 (20%) 1 (5.9%) ​
 Other 10 (10%) 2 (3.8%) 2 (11%) 1 (6.7%) 4 (24%) ​
 Surveillance biopsy 55 (53%) 44 (83%) 8 (44%) 2 (13%) 1 (5.9%) <0.001
Months post-transplant 6 [4, 12] 6 [4, 12] 7 [4, 9] 3 [3, 9] 23 [8, 46] 0.001
Mean proteinuria (grams/L) (SD) 0.24 (0.61) 0.10 (0.25) 0.25 (0.44) 0.80 (1.65) 0.35 (0.44) 0.081
eGFR at biopsy (ml/min/1.73m2) 49 [38, 60] 52 [44, 62] 43 [28, 53] 47 [33, 54] 42 [36, 53] 0.050

Median [Q1, Q3]; n (%), SD, standard deviation.

Kruskal-Wallis rank sum test; Fisher’s Exact Test for Count Data with simulated p-value (based on 2000 replicates); Pearson’s Chi-squared test.

ESRD, end-stage renal disease. eGFR, estimated glomerular filtration rate.

Exclusion criteria included simultaneous multi-organ transplantation, prior non-renal organ or bone marrow transplantation, pregnancy, and incomplete biomarker data (urine CXCL10, plasma dd-cfDNA, and DSA). For those patients who underwent multiple biopsies, only the last biopsy was included. ESRD = end-stage renal disease, eGFR = estimated glomerular filtration rate in mL/min/1.73m2.

 Dependent Variables

Cases were defined as patients with diagnoses of AMR, Banff ≥1A TCMR, or borderline TCMR as classified by the Banff 2022 criteria [11]. Ambiguous classifications were adjudicated by a second nephrologist (JH) and pathologist (IG). Inadequate biopsies not graded by adhering to the Banff criteria were excluded. Biopsy-confirmed BK nephropathy, pyelonephritis, and recurrent glomerular disease were also excluded.

Per Banff 2022 criteria, the AMR category included all diagnoses of active AMR, probable AMR, microvascular inflammation (MVI), mixed rejection, and probable mixed rejection [11]. Diagnoses of chronic AMR were not excluded provided there was a concomitant active component as characterized by glomerulitis, peritubular capillaritis or C4d positivity. Borderline TCMR was defined using the i1t1 threshold. The biopsy-proven acute rejection (BPAR) category was a combination of the AMR and Banff ≥1A TCMR categories with borderline TCMR excluded.

Controls (NR) were defined as patients with either normal histology or low-grade inflammation that did not meet the current Banff diagnostic criteria for BPAR or borderline TCMR.

 Independent Variables Urinary CXCL10

Urine samples were collected immediately prior to biopsy and supernatants processed and stored at −80 °C. Urine CXCL10 was measured at the Transplant Immunology Lab, Canadian Blood Services, University of Manitoba using the Meso Scale V-Plex assay, using previously reported methods [37]. The subclinical rejection threshold used to calculate sensitivity and specificity for urine CXCL10 was > 13 pg/mL for male recipients over 2 weeks post-transplant, > 33 pg/mL for female recipients from 2 weeks to 5 months post-transplant, and > 13 pg/mL for female recipients > 6 months post-transplant [37, 38]. Units of measurement were picograms per ml.

 Donor Derived Cell-free DNA

cfDNA was extracted from EDTA plasma samples collected immediately prior to the biopsy, following a protocol designed to avoid genomic DNA contamination. The percent dd-cfDNA was determined using the AlloSeq cfDNA Assay and Software (CareDx, Brisbane, CA) by the manufacturer. Due to the significant number of borderline TCMR cases (n = 18) in our study, coupled with the fact that our centers perform protocol biopsies as part of standard care, we selected an injury threshold of 0.5%. Others have previously compared sensitivity and specificity at various thresholds and have found sensitivity for rejection higher at this threshold [36, 39, 40]. Prior comparison of this assay versus the Allosure kidney assay has shown high correlation [41].

 Donor-Specific Antibodies

DSA was assessed in samples collected either at the time of biopsy or in the preceding month. Analysis was conducted at the HLA laboratory of the McGill University Health Centre using a threshold of 1000 mean fluorescence intensity (MFI) units of the immunodominant DSA to dichotomize DSA as positive vs. negative per Canadian standards [42].

 Statistical Analyses

Individual characteristics of the biopsies in each group were summarized within groups and overall using quartiles (Q1, median, and Q3) for numerical variables and counts and percentages for categorical variables. Kruskal-Wallis was used to compare the distribution of numeric variables between groups (or Wilcoxon tests for two-group comparisons). Chi-squared tests were used to compare categorical variables; Fisher’s Exact test was used when cell sizes were smaller than 5.

Logistic regression models were used to examine relationships between plasma %dd-cfDNA, urine CXCL10, serum creatinine, and rejection events both with and without consideration of DSA depending on the rejection phenotype. Generalized estimating equations were used in sensitivity analyses to fit similar models to a dataset incorporating all biopsies with complete data, assuming an exchangeable working correlation matrix to account for within-participant correlations. Due to skewness, plasma % dd-cfDNA, urine CXCL10, and serum creatinine concentrations were analyzed on a log10 scale. Their association with Banff scores were described using Pearson correlations. Sensitivity and specificity for %dd-cfDNA and urine CXCL10 using thresholds as described above are also reported. As the small number of AMR events led to difficulties fitting conventional maximum likelihood logistic regression models, parameter estimates for the AMR vs. NR model were obtained via Firth’s penalized likelihood method [43]. Differences in AIC (Akaike Information Criterion) values were used as a comparative measure of model fit, with lower AIC values indicating a better fitting model. Model discrimination was assessed by evaluating the AUC (area under the receiver-operator characteristic curve). Statistical analyses were performed using R (version 4.4.2) and the logistf package (version 1.26.0). In general, p-values less than 0.05 were described as statistically significant, without correction for multiple inference.

 Results

Between 26 February 2011 and 8 June 2021, 1,215 patients underwent renal transplantation at our institutions (Figure 1). Of these patients, 1,028 enrolled in our biobank and 1,259 biopsies were performed. Although our initial cohort of cases and controls encompassed 181 biopsies from 160 patients, 60 biopsies were ultimately excluded due to inadequate tissue and/or incomplete Banff scores (n = 23), insufficient cell free DNA for analysis (n = 32), or urine samples that were not collected (n = 5). As a result, 121 biopsies from 103 patients had measurements for all traditional and novel biomarkers as well as sufficient tissue to provide Banff scores. Demographic characteristics, other than shorter dialysis exposure and cause of end-stage renal disease, were similar between our cohort and excluded patients (see Supplementary Table S1). As there was no significant difference in our results when we used a generalized estimating equation (GEE) model to account for multiple biopsies per patient, our primary analytic cohort comprised 103 biopsies: one per patient, with the most recent biopsy selected (see Supplementary Table S2; Supplementary Figure S1). Within this cohort, 17.5% (n = 18) of the biopsies were categorized as Banff Borderline TCMR, 14.6% (n = 15) were Banff ≥1A TCMR, and 16.5% (n = 17) were AMR/mixed rejection (see Table 1).

Study Population. Out of a total of 1259 biopsies performed during the study period as part of our biobank, 181 biopsies were initially selected from 160 patients. However, 60 biopsies were subsequently excluded due to incomplete biomarker data or an inadequate biopsy not adhering to the Banff criteria. The remaining 121 biopsies were selected from 103 patients with one of the following diagnoses as per Banff 2022 criteria: no rejection, borderline TCMR, Banff ≥1A TCMR, and AMR (including probable AMR and MVI). The primary analytic cohort encompassed 103 biopsies from 103 patients. inc = included.

Flowchart diagram shows progression from 1,215 kidney transplants, 1,028 patients, and 1,259 biopsies to an initial cohort of 181 biopsies, reduced by 60 exclusions, resulting in a primary analytic cohort of 121 biopsies and 103 patients, divided into no rejection (53 biopsies), borderline TCMR (18), Banff≥1A TCMR (15), and AMR including MVI (17).

Demographic characteristics of the overall cohort and by diagnostic category are presented in Table 1; Supplementary Table S3. All patients underwent ABO-compatible transplantation and had a negative flow cytometric or complement-dependent cytotoxicity crossmatch prior to transplantation as well as no pre-transplant DSA. AMR patients tended to be younger and were less likely to have diabetes and delayed graft function. They were also less likely to have been treated with maintenance immunosuppression consisting of tacrolimus, MMF, and prednisone at the time of rejection. Although most of the TCMR and AMR biopsies were performed due to graft dysfunction, the AMR biopsies were generally performed later post-transplant.

Biomarker levels are presented in Table 2; Figure 2, stratified by diagnostic category. The distribution of Banff scores by rejection category is also provided in Table 2. Log10%dd-cfDNA is higher in patients with AMR while log10 serum creatinine and log10CXCL10 show smaller between-group differences (Figure 2).

A comparison of biomarker levels and Banff scores by diagnostic classification.

Characteristic No RejectionN = 53 a BorderlineN = 18 a TCMRN = 15 a AMRN = 17 a OverallN = 103 a p-value b
Urine CXCL10 (pg/mL) 7 [4, 14] 9 [1, 25] 17 [6, 42] 13 [9, 63] 9 [4, 22] 0.070
CXCL10 positive 13 (25%) 8 (44%) 8 (53%) 9 (53%) 38 (37%) 0.056
Serum creatinine (μmol/L) 119 [109,135] 138 [125, 203] 121 [116, 176] 152 [125, 172] 125 [114, 159] 0.008
%dd-cfDNA 0.20 [0.14, 0.30] 0.37 [0.17, 0.60] 0.34 [0.14, 0.58] 1.73 [0.76, 3.25] 0.26 [0.17, 0.56] <0.001
%dd-cfDNA positive c 3 (5.7%) 7 (39%) 5 (33%) 15 (88%) 30 (29%) <0.001
DSA positive (de novo) 1 (1.9%) 0 (0%) 0 (0%) 12 (71%) 13 (13%) <0.001
Glomerulitis (g score) ​ ​ ​ ​ ​ <0.001
 0 53 (100%) 18 (100%) 15 (100%) 6 (35%) 92 (89%) ​
 1 0 (0%) 0 (0%) 0 (0%) 6 (35%) 6 (5.8%) ​
 2 0 (0%) 0 (0%) 0 (0%) 4 (24%) 4 (3.9%) ​
 3 0 (0%) 0 (0%) 0 (0%) 1 (5.9%) 1 (1.0%) ​
Interstitial inflammation (i score) ​ ​ ​ ​ ​ <0.001
 0 52 (98%) 0 (0%) 1 (6.7%) 10 (59%) 63 (61%) ​
 1 1 (1.9%) 18 (100%) 3 (20%) 2 (12%) 24 (23%) ​
 2 0 (0%) 0 (0%) 9 (60%) 3 (18%) 12 (12%) ​
 3 0 (0%) 0 (0%) 2 (13%) 2 (12%) 4 (3.9%) ​
Tubulitis (t score) ​ ​ ​ ​ ​ <0.001
 0 32 (60%) 0 (0%) 1 (6.7%) 8 (47%) 41 (40%) ​
 1 14 (26%) 13 (72%) 2 (13%) 6 (35%) 35 (34%) ​
 2 6 (11%) 5 (28%) 6 (40%) 1 (5.9%) 18 (17%) ​
 3 1 (1.9%) 0 (0%) 6 (40%) 2 (12%) 9 (8.7%) ​
Intimal arteritis (v score) ​ ​ ​ ​ ​ <0.001
 0 52 (100%) 18 (100%) 9 (60%) 14 (82%) 93 (91%) ​
 1 0 (0%) 0 (0%) 2 (13%) 2 (12%) 4 (3.9%) ​
 2 0 (0%) 0 (0%) 4 (27%) 1 (5.9%) 5 (4.9%) ​
Peritubular capillaritis (ptc score) ​ ​ ​ ​ ​ <0.001
 0 51 (96%) 17 (94%) 9 (60%) 3 (18%) 80 (78%) ​
 1 2 (3.8%) 1 (5.6%) 3 (20%) 5 (29%) 11 (11%) ​
 2 0 (0%) 0 (0%) 2 (13%) 4 (24%) 6 (5.8%) ​
 3 0 (0%) 0 (0%) 1 (6.7%) 5 (29%) 6 (5.8%) ​
c4d deposition in PTC ​ ​ ​ ​ ​ <0.001
 0 53 (100%) 18 (100%) 14 (93%) 6 (38%) 91 (89%) ​
 2 0 (0%) 0 (0%) 0 (0%) 2 (13%) 2 (2.0%) ​
 3 0 (0%) 0 (0%) 1 (6.7%) 8 (50%) 9 (8.8%) ​
GBM double contours (cg score) ​ ​ ​ ​ ​ <0.001
 0 52 (98%) 17 (94%) 15 (100%) 9 (53%) 93 (90%) ​
 1 1 (1.9%) 1 (5.6%) 0 (0%) 3 (18%) 5 (4.9%) ​
 2 0 (0%) 0 (0%) 0 (0%) 2 (12%) 2 (1.9%) ​
 3 0 (0%) 0 (0%) 0 (0%) 3 (18%) 3 (2.9%) ​
Interstitial fibrosis (ci score) ​ ​ ​ ​ ​ <0.001
 0 35 (66%) 6 (33%) 3 (20%) 4 (24%) 48 (47%) ​
 1 17 (32%) 9 (50%) 12 (80%) 12 (71%) 50 (49%) ​
 2 0 (0%) 3 (17%) 0 (0%) 1 (5.9%) 4 (3.9%) ​
 3 1 (1.9%) 0 (0%) 0 (0%) 0 (0%) 1 (1.0%) ​
Tubular atrophy (ct score) ​ ​ ​ ​ ​ 0.002
 0 25 (47%) 1 (5.6%) 3 (20%) 3 (18%) 32 (31%) ​
 1 27 (51%) 16 (89%) 12 (80%) 12 (71%) 67 (65%) ​
 2 0 (0%) 1 (5.6%) 0 (0%) 2 (12%) 3 (2.9%) ​
 3 1 (1.9%) 0 (0%) 0 (0%) 0 (0%) 1 (1.0%) ​
Arterial intimal thickening (cv score) ​ ​ ​ ​ ​ 0.4
 0 20 (40%) 4 (24%) 3 (20%) 5 (31%) 32 (33%) ​
 1 19 (38%) 5 (29%) 8 (53%) 8 (50%) 40 (41%) ​
 2 10 (20%) 7 (41%) 3 (20%) 2 (13%) 22 (22%) ​
 3 1 (2.0%) 1 (5.9%) 1 (6.7%) 1 (6.3%) 4 (4.1%) ​
Arteriolar hyalinosis (ah score) ​ ​ ​ ​ ​ 0.069
 0 29 (55%) 8 (44%) 6 (40%) 4 (24%) 47 (46%) ​
 1 20 (38%) 8 (44%) 6 (40%) 6 (35%) 40 (39%) ​
 2 4 (7.5%) 1 (5.6%) 2 (13%) 5 (29%) 12 (12%) ​
 3 0 (0%) 1 (5.6%) 1 (6.7%) 2 (12%) 4 (3.9%) ​

Median [Q1, Q3]; n (%).

Kruskal-Wallis rank sum test; Pearson’s Chi-squared test; Fisher’s exact test; NA.

dd-cf DNA >0.5%.

Absolute biomarker levels for urine CXCL10 (pg/mL), serum creatinine (μmol/L), and plasma dd-cfDNA (%) are compared among the four major diagnostic classifications (no rejection, borderline TCMR, Banff ≥1A TCMR, and AMR). The percentage of patients positive for urine CXCL10 and %dd-cfDNA are based on pre-defined thresholds. The subclinical rejection threshold used for urine CXCL10 was >13 pg/mL for male recipients over 2 weeks post-transplant, >33 pg/mL for female recipients from 2 weeks to 5 months post-transplant, and >13 pg/mL for female recipients >6 months post-transplant. An injury threshold of 0.5% dd-cfDNA was used. DSA is presented as a categorical variable with positivity defined as a minimum of 1000 MFI units of the immunodominant DSA.

A comparison of log-transformed biomarker levels (urine CXCL10, plasma %dd-cfDNA, and serum creatinine) by diagnostic classification (no rejection, borderline, TCMR, and AMR).

Three side-by-side box plots compare biomarker levels by diagnostic classification: CXCL10, dd-cfDNA, and Creatinine, each grouped by No Rejection, Borderline, TCMR, and AMR. Statistical significance values are indicated for select pairwise comparisons above the boxes. Association With Acute and Chronic Banff Scores

Next, we sought to investigate the correlation between Banff scores and urine CXCL10 and plasma dd-cfDNA, respectively (Figure 3). Overall, log10%dd-cfDNA correlated more strongly than log10CXCL10 with glomerulitis (r = 0.55, p < 0.001 vs. r = 0.25, p = 0.012), peritubular capillaritis (r = 0.47, p < 0.001 vs. r = 0.23, p = 0.019), peritubular capillary c4d staining (r = 0.47, p < 0.001 vs. r = 0.26, p = 0.008), and glomerular basement membrane double contours (r = 0.41, p < 0.001 vs. r = 0.12, p = 0.226). In contrast, although neither log10%dd-cfDNA nor log10CXCL10 correlated strongly with tubulitis, performance characteristics favored log10CXCL10 (r = 0.28, p = 0.004 vs. r = 0.054, p = 0.588). The correlation was similar when interstitial inflammation scores were compared (Figure 3). In contrast, there was poor correlation with both log10%dd-cfDNA and log10CXCL10 and chronic Banff scores (Supplementary Figure S2).

A comparison of the association of log-transformed biomarker levels (urine CXCL10 and plasma %dd-cfDNA) with the Banff scores required to diagnose acute rejection (g = glomerulitis, i = interstitial inflammation, ptc = peritubular capillaritis, t = tubulitis, and v = intimal arteritis).

Scatterplot matrix shows the association of acute rejection Banff scores with CXCL10 (left) and dd-cfDNA (right), with each subplot presenting correlation coefficient r and statistical significance p-value for different Banff criteria (c4d, g, i, ptc, t, v), and regression lines fitted to the data points. AMR vs. No Rejection

Because of the small number of AMR biopsies, the multi-predictor model for AMR vs. NR was fit using Firth’s penalized likelihood method (Figure 4). The log10 serum creatinine OR was 1.33 (95% CI 0.000445 to 5480, p = 0.943), log10CXCL10 OR 0.794 (95% CI 0.0611 to 6.76, p = 0.842), DSA OR 13.0 (95% CI 1.34 to 220, p = 0.027), and log10%dd-cfDNA OR 40.8 (95% CI 4.13 to 925, p < 0.001). When %dd-cfDNA was used as a categorical variable, the log10 serum creatinine OR was 6.98 (95% CI 0.00417 to 34,500, p = 0.619), log10CXCL10 OR 1.01 (95% CI 0.109 to 8.82, p = 0.990), DSA OR 10.4 (95% CI 1.16 to 157, p = 0.037), and dd-cfDNA >0.5% OR 21.9 (95% CI 3.74 to 180, p < 0.001). These data demonstrated that measurement of %dd-cfDNA improved AMR diagnosis independent of DSA.

Logistic regression analysis for AMR. (A) Odds ratio for log10serum creatinine, log10urine CXCL10, DSA, and either log10%dd-cfDNA or %dd-cfDNA >0.5 using Firths penalized likelihood method. (B) Concomitant Receiver Operating Characteristic (ROC) analysis using various combinations of biomarkers. (C) Concomitant Akaike Information Criterion (AIC) analysis. (Creat = creatinine).

Table summarizing AMR logistic regression model odds ratios and p-values for different biomarkers, including serum creatinine, urine CXCL10, DSA, and percent dd-cfDNA. Receiver operating characteristic curve compares diagnostic sensitivity and specificity for various biomarker models, showing AUC values. Akaike information criterion plot ranks statistical models based on model complexity and information loss, with a legend matching dot colors to biomarker combinations.

Furthermore, AUCs for distinguishing AMR vs. NR increased from 0.696 (95% CI 0.527–0.865) for log10 serum creatinine alone to 0.967 (95% CI 0.930–1.000) and 0.959 (0.915–1.000) when log10 serum creatinine and either log10%dd-cfDNA or dd-cfDNA >0.5% were combined, respectively (Table 3). In contrast, the AUC for detecting AMR was 0.706 (95% CI 0.544–0.867) when log10 serum creatinine and log10CXCL10 were combined.

Comparison of AUCs with 95% confidence intervals for all models of AMR, Banff ≥1A TCMR, and BPAR (Banff ≥1A TCMR and AMR), with percent dd-cfDNA included as either a continuous variable or applying a threshold of 0.5%.

Model AMR (95%CI) TCMR (95%CI) BPAR (95%CI)
Log10serum creatinine 0.696 (0.527–0.865) 0.621 (0.452–0.789) 0.661 (0.534–0.787)
DSA 0.844 (0.730–0.957) ​ 0.678 (0.591–0.765)
%dd-cfDNA >0.5 0.913 (0.828–0.998) 0.638 (0.510–0.766) 0.784 (0.693–0.875)
Log10 (%dd-cfDNA) 0.973 (0.941–1.000) 0.672 (0.493–0.851) 0.832 (0.729–0.934)
Log10 (cxcl10) 0.691 (0.540–0.841) 0.654 (0.482–0.826) 0.674 (0.552–0.796)
Log10 serum creatinine + DSA 0.896 (0.774–1.000) ​ 0.763 (0.649–0.878)
Log10serum creatinine + %dd-cfDNA >0.5 0.959 (0.915–1.000) 0.724 (0.565–0.883) 0.849 (0.756–0.941)
Log10serum creatinine + log10 (%dd-cfDNA) 0.967 (0.930–1.000) 0.682 (0.497–0.867) 0.833 (0.729–0.937)
Log10serum creatinine + log10 (cxcl10) 0.706 (0.544–0.867) 0.717 (0.550–0.884) 0.713 (0.592–0.834)
DSA + %dd-cfDNA >0.5 0.932 (0.847–1.000) ​ 0.791 (0.699–0.883)
DSA + log10 (%dd-cfDNA) 0.979 (0.954–1.000) ​ 0.830 (0.728–0.932)
DSA + log10 (cxcl10) 0.851 (0.705–0.996) ​ 0.750 (0.631–0.869)
Log10 (cxcl10) + %dd-cfDNA >0.5 0.920 (0.808–1.000) 0.725 (0.560–0.889) 0.828 (0.724–0.933)
Log10 (cxcl10) + log10 (%dd-cfDNA) 0.972 (0.940–1.000) 0.688 (0.516–0.860) 0.831 (0.729–0.934)
Log10 (cxcl10) + DSA + %dd-cfDNA >0.5 0.931 (0.820–1.000) ​ 0.829 (0.724–0.934)
Log10 (cxcl10) + DSA + log10 (%dd-cfDNA) 0.979 (0.953–1.000) ​ 0.829 (0.725–0.933)
Log10serum creatinine + DSA + %dd-cfDNA >0.5 0.977 (0.949–1.000) ​ 0.855 (0.765–0.945)
Log10serum creatinine + DSA + log10 (%dd-cfDNA) 0.981 (0.957–1.000) ​ 0.832 (0.729–0.935
Log10serum creatinine + log10 (cxcl10) + DSA 0.878 (0.748–1.000) ​ 0.793 (0.682–0.904)
Log10serum creatinine + log10 (cxcl10) + %dd-cfDNA >0.5 0.959 (0.915–1.000) 0.767 (0.604–0.931) 0.862 (0.771–0.953
Log10serum creatinine + log10 (cxcl10) + log10 (%dd-cfDNA) 0.967 (0.930–1.000) 0.725 (0.554–0.896) 0.831 (0.727–0.936)
Log10serum creatinine + log10 (cxcl10) + DSA + %dd-cfDNA >0.5 0.978 (0.951–1.000) ​ 0.866 (0.775–0.956
Log10serum creatinine + log10 (cxcl10) + DSA + log10 (%dd-cfDNA) 0.980 (0.955–1.000) ​ 0.826 (0.720–0.933
Banff ≥1A TCMR Vs. No Rejection

Next, we compared the performance of serum creatinine, %dd-cfDNA, and urine CXCL10 in the detection of Banff ≥1A TCMR. AUCs and AICs were compared (Figure 5; Table 3). The log10 serum creatinine OR was 95.7 (95% CI 0.36 to 41,100, p = 0.120), log10CXCL10 OR 2.68 (95% CI 0.850 to 9.73, p = 0.110), and log10%dd-cfDNA OR 3.20 (95% CI 0.560 to 22.1, p = 0.200). When the 0.5% dd-cfDNA threshold was used, the log10 serum creatinine OR was 91.3 (95% CI 0.260 to 44,300, p = 0.130), log10CXCL10 OR 2.65 (95% CI 0.830 to 9.87, p = 0.120), and dd-cfDNA >0.5% OR 5.37 (95% CI 1.04 to 31.5, p = 0.047). Therefore, dd-cfDNA >0.5% was the only independent predictor of Banff ≥1A TCMR in this cohort. However, in a model without %dd-cfDNA, log10CXCL10 detected TCMR independently of serum creatinine (OR 3.12, 95% CI 1.09–10.4, p = 0.043) (data not shown).

Logistic regression analysis for Banff ≥1A TCMR. (A) Odds ratio for log10serum creatinine, log10urine CXCL10, and either log10%dd-cfDNA or %dd-cfDNA >0.5 using logistic regression. (B) Concomitant Receiver Operating Characteristic (ROC) analysis using various combinations of biomarkers. (C) Concomitant Akaike Information Criterion (AIC) analysis. (Creat = creatinine).

Panel A presents a table showing TCMR logistic regression results for biomarkers, including odds ratios, confidence intervals, and p-values; Panel B displays receiver operating characteristic curves for different biomarker models with reported area under the curve values; Panel C presents an Akaike information criterion plot for model comparison, with legend identifying each model by color.

AUCs for distinguishing TCMR vs. NR increased from 0.621 (95% CI 0.452–0.789) for log10 serum creatinine alone to 0.717 (95% CI 0.550–0.884) when log10 serum creatinine and log10CXCL10 were combined (Table 3). The AUC increased to 0.682 (95% CI 0.497–0.867) and 0.724 (95% CI 0.565–0.883) when log10 serum creatinine and either log10%dd-cfDNA or dd-cfDNA >0.5% were combined, respectively. When log10CXCL10 and either log10%dd-cfDNA or dd-cfDNA >0.5% were added to log10 serum creatinine, AUC increased to 0.725 (95% CI 0.554–0.896) and 0.767 (95% CI 0.604–0.931), respectively. That is, there was an improvement in the diagnosis of Banff ≥1A TCMR with the addition of both urine CXCL10 and %dd-cfDNA to serum creatinine.

 BPAR vs. No Rejection

We also assessed performance in the diagnosis of BPAR (AMR and Banff ≥1A TCMR) vs. NR. Models were fit using various combinations (Figure 6; Table 3). In a multi-predictor logistic regression model, the log10 serum creatinine OR was 17.4 (95% CI 0.160–2640, p = 0.200), log10CXCL10 OR 2.00 (95% CI 0.680–6.32, p = 0.200), DSA OR 8.58 (95% CI 0.870–200, p = 0.092), and log10%dd-cfDNA OR 11.1 (95% CI 2.24–75.1, p = 0.006). Similarly, when dichotomized, dd-cfDNA >0.5% OR 13.4 (95% CI 3.14–75.8, p = 0.001) remained the only significant predictor of BPAR.

Logistic regression analysis for BPAR (Banff ≥1A TCMR and AMR). (A) Odds ratio for log10serum creatinine, log10urine CXCL10, DSA, and either log10%dd-cfDNA or %dd-cfDNA >0.5 using logistic regression. (B) Concomitant Receiver Operating Characteristic (ROC) analysis using various combinations of biomarkers. (C) Concomitant Akaike Information Criterion (AIC) analysis. (Creat = creatinine).

Figure containing three panels comparing BPAR logistic regression model performance. Panel A shows a table of biomarkers with odds ratios, confidence intervals, and p-values. Panel B displays a ROC curve comparing model sensitivities and area under the curve values. Panel C presents an AIC scatterplot for multiple models, with a color-coded legend identifying each model.

AUCs for distinguishing BPAR vs. NR increased from 0.661 (95% CI 0.534–0.787) for log10 serum creatinine alone to 0.833 (95% CI 0.729–0.937) and 0.849 (95% CI 0.756–0.941) when log10 serum creatinine and either log10%dd-cfDNA or dd-cfDNA >0.5% were combined, respectively (Table 3). AUC increased to 0.713 (95% CI 0.592–0.834) when log10CXCL10 was added to log10 serum creatinine.

 Borderline TCMR vs. No Rejection

Similar analyses were performed comparing borderline TCMR vs. NR. In a multi-predictor logistic regression model, the log10 serum creatinine OR was 939 (95% CI 5.43–309732, p = 0.014), log10CXCL10 OR 1.19 (95% CI 0.360–3.96, p = 0.800), and log10%dd-cfDNA OR 2.59 (95% CI 0.520–14.1, p = 0.200) (Figure 7). When the 0.5% dd-cfDNA threshold was used, the log10 serum creatinine OR was 386 (95% CI 1.57–180,000, p = 0.043), log10CXCL10 OR 1.27 (95% CI 0.380–4.33, p = 0.700), and dd-cfDNA >0.5% OR 5.85 (95% CI 1.19–33.5, p = 0.032). In this model, dd-cfDNA >0.5% was a significant independent predictor of borderline TCMR.

Logistic regression analysis for Borderline TCMR. (A) Odds ratio for log10serum creatinine, log10urine CXCL10, and either log10%dd-cfDNA or %dd-cfDNA >0.5 using logistic regression. (B) Concomitant Receiver Operating Characteristic (ROC) analysis using various combinations of biomarkers. (C) Concomitant Akaike Information Criterion (AIC) analysis. (Creat = creatinine).

Figure composed of three panels. Panel A shows a logistic regression table with biomarkers, odds ratios, confidence intervals, and p-values comparing serum creatinine, urine CXCL10, and dd-cfDNA for borderline TCMR, with two model columns. Panel B displays overlapping ROC curves comparing sensitivity versus 1-specificity for creatinine, creatinine plus CXCL10, and creatinine plus dd-cfDNA, and indicates AUC values of 0.768, 0.731, and 0.729, respectively. Panel C presents an Akaike information criterion (AIC) scatter plot for six regression models, color-coded to correspond with different biomarker combinations listed in the legend at right.

AUCs for distinguishing borderline TCMR vs. NR increased from 0.729 (95% CI 0.590–0.868) for log10 serum creatinine alone to 0.768 (95% CI 0.631–0.906) when log10 serum creatinine and dd-cfDNA >0.5% were combined (Table 4).

Comparison of AUCs with 95% confidence intervals for all models of borderline TCMR with percent dd-cfDNA included as either a continuous variable or applying a threshold of 0.5%.

Model AUC 95% CI
Log10serum creatinine 0.729 0.590–0.868
%dd-cfDNA >0.5 0.666 0.546–0.786
Log10 (% dd-cfDNA) 0.661 0.486–0.836
Log10(CXCL10) 0.508 0.329–0.687
Log10serum creatinine + %dd- cfDNA >0.5 0.768 0.631–0.906
Log10serum creatinine + log10 (%dd-cfDNA) 0.737 0.603–0.871
Log10serum creatinine + log10(CXCL10) 0.731 0.595–0.866
Log10(CXCL10) + % dd-cfDNA >0.5 0.724 0.573–0.876
Log10(CXCL10) + log10 (% dd-cfDNA) 0.647 0.471–0.822
Log10serum creatinine + log10(CXCL10) + % dd-cfDNA >0.5 0.781 0.654–0.908
Log10serum creatinine + log10(CXCL10) + log10 (%dd- cfDNA) 0.753 0.627–0.879
Sensitivity, Specificity and Bivariate Scatter Plot for CXCL10 and dd-cfDNA

Sensitivities and specificities, using the aforementioned thresholds, were calculated for each biomarker alone as well as together, if either were positive, for AMR, Banff ≥1A TCMR, and borderline TCMR (see Table 5). While performance characteristics for AMR favored %dd-cfDNA alone, sensitivity for both TCMR and borderline TCMR increased when urine CXCL10 was added to %dd-cfDNA. Sensitivity increased from 0.33 to 0.67 for TCMR and from 0.39 to 0.72 for borderline TCMR. Given each diagnostic category was compared to NR, specificities did not change across groups. A bivariate scatter plot of log10CXCL10 vs. log10%dd-cfDNA, with ellipses centered at the mean values for each group, for AMR, Banff ≥1A TCMR, borderline TCMR, and no rejection (see Figure 8) corroborates these observations, with differences associated with AMR favoring dd-cfDNA. For differences associated with Banff ≥1A TCMR, both urine CXCL10 and dd-cfDNA performed modestly. Differences associated with borderline TCMR slightly favored dd-cfDNA.

Sensitivity and specificity for %dd-cfDNA and urine CXCL10 in AMR, Banff ≥1A TCMR, and borderline TCMR.

AMR vs. NR Sensitivity Specificity
%dd-cfDNA positive 0.88 (0.64–0.99) 0.94 (0.84–0.99)
CXCL10 positive 0.53 (0.28–0.77) 0.75 (0.62–0.86)
%dd-cfDNA or CXCL10 positive 0.94 (0.71–1.00) 0.72 (0.58–0.83)
TCMR vs. NR Sensitivity
%dd-cfDNA positive 0.33 (0.12–0.62)
CXCL10 positive 0.53 (0.27–0.79)
%dd-cfDNA or CXCL10 positive 0.67 (0.38–0.88)
Borderline vs. NR Sensitivity
%dd-cfDNA positive 0.39 (0.17–0.64)
CXCL10 positive 0.44 (0.22–0.69)
%dd-cfDNA or CXCL10 positive 0.72 (0.47–0.90)

The percentage of patients positive for urine CXCL10 and dd-cfDNA are based on pre-defined thresholds. The subclinical rejection threshold used for urine CXCL10 was >13 pg/mL for male recipients over 2 weeks post-transplant, >33 pg/mL for female recipients from 2 weeks to 5 months post-transplant, and >13 pg/mL for female recipients >6 months post-transplant. A threshold of 0.5% dd-cfDNA was used.

Bivariate scatter plot of log10 urine CXCL10 versus log10 %dd-cfDNA for no rejection, borderline TCMR, Banff ≥1A TCMR, and AMR. Ellipses are centered at the mean values in each target group.

Bivariate scatter plot showing log10-transformed CXCL10 versus log10-transformed percent dd-cfDNA. Data points are grouped and color-coded by four categories: No rejection (gray), Borderline (orange), TCMR (blue), and AMR (red), with translucent ellipses highlighting the data range for each group. A legend to the right matches color to group name. Discussion

In this case-controlled study, we have shown that %dd-cfDNA is a better non-invasive biomarker of AMR than serum creatinine and urine CXCL10. However, when Banff ≥1A TCMR was assessed, %dd-cfDNA and urine CXCL10 were synergistic. Although performance characteristics for TCMR slightly favored %dd-cfDNA in our logistic regression model, their combination enhanced sensitivity and AUC. Over the last 15 years, various non-invasive diagnostic biomarkers of kidney transplant rejection have been evaluated, with the most data on urine chemokines, plasma dd-cfDNA, and blood gene expression profiling [13, 44]. Positive predictive values have remained low due to the low incidence of rejection under modern immunosuppression, the inherent limitations of histological sampling which may underestimate rejection, and confounders such as infection and non-alloimmune acute kidney injury. As a result, they appear to be more promising as non-invasive “rule out” tests for rejection in stable patients rather than as stand-alone diagnostic tests. At the same time, further study on how these tests may be used as adjuncts to histology are necessary.

Despite advances in our understanding of the diagnostic accuracy of urine CXCL10 and dd-cfDNA in kidney transplant rejection, few studies [45] have simultaneously measured these two biomarkers in the same cohort. In this study we showed urine CXCL10 was a better correlate of tubulitis than %dd-cfDNA and, when compared to serum creatinine alone, independently improved the non-invasive diagnosis of Banff ≥1A TCMR but did not add to AMR detection. Conversely, plasma %dd-cfDNA correlated more strongly than urine CXCL10 with microvascular inflammation and independently improved the non-invasive diagnosis of both AMR and Banff ≥1A TCMR. Overall, our histology correlations are consistent with what others have reported for both %dd-cfDNA and urine CXCL10 [20, 29, 46]. Although Rabant et al. demonstrated that urine CXCL10 independently improves AMR diagnosis, %dd-cfDNA was not included in that analysis [47]. Furthermore, the AUC for AMR in this study, when compared to %dd-cfDNA performance, was relatively modest at 0.702.

Overall, this study—the first to simultaneously measure these two novel biomarkers in a meticulously characterized clinical-pathological cohort including protocol biopsies—demonstrates that urine CXCL10 may not be helpful as a diagnostic biomarker of AMR when plasma %dd-cfDNA and DSA are available. Our study also adds to prior work demonstrating that %dd-cfDNA improves the non-invasive detection of AMR independently of serum creatinine and DSA [10, 39].

In contrast, both urine CXCL10 and %dd-cfDNA were modest predictors of Banff ≥1A TCMR. These data are consistent with recent studies showing modest performance for both urine CXCL10 and %dd-cfDNA for detecting borderline TCMR and/or Banff ≥1A TCMR [28, 35, 36, 48]. The large proportion of TCMR without vascular lesions in our study may explain the lower %dd-cfDNA diagnostic performance compared to other studies [29]. The decrease in performance of urine CXCL10 when %dd-cfDNA was included in our multivariable analysis suggests some diagnostic overlap. This may be explained by the fact that both correlate similarly with interstitial inflammation (Figure 3). In addition, the inclusion of biopsies with low-grade inflammation in the NR group, 40% of which had tubulitis scores ≥ t1 and would have met the 2015 definition of borderline, although appropriate, may have decreased the performance characteristics of urine CXCL10. Nonetheless, its greater sensitivity compared to %dd-cfDNA for both Banff ≥1A and borderline TCMR supports its potential as a screening tool. Although other diagnostic advantages of urine CXCL10 may lie in its higher correlation with tubulitis, its early identification of post-transplant infections such as BK and ease of sample collection, in our study, %dd-cfDNA, due to its higher specificity, more effectively categorized low-grade inconsequential inflammation such as isolated tubulitis as no rejection [15, 16]. As a result, one approach might be to use urine CXCL10 and %dd-cfDNA sequentially to more accurately non-invasively detect Banff ≥1A and borderline TCMR. Given that both of these rejection phenotypes are clinically important due to the risk of developing persistent chronic active TCMR, de novo DSA, or AMR [49–51], defining optimally targeted screening strategies remains a priority.

Limitations of our study include potential over-estimation of diagnostic performance based on a curated case-control cohort with exclusion of non-rejection diagnoses. The cohort consisted of 47% clinical indication biopsies wherein the decision to biopsy was already made using serum creatinine and clinical judgement. Furthermore, as centers that perform protocol biopsies, there is a relative over-representation of more severe rejection phenotypes (AMR and Banff ≥1A TCMR) compared to borderline TCMR. While borderline TCMR is the most common rejection phenotype we see within the first year (unpublished data), AMR is less frequent yet increases over time. In addition, while urine CXCL10 was selected due to having an assay with well-developed performance characteristics and pre-analytical sample handling data and the potential to be used in clinical labs [37], exclusion of urine CXCL9, a urine chemokine with more favorable diagnostic characteristics [48] is a limitation. Further, due to the small size of our cohort, we were unable to attempt to distinguish among AMR subsets, e.g., our AMR group includes a heterogeneous mix of mixed rejection as well as DSA negative, c4d negative microvascular inflammation (MVI). We do know that these phenotypes have different long-term prognoses and that performance characteristics of the biomarkers may differ [52]. Finally, this exploratory study was not designed or powered to assess whether absolute biomarker levels reflect degree of injury and whether changes post-therapy reflect resolution vs. persistent inflammation. Nevertheless, this analysis offers new and important insights into the relative associations of urine CXCL10 and plasma dd-cfDNA with different types of rejection.

Taken together, the potential for individualized biomarker-guided screening strategies for AMR and TCMR deserves further investigation in larger unselected prospective observational cohort studies to provide real-world diagnostic performance characteristics. Such studies will enhance practical clinical implementation of these biomarkers into post-transplant care [44, 53].

 Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

 Ethics Statement

The studies involving humans were approved by Research ethics board of the Centre hospitalier de l’Université de Montréal. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

 Author Contributions

DF and RB: Participated in research design, writing of the paper, the performance of the research, and data analysis. CS, NT, TV, and CD: Participated in performance of the research, contributed reagents and analytic tools, and participated in data analysis. MB: Participated in performance of the research. FG: Participated in performance of the research and contributed reagents and analytic tools. JB: Participated in performance of the research and participated in data analysis. SC, CL, and HC: Participated in writing of the paper and data analysis. JH: Participated in research design and writing of the paper, contributed reagents and analytic tools, and participated in data analysis. All authors contributed to the article and approved the submitted version.

 Conflict of Interest

NT, TV, and SC were employed by the company CareDx.

The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

 Generative AI Statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

 Acknowledgements

We thank the Transplant Immunology laboratory at the University of Manitoba, particularly Denise Pochinco, and CareDx Inc in Brisbane, California for reagent and technical support provided for urine CXCL10 and plasma donor-derived cell-free DNA measurement, respectively. We also thank Dr Ian Gibson (IG, Department of Pathology, University of Manitoba) and Dr Mathieu Latour (Department of Pathology, Université de Montréal) for helpful discussions. We also thank Dr Robert Woodward (Chief Scientific Officer, CareDx) and Dr Peale Chuang (Kidney Transplant Medical Director, CareDx) for manuscript review and feedback. We also thank Belinda Abidi at the Hopital Maisonneuve-Rosemont and Stéphanie Lariviere-Beaudoin, Dr Francis Migneault, Dr Julie Turgeon, and Dr Marie Josée Hebert at the Centre de Recherche du CHUM (CRCHUM) for administrative and logistical support.

 Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontierspartnerships.org/articles/10.3389/ti.2026.15517/full#supplementary-material

References Lentine KL Smith JM Lyden GR Miller JM Dolan TG Bradbrook K OPTN/SRTR 2022 Annual Data Report: Kidney. Am J Transpl (2024) 24(2S1):S19–S118. 10.1016/j.ajt.2024.01.012 38431360 El-Zoghby ZM Stegall MD Lager DJ Kremers WK Amer H Gloor JM Identifying Specific Causes of Kidney Allograft Loss. Am J Transpl (2009) 9(3):527–35. 10.1111/j.1600-6143.2008.02519.x 19191769 Loheac C Aubert O Loupy A Legendre C . Identifying the Specific Causes of Kidney Allograft Loss: A Population-based Study. Nephrol Ther (2018) 14(Suppl. 1):S39–S50. 10.1016/j.nephro.2018.02.018 29606262 Van Loon E Senev A Lerut E Coemans M Callemeyn J Van Keer JM Assessing the Complex Causes of Kidney Allograft Loss. Transplantation (2020) 104(12):2557–66. 10.1097/TP.0000000000003192 32091487 Sellares J de Freitas DG Mengel M Reeve J Einecke G Sis B Understanding the Causes of Kidney Transplant Failure: The Dominant Role of Antibody-Mediated Rejection and Nonadherence. Am J Transpl (2012) 12(2):388–99. 10.1111/j.1600-6143.2011.03840.x 22081892 Nankivell BJ Shingde M Keung KL Fung CLS Borrows RJ O'Connell PJ The Causes, Significance and Consequences of Inflammatory Fibrosis in Kidney Transplantation: The Banff i-IFTA Lesion. Am J Transpl (2018) 18(2):364–76. 10.1111/ajt.14609 29194971 Lefaucheur C Louis K Morris AB Taupin JL Nickerson P Tambur AR Clinical Recommendations for Posttransplant Assessment of Anti-HLA (Human Leukocyte Antigen) Donor-Specific Antibodies: A Sensitization in Transplantation: Assessment of Risk Consensus Document. Am J Transpl (2023) 23(1):115–32. 10.1016/j.ajt.2022.11.013 36695614 Blydt-Hansen TD Sharma A Gibson IW Wiebe C Langlois V Rush D Validity and Utility of Urinary CXCL10/Cr Immune Monitoring in Pediatric Kidney Transplant Recipients. Am J Transpl (2021) 21(4):1545–55. 10.1111/ajt.16336 33034126 Van Loon E Tinel C de Loor H Bossuyt X Callemeyn J Coemans M Automated Urinary Chemokine Assays for Noninvasive Detection of Kidney Transplant Rejection: A Prospective Cohort Study. Am J Kidney Dis (2024) 83(4):467–76. 10.1053/j.ajkd.2023.07.022 37777058 Halloran PF Reeve J Madill-Thomsen KS Demko Z Prewett A Gauthier P Antibody-Mediated Rejection Without Detectable Donor-specific Antibody Releases Donor-Derived Cell-Free DNA: Results from the Trifecta Study. Transplantation (2023) 107(3):709–19. 10.1097/TP.0000000000004324 36190186 Naesens M Roufosse C Haas M Lefaucheur C Mannon RB Adam BA The Banff 2022 Kidney Meeting Report: Re-Appraisal of Microvascular Inflammation and the Role of Biopsy-Based Transcript Diagnostics. Am J Transpl (2023) 24:338–49. 10.1016/j.ajt.2023.10.016 38032300 Parajuli S Reville PK Ellis TM Djamali A Mandelbrot DA . Utility of Protocol Kidney Biopsies for De Novo Donor-Specific Antibodies. Am J Transpl (2017) 17(12):3210–8. 10.1111/ajt.14466 28805293 Huang E Mengel M Clahsen-van Groningen MC Jackson AM . Diagnostic Potential of Minimally Invasive Biomarkers: A Biopsy-Centered Viewpoint from the Banff Minimally Invasive Diagnostics Working Group. Transplantation (2023) 107(1):45–52. 10.1097/TP.0000000000004339 36508645 Hirt-Minkowski P Schaub S . Urine CXCL10 as a Biomarker in Kidney Transplantation. Curr Opin Organ Transpl (2024) 29(2):138–43. 10.1097/MOT.0000000000001135 38235748 Jackson JA Kim EJ Begley B Cheeseman J Harden T Perez SD Urinary Chemokines CXCL9 and CXCL10 Are Noninvasive Markers of Renal Allograft Rejection and BK Viral Infection. Am J Transpl (2011) 11(10):2228–34. 10.1111/j.1600-6143.2011.03680.x 21812928 Weseslindtner L Hedman L Wang Y Strassl R Helanterä I Aberle SW Longitudinal Assessment of the CXCL10 Blood and Urine Concentration in Kidney Transplant Recipients with BK Polyomavirus Replication-a Retrospective Study. Transpl Int (2020) 33(5):555–66. 10.1111/tri.13584 31981424 Hricik DE Nickerson P Formica RN Poggio ED Rush D Newell KA Multicenter Validation of Urinary CXCL9 as a Risk-Stratifying Biomarker for Kidney Transplant Injury. Am J Transpl (2013) 13(10):2634–44. 10.1111/ajt.12426 23968332 Ho J Rush DN Karpinski M Storsley L Gibson IW Bestland J Validation of Urinary CXCL10 as a Marker of Borderline, Subclinical, and Clinical Tubulitis. Transplantation (2011) 92(8):878–82. 10.1097/TP.0b013e31822d4de1 21876477 Schaub S Nickerson P Rush D Mayr M Hess C Golian M Urinary CXCL9 and CXCL10 Levels Correlate with the Extent of Subclinical Tubulitis. Am J Transpl (2009) 9(6):1347–53. 10.1111/j.1600-6143.2009.02645.x 19459809 Hirt-Minkowski P Amico P Ho J Gao A Bestland J Hopfer H Detection of Clinical and Subclinical Tubulo-Interstitial Inflammation by the Urinary CXCL10 Chemokine in a Real-Life Setting. Am J Transpl (2012) 12(7):1811–23. 10.1111/j.1600-6143.2012.03999.x 22390571 Matz M Beyer J Wunsch D Mashreghi MF Seiler M Pratschke J Early Post-Transplant Urinary IP-10 Expression After Kidney Transplantation Is Predictive of Short- and Long-Term Graft Function. Research Support, Non-u.s. Gov'T. Kidney International (2006) 69(9):1683–90. 10.1038/sj.ki.5000343 16572110 Hirt-Minkowski P Ho J Gao A Amico P Koller MT Hopfer H Prediction of Long-Term Renal Allograft Outcome by Early Urinary CXCL10 Chemokine Levels. Transpl Direct (2015) 1(8):e31. 10.1097/TXD.0000000000000537 27500231 Hauser IA Spiegler S Kiss E Gauer S Sichler O Scheuermann EH Prediction of Acute Renal Allograft Rejection by Urinary Monokine Induced by IFN-Gamma (MIG). J Am Soc Nephrol (2005) 16(6):1849–58. 10.1681/ASN.2004100836 15857922 Ho J Schaub S Wiebe C Gao A Wehmeier C Koller MT Urinary CXCL10 Chemokine Is Associated with Alloimmune and Virus Compartment-Specific Renal Allograft Inflammation. Transplantation (2018) 102(3):521–9. 10.1097/TP.0000000000001931 28902772 Hirt-Minkowski P Handschin J Stampf S Hopfer H Menter T Senn L Randomized Trial to Assess the Clinical Utility of Renal Allograft Monitoring by Urine CXCL10 Chemokine. J Am Soc Nephrol (2023) 34(8):1456–69. 10.1681/ASN.0000000000000160 37228005 Gupta G Athreya A Kataria A . Biomarkers in Kidney Transplantation: A Rapidly Evolving Landscape. Transplantation (2024) 109:418–27. 10.1097/TP.0000000000005122 39020463 Jahr S Hentze H Englisch S Hardt D Fackelmayer FO Hesch RD DNA Fragments in the Blood Plasma of Cancer Patients: Quantitations and Evidence for Their Origin from Apoptotic and Necrotic Cells. Cancer Res (2001) 61(4):1659–65.11245480 Bloom RD Bromberg JS Poggio ED Bunnapradist S Langone AJ Sood P Cell-Free DNA and Active Rejection in Kidney Allografts. J Am Soc Nephrol (2017) 28(7):2221–32. 10.1681/ASN.2016091034 28280140 Aubert O Ursule-Dufait C Brousse R Gueguen J Racapé M Raynaud M Cell-Free DNA for the Detection of Kidney Allograft Rejection. Nat Med (2024) 30(8):2320–7. 10.1038/s41591-024-03087-3 38824959 Sigdel TK Archila FA Constantin T Prins SA Liberto J Damm I Optimizing Detection of Kidney Transplant Injury by Assessment of Donor-Derived Cell-free DNA Via Massively Multiplex PCR. J Clin Med (2018) 8(1):1–17. 10.3390/jcm8010019 30583588 Xiao H Gao F Pang Q Xia Q Zeng X Peng J Diagnostic Accuracy of Donor-Derived Cell-Free DNA in Renal-Allograft Rejection: A Meta-Analysis. Transplantation (2021) 105(6):1303–10. 10.1097/tp.0000000000003443 32890130 Bu L Gupta G Pai A Anand S Stites E Moinuddin I Clinical Outcomes From the Assessing Donor-Derived Cell-Free DNA Monitoring Insights of Kidney Allografts with Longitudinal Surveillance (ADMIRAL) Study. Kidney International (2022) 101(4):793–803. 10.1016/j.kint.2021.11.034 34953773 Bromberg JS Bunnapradist S Samaniego-Picota M Anand S Stites E Gauthier P Elevation of Donor-Derived Cell-Free DNA Before Biopsy-Proven Rejection in Kidney Transplant. Transplantation (2024) 108(9):1994–2004. 10.1097/TP.0000000000005007 38595232 Parajuli S Garg N Dodin B Breyer I Zona E Patel S Changes in Donor-Derived Cell-Free DNA Before and After Rejection and De Novo DSA Detection in Primary and Repeat Kidney Transplant Recipients. Clin Transpl (2024) 38(11):e70019. 10.1111/ctr.70019 39498966 Huang E Sethi S Peng A Najjar R Mirocha J Haas M Early Clinical Experience Using Donor-Derived Cell-Free DNA to Detect Rejection in Kidney Transplant Recipients. Am J Transpl (2019) 19(6):1663–70. 10.1111/ajt.15289 30725531 Benning L Morath C Fink A Rudek M Speer C Kälble F Donor-Derived Cell-Free DNA (dd-cfDNA) in Kidney Transplant Recipients with Indication Biopsy-Results of a Prospective Single-Center Trial. Transpl Int (2023) 36:11899. 10.3389/ti.2023.11899 38020751 Ho J Schaub S Jackson AM Balshaw R Carroll R Cun S Multicenter Validation of a Urine CXCL10 Assay for Noninvasive Monitoring of Renal Transplants. Transplantation (2023) 107(7):1630–41. 10.1097/TP.0000000000004554 36949034 Ho J Sharma A Kroeker K Carroll R De Serres S Gibson IW Multicentre Randomised Controlled Trial Protocol of Urine CXCL10 Monitoring Strategy in Kidney Transplant Recipients. BMJ Open (2019) 9(4):e024908. 10.1136/bmjopen-2018-024908 30975673 Obrisca B Butiu M Sibulesky L Bakthavatsalam R Smith KD Gimferrer I Combining Donor-Derived Cell-Free DNA and Donor Specific Antibody Testing as Non-Invasive Biomarkers for Rejection in Kidney Transplantation. Sci Rep (2022) 12(1):15061. 10.1038/s41598-022-19017-7 36064740 Bromberg JS Brennan DC Taber DJ Cooper M Anand S Akalin E Donor-Derived Cell-Free DNA Significantly Improves Rejection Yield in Kidney Transplant Biopsies. Am J Transpl (2025) 25(12):2529–42. 10.1016/j.ajt.2025.07.2484 40780562 Loupy A Certain A Tangprasertchai NS Racapé M Ursule-Dufait C Benbadi K Evaluation of a Decentralized Donor-Derived Cell-Free DNA Assay for Kidney Allograft Rejection Monitoring. Transpl Int (2024) 37:13919. 10.3389/ti.2024.13919 39741495 Tambur AR Campbell P Claas FH Feng S Gebel HM Jackson AM Sensitization in Transplantation: Assessment of Risk (STAR) 2017 Working Group Meeting Report. Am J Transpl (2018) 18(7):1604–14. 10.1111/ajt.14752 29603613 Heinze G Schemper M . A Solution to the Problem of Separation in Logistic Regression. Stat Med (2002) 21(16):2409–19. 10.1002/sim.1047 12210625 Park S Sellares J Tinel C Anglicheau D Bestard O Friedewald JJ . European Society of Organ Transplantation Consensus Statement on Testing for Non-Invasive Diagnosis of Kidney Allograft Rejection. Transpl Int (2023) 36:12115. 10.3389/ti.2023.12115 38239762 Mayer KA Omic H Weseslindtner L Doberer K Reindl-Schwaighofer R Viard T Levels of Donor-Derived Cell-Free DNA and Chemokines in BK polyomavirus-associated Nephropathy. Clin Transpl (2022) 36(11):e14785. 10.1111/ctr.14785 35894263 Macioniene E Simanavicius M Vitkauskaite M Vickienė A Staučė R Vinikovas A Urinary Chemokines CXCL9 and CXCL10 Are Non-Invasive Biomarkers of Kidney Transplant Rejection. Ann Transpl (2024) 29:e944762. 10.12659/AOT.944762 39402819 Rabant M Amrouche L Lebreton X Aulagnon F Benon A Sauvaget V Urinary C-X-C Motif Chemokine 10 Independently Improves the Noninvasive Diagnosis of Antibody-Mediated Kidney Allograft Rejection. J Am Soc Nephrol (2015) 26(11):2840–51. 10.1681/ASN.2014080797 25948873 Goutaudier V Aubert O Racape M Truchot A Sablik M Raynaud M Detection of Kidney Allograft Rejection Using Urinary Chemokines. J Am Soc Nephrol (2025) 36:2228–40. 10.1681/ASN.0000000742 40504617 Rampersad C Balshaw R Gibson IW Ho J Shaw J Karpinski M The Negative Impact of T Cell-Mediated Rejection on Renal Allograft Survival in the Modern Era. Am J Transpl (2022) 22(3):761–71. 10.1111/ajt.16883 34717048 Wiebe C Rush DN Gibson IW Pochinco D Birk PE Goldberg A Evidence for the Alloimmune Basis and Prognostic Significance of Borderline T Cell-Mediated Rejection. Am J Transpl (2020) 20(9):2499–508. 10.1111/ajt.15860 32185878 Bouatou Y Viglietti D Pievani D Louis K Duong Van Huyen JP Rabant M Response to Treatment and long-term Outcomes in Kidney Transplant Recipients with Acute T Cell-Mediated Rejection. Am J Transpl (2019) 19(7):1972–88. 10.1111/ajt.15299 30748089 Sablik M Sannier A Raynaud M Goutaudier V Divard G Astor BC Microvascular Inflammation of Kidney Allografts and Clinical Outcomes. N Engl J Med (2024) 392:763–76. 10.1056/NEJMoa2408835 39450752 Tinel C Sauvaget V Aouni L Lamarthée B Terzi F Legendre C Transforming Kidney Transplant Monitoring with Urine CXCL9 and CXCL10: Practical Clinical Implementation. Sci Rep (2024) 14(1):20357. 10.1038/s41598-024-70390-x 39223175