Detection of Clinical and Subclinical Tubulo-Interstitial Inflammation by the Urinary CXCL10 Chemokine in a Real-Life Setting

Authors


Stefan Schaub, schaubs@uhbs.ch

Abstract

Urinary CXCL10 is a promising noninvasive biomarker for tubulo-interstitial allograft inflammation, but its diagnostic characteristics have not been assessed in a real-life setting. We investigated urinary CXCL10 in 213 consecutive renal allograft recipients having 362 surveillance biopsies at 3/6 months and 80 indication biopsies within the first year posttransplant. Allograft histology results were classified as (i) acute Banff score zero, (ii) interstitial infiltrates only, (iii) tubulitis t1, (iv) tubulitis t2–3 and (v) isolated vascular compartment inflammation. For clinical and subclinical pathologies, urinary CXCL10 correlated well with the extent of tubulo-interstitial inflammation. To determine diagnostic characteristics of urinary CXCL10, histological groups were separated into two categories: no relevant inflammation (i.e. acute Banff score zero and interstitial infiltrates only) versus all other pathologies (i.e. tubulitis t1–3 and isolated vascular compartment inflammation). For subclinical pathologies, AUC was 0.69 (sensitivity 61%, specificity 72%); for clinical pathologies, AUC was 0.74 (sensitivity 63%, specificity 80%). A urinary CXCL10-guided biopsy strategy would have reduced performance of surveillance and indication biopsies by 61% and 64%, respectively. Missed (sub)clinical pathologies were mostly tubulitis t1 and isolated vascular compartment lesions. In real life, urinary CXCL10 had clinically useful diagnostic properties making it a candidate biomarker to guide allograft biopsies.

Abbreviations: 
AUC

area under the curve

CXCL10

CXC chemokine ligand 10

Introduction

Even using current immunosuppressive regimens rejection is still the most important cause for renal allograft loss (1). At the same time, the clinical presentation of renal allograft rejection has changed dramatically. While the incidence of renal allograft rejection with functional impairment assessed by serum creatinine (i.e. clinical allograft rejection) decreased to 10–15% using tacrolimus-based immunosuppression (2), subclinical rejection including borderline changes occurs in 10–30% of patients within the first year posttransplant (3–6). As persisting subclinical rejection is associated with progression to chronic irreversible damage and slow deterioration of allograft function (5,7–9), its detection is important allowing for timely therapeutic interventions (10).

However, it is a matter of debate in which patients and at which time points posttransplant surveillance biopsies should ideally be performed (11). Clearly, a noninvasive screening biomarker correlating well with subclinical rejection would be very helpful to guide performance of surveillance biopsies. Several studies have demonstrated that the urinary CXCL10 chemokine is a promising biomarker for this purpose. Indeed, urinary CXCL10 predicts acute clinical allograft inflammation (i.e. rejection and polyomavirus-BK nephropathy) (12–14). Furthermore, it correlated with the extent of subclinical tubulitis in two independent patient cohorts (15,16).

A key limitation of these studies is that they included selected patient groups (12–16). As a consequence, the frequency of the investigated pathologies did not reflect real life, and thus cut-off determination of urinary CXCL10 for prediction of clinical/subclinical allograft inflammation and calculation of its diagnostic characteristics is problematic. The aim of this study was to investigate urinary CXCL10 as a noninvasive biomarker for clinical and subclinical allograft inflammation in a real-life setting of 213 consecutive patients having 362 surveillance biopsies at 3 and 6 months, as well as 80 indication biopsies within the first year posttransplant.

Materials and Methods

Patient population

Since March 2003, midstream urine samples are obtained immediately before clinically indicated or surveillance biopsies are performed. The study was approved by the Ethics committee of the University of Basel and all participating patients gave written informed consent. From October 2005 to March 2009, 228 consecutive patients received a kidney allograft at our center. These patients represent an independent cohort and were not part of the previously reported biomarker discovery study investigating urinary CXCL10 (16). Fifteen of 228 patients (7%) did not provide any urine samples at the time of allograft biopsies, giving a final population of 213 patients, from whom at least one allograft biopsy with a corresponding urine sample was obtained. Three hundred sixty-two surveillance biopsies with corresponding urine samples were available: 176/213 patients (83%) had a surveillance biopsy at 3 months, 186/213 patients (87%) at 6 months posttransplant, 154/213 patients (72%) had both. In addition, 70/213 patients (33%) had 80 clinically indicated biopsies within the first year posttransplant with corresponding urine samples. Thus, the whole study sample collection consists of 442 allograft biopsies with corresponding urine samples from 213 consecutive patients.

Immunosuppressive regimens

Initial immunosuppression was selected based on the presence/absence of donor-specific HLA-antibodies (HLA-DSA), ABO-blood group compatibility and HLA-matching (4).

Recipients of an HLA-identical allograft did not receive induction therapy. They started on a triple immunosuppression with tacrolimus (Tac; Prograf, Astellas), mycophenolate-mofetil (MMF; CellCept, Roche, Switzerland) and prednisone (P), which was reduced within the first 3 months to a dual therapy with low dose Tac-MMF (Tac trough levels 4–6 ng/mL).

Recipients of an allograft with one or more HLA-mismatches but no HLA-DSA (i.e. normal risk patient) received induction therapy with basiliximab (Simulect, Novartis Switzerland), and triple therapy either with Tac-MMF-P, or a steroid-free regimen consisting of Tac, mycophenolate-sodium (MPS; Myfortic, Novartis) and sirolimus or everolimus (Rapamune, Wyeth or Certican, Novartis). Immunosuppression was reduced within the first 6 months with the aim to establish a dual therapy in the long-term with Tac-MMF/MPS or sirolimus-MPS or everolimus-MPS.

Recipients of an allograft with HLA-DSA (i.e. high risk patient) received induction therapy consisting of a polyclonal anti-T-lymphocyte globulin (ATG; ATG-Fresenius, Fresenius Medical Care) and intravenous immunoglobulins (IvIg) as reported previously (17,18). Maintenance immunosuppression consisted of Tac-MMF-P.

ABO-incompatible transplants received one dose of rituximab (Mabthera, Roche) 4 weeks prior to the transplant, pretransplant immunoabsorption, induction therapy with basiliximab, and maintenance immunosuppression with Tac-MMF-P as previously reported (19).

Evaluation of allograft biopsies

Clinically indicated allograft biopsies were performed when serum creatinine increased by >20% from baseline. Surveillance biopsies were scheduled at 3 and 6 months posttransplant. All biopsy specimens (two cores obtained with a 16 gauge needle) were evaluated by light microscopy, immunofluorescence (C4d, HLA-DR), and immunohistochemistry (SV40 large T-antigen). Acute and chronic Banff scores were assessed and biopsies were assigned to five groups according to the acute scores (20):

  • 1Acute score zero (i.e. t0 i0 g0 v0 ptc0).
  • 2Interstitial infiltrates only (i.e. t0 i1–3 g0 v0 ptc0).
  • 3Tubulitis t1 plus any other inflammation (i.e. t1 i0–3 g0–3 v0–3 ptc0–3).
  • 4Tubulitis t2–3 plus any other inflammation (i.e. t2–3 i0–3 g0–3 v0–3 ptc0–3).
  • 5Isolated vascular compartment inflammation (i.e. t0 i0–3 g0–3 v0–3 ptc0–3).

Definition of Polyomavirus BK viremia and urinary tract infection

Screening for Polyomavirus BK (BKV) replication was done according to a standard protocol consisting of urine cytology for decoy cells (21). Patients with positive decoy cells were tested for plasma BKV-loads by quantitative real-time PCR at the following visit as described previously (21). Positive BKV-viremia at the time of the biopsy was defined as any detectable BKV replication (i.e. ≥100 copies/mL).

Patients with significant leukocyturia and a positive urine culture equal or less than seven days prior to an allograft biopsy were considered as having a urinary tract infection (UTI) at the time of the biopsy.

Treatment of subclinical allograft inflammation

Patients with subclinical rejection including tubulitis t1 were treated, mostly with steroids and increasing maintenance immunosuppression. Patients with tubulo-interstitial allograft inflammation due to active BKV-infection were mostly treated with reduction of immunosuppression as reported previously in detail (22).

Urine protein analyses

Measurement of total protein (benzethonium chloride method) and creatinine (enzymatic method) were performed on a Modula clinical chemistry analyzer (Roche Diagnostics). Urinary α1-microglobulin (α1m) was measured by ELISA (Beckman-Coulter nephelometry system, Brea, CA, USA). Urinary CXCL10 measurements were performed retrospectively on midstream urine samples (spun to remove cellular elements, stored at −80°C without any additives) with a sandwich ELISA. Briefly, plates were coated overnight at 4°C with 0.1 μg/mL rabbit anti-human CXCL10 polyclonal antibody (Peprotech, Catalog #500-P93), washed and then blocked (blocking buffer: 0.17% BSA, 0.02% NaN3 in 1× PBS). Each sample (50 μL urine) was performed in duplicate at neat, 1:2, 1:4 and 1:8 serial dilutions (dilution buffer: 0.085% BSA, 0.05% Tween 20, 0.02% NaN3 in 1× PBS), incubated overnight at 4°C, and then washed (wash buffer: 0.05% Tween 20, 0.02% NaN3 in 1× PBS). Biotinylated rabbit anti-human CXCL10 secondary polyclonal antibody (Peprotech, Catalog #500-P93Bt) was added at 0.1 μg/mL and incubated overnight at 4°C. Plates were then washed, developed with streptavidin-alkaline phosphatase/PNPP, and read at 405 nm/690 nm (Biotek Synergy 4 microplate reader, Gen 5 software, Fisher Scientific). CXCL10 duplicates demonstrated very good correlation (r2= 0.99). The intra-assay and interassay coefficients of variation were 3.1% and 11%, respectively. The detection limit of the ELISA assay was 1.95 ng/L. Urine samples with CXCL10 concentration below this value (13% of surveillance and 6% of indication biopsies) were included in the analysis as half the detection limit (i.e. 0.975 ng/L). In order to correct for different urine dilution, excretion of urine proteins are given in relation to urine creatinine (i.e. mg or ng protein/mmol creatinine). CXCL10 measurements were performed on blinded samples in Winnipeg (Canada) and thereafter analyzed in Basel (Switzerland).

Statistical analysis

We used JMP software version 8.0 (SAS Institute Inc., Cary, NC, USA) for statistical analysis. For categorical data, Fisher's exact test or Pearson's chi-square test was used. Parametric continuous data were analyzed by Student's t tests. For nonparametric continuous data, the Wilcoxon rank-sum or Kruskal-Wallis rank sum tests were used for analysis. Significant results in the Kruskal-Wallis rank sum test were further analyzed with pair-wise nonparametric tests. A p-value < 0.05 (two-tailed) was considered to indicate statistical significance.

Results

Characteristics of surveillance biopsies

Demographic data of the 362 surveillance biopsies are summarized in Table 1. Among the five histologically defined groups, there were slight differences regarding the maintenance immunosuppression, but all other donor and recipient characteristics were equally distributed. By definition, all acute Banff scores (t, i, v, g, ptc) were significantly different across the groups (all p < 0.0001). However, all chronic Banff scores were similar (p ≥ 0.16). BKV-viremia was observed more often in the tubulitis t1 and tubulitis t2–3 group. Serum creatinine, estimated glomerular filtration rate (eGFR), and the percentage of patients with a low eGFR (i.e. <40 mL/min) was not different across the five groups (p ≥ 0.43). Median donor age was significantly higher in patients with a low eGFR compared to patients with an eGFR > 40 mL/min (62 years vs. 48 years; p < 0.0001), and DGF had occurred more often (39% vs. 17%; p < 0.0001) suggesting that low eGFR is most likely related to preexisting donor kidney damage and/or ischemia–reperfusion injury. Total urine protein/creatinine ratio, and urinary α1m/creatinine ratio were not different across the five groups (p ≥ 0.21).

Table 1.  Characteristics of 362 surveillance biopsies obtained at 3 and 6 months posttransplant, grouped according to histology results
  1. Tac = tacrolimus; MMF = mycophenolate-mofetil; MPS = mycophenolate-sodium; mTOR = sirolimus or everolimus; P = prednisone; ATG = polyclonal anti-thymocyte globulin; IvIg = intravenous immunoglobulins; eGFR = estimated glomerular filtration rate calculated by the MDRD formula.

 Acute score zero (n = 206)Interstitial infiltrates only (n = 37)Tubulitis t1 + any i/g/v/ptc (n = 86)Tubulitis t2-3 + any i/g/v/ptc (n = 21)Isolated vascular compartment inflammation (n = 12)p Level
Recipient
 Age, median (IQR)54 (43–62)55 (45–60)55 (44–63)46 (38–60)58 (47–64)0.36
 Female, n (%)56 (27)12 (32)31 (36)3 (14)7 (58)0.05
Primary disease
 ADPKD34 (16)8 (22)17 (20)7 (33)4 (34)0.26
 Diabetic22 (11)7 (19)11 (13)1 (8) 
 Vascular19 (9)3 (8)7 (8)2 (10)1 (8) 
 Glomerulopathy82 (40)13 (35)22 (26)5 (24)5 (42) 
 Other49 (24)6 (16)29 (33)7 (33)1 (8) 
Immunological risk
 Normal risk, n (%)164 (80)29 (78)62 (72)19 (91)7 (58)0.13
 HLA-DSA, n (%)34 (16)5 (14)16 (19)2 (9)5 (42) 
 ABOi, n (%)8 (4)3 (8)8 (9) 
Induction therapy
 Basiliximab, n (%)153 (75)31 (84)69 (80)19 (90)7 (58)0.30
 ATG +/– IvIg, n (%)46 (22)6 (16)16 (19)2 (10)5 (42) 
 None, n (%)7 (3)1 (1) 
Baseline IS
 Tac-MMF-P, n (%)116 (56)20 (54)42 (49)8 (38)8 (67)0.04
 Tac-MPS-mTOR, n (%)86 (42)17 (46)37 (43)10 (48)3 (25) 
 Other, n (%)4 (2)7 (8)3 (14)1 (8) 
Donor
 Age, median (IQR)53 (45–63)53 (40–62)53 (43–63)48 (35–61)60 (45–67)0.42
 Deceased donor, n (%)105 (51)21 (57)44 (51)14 (67)9 (75)0.34
 DGF, n (%)44 (21)13 (35)17 (20)5 (24)5 (42)0.18
 HLA-A-B-DR MM, n with 0/1/2/3/4/5/69/12/25/59/44/46/111 /0/3/9/8/13/32/0/11/16/22/23/120/2/2/6/7/3/10/2/1/1/4/1/30.07
Surveillance biopsy
 Month 3/6, n107/9919/1841/466/153/90.13
 Days post-TRP, median (IQR)130 (94–185)101 (93–188)173 (93–190)182 (103–197)182 (118–195)0.08
 Acute score zero (n = 206)Interstitial infiltrates only (n = 37)Tubulitis t1 + any i/g/v/ptc (n = 86)Tubulitis t2-3 + any i/g/v/ptc (n = 21)Isolated vascular compartment inflammation (n = 12)p Level
Prior clinical rejection, n (%)12 (6)1 (3)9 (10)1 (5)0.37
Allograft histology
 Glomeruli, median (IQR)17 (13–25)12 (11–21)18 (13–25)19 (12–27)18 (16–23)0.16
 Acute i-score, n with 0/1/2/3206/0/0/00/32/4/15/66/12/30/4/12/59/3/0/0<0.0001
 Acute t-score, n with 0/1/2/3206/0/0/037/0/0/00/86/0/00/0/16/512/0/0/0<0.0001
 Acute v-score, n with 0/1/2/3206/0/0/037/0/0/080/6/0/018/2/1/O5/7/0/0<0.0001
 Acute g-score, n with 0/1/2/3206/0/0/037/0/0/075/9/2/021/0/0/09/3/0/0<0.0001
 Acute ptc-score, n with 0/1/2/3206/0/0/037/0/0/075/10/1 /021/0/0/09/3/0/0<0.0001
 Chronic i-score, n with 0/1/2/3121/70/13/220/14/0/346/31/7/212/6/3/06/4/2/00.19
 Chronic t-score, n with 0/1/2/3124/68/12/221/13/0/346/32/6/211/7/3/06/4/2/00.16
 Chronic v-score, n with 0/1/2/3162/65/14/127/9/1/050/30/5/114/5/1/16/4/1/10.37
 Chronic g-score, n with 0/1/2/3204/2/0/037/0/0/086/0/0/021/0/0/011/0/1/O0.33
 C4d positive, n (%)17 (8)8 (22)17 (20)03 (25)0.004
 BKV-viremia, n (%)7 (3)4 (11)15 (17)9 (43)0<0.0001
Allograft function
 Serum creatinine, median (IQR)135 (112–165)134 (105–158)132 (107–169)153 (123–212)135 (94–180)0.43
 eGFR, median (IQR)47 (39–58)51 (45–59)47 (37–58)43 (31–57)48 (36–58)0.57
 eGFR <40ml/min, n (%)60 (29)8 (22)25 (29)7 (33)4 (33)0.87
UTI, n (%)1 (0.5)4 (5)1 (5)0.07
Urine protein
 Prot/creat ratio, median (IQR)13 (8–21)13 (10–24)14 (9–24)12 (8–19)12 (8–15)0.40
 α1m/creat ratio, median (IQR)4.1 (2.5–8.1)6.2 (3.7–8.5)4.6 (3.0–7.6)6.5 (3.2–10.8)4.7 (1.8–5.4)0.21
 CXCL10/creat ratio, median (IQR)0.65 (0.35–1.58)1.19 (0.59–3.33)1.89 (0.66–6.66)5.50 (1.41–14.29)1.74 (0.45–3.83)<0.0001

Urinary CXCL10 and subclinical allograft inflammation

Urinary CXCL10/creatinine ratios were significantly different among the five groups (p < 0.0001; Table 1). Median CXCL10/creatinine ratio was lowest in the acute score zero group (0.65 ng/mmol) with a stepwise increase to the interstitial infiltrates only group (1.19 ng/mmol; p = 0.004), the tubulitis t1 group (1.89 ng/mmol; p < 0.0001) and the tubulitis t2–3 group (5.5 ng/mmol; p < 0.0001; Figure 1A). The isolated vascular compartment inflammation group had a higher median urinary CXCL10/creatinine ratio (1.74 ng/mmol) than the acute score zero group, but this did not reach statistical significance (p = 0.07; Figure 1A).

Figure 1.

Correlation of urinary CXCL10/creatinine ratio with allograft histology. (A) Subclinical pathologies (n = 362). (B) Clinical pathologies (n = 80). Biopsies with BKV-viremia and/or urinary tract infection (UTI) are marked with colored circles.

When biopsies with concomitant BKV-viremia and/or UTI were excluded from the analysis (n = 39), urinary CXCL10/creatinine ratios were still different across the five groups (p = 0.0002). The acute score zero group had the lowest median urinary CXCL10/creatinine ratio (0.63 ng/mmol), with a stepwise increase to the interstitial infiltrates only group (1.12 ng/mmol; p = 0.01), the tubulitis t1 group (1.25 ng/mmol; p = 0.0005) and the tubulitis t2–3 group (1.92 ng/mmol; p = 0.009). The isolated vascular compartment inflammation group had a higher median urinary CXCL10/creatinine ratio (1.74 ng/mmol) than the acute score zero group, but this did not reach statistical significance (p = 0.05).

Diagnostic characteristics of urinary CXCL10 for detection of subclinical allograft inflammation

For this analysis the five groups were separated into two categories. Category one contained the acute score zero and the interstitial infiltrates only group (i.e. no relevant inflammation category; n = 243). The second category comprised the tubulitis t1, the tubulitis t2–3 and the isolated vascular compartment inflammation group (i.e. inflammation category; n = 119). The prevalence of subclinical allograft inflammation was therefore 33% (i.e. 119/362 surveillance biopsies). The demographic data of these two categories are summarized in Table 2. Serum creatinine, eGFR, total urine protein/creatinine ratio, and urinary α1m/creatinine ratio were not different among the two categories (p ≥ 0.32), while urinary CXCL10/creatinine ratio was significantly higher in the second group (median 2.15 ng/mmol vs. 0.72 ng/mmol; p < 0.0001).

Table 2.  Characteristics of 362 surveillance biopsies grouped into the “no relevant inflammation” category (i.e. acute Banff score zero and interstitial infiltrates only; n = 243) and the “inflammation” category (i.e. Tubulitis t1 and Tubulitis t2–3 and isolated vascular compartment inflammation; n = 119)
 No relevant inflammation (n = 243)Inflammation (n = 119)p Level
Recipient
 Age, median (IQR)55 (43–61)55 (44–62)0.55
 Female, n (%)68 (28)41 (34)0.22
Primary disease
  ADPKD42 (17)28 (24)0.11
  Diabetic29 (12)12 (10) 
  Vascular22 (9)10 (8) 
  Glomerulopathy95 (39)32 (27) 
  Other55 (23)37 (31) 
Immunological risk
  Normal risk, n (%)193 (79)88 (74)0.46
  HLA-DSA, n (%)39 (16)23 (19) 
  ABOİ, n (%)11 (5)8 (7) 
Induction therapy
  Basiliximab, n (%)184 (76)95 (80)0.40
  ATG +/− Ivlg, n (%)52 (21)23 (19) 
  None, n (%)7 (3)1 (1) 
Baseline IS
  Tac-MMF-P, n (%)136 (56)58 (49)0.003
  Tac-MPS-mTOR, n (%)103 (42)50 (42) 
  Other, n (%)4 (2)11 (9) 
Donor
 Age, median (IQR)53 (44–63)54 (42–63)0.86
 Deceased donor, n (%)126 (52)67 (56)0.43
 DGF, n (%)57 (23)27 (23)0.90
 HLA-A-B-DR MM, n with 0/1/2/3/4/5/610/12/28/68/52/59/142/4/14/23/33/27/160.08
Surveillance biopsy
 Month 3/6, n126/11750/690.09
 Days post-TRP, median (IQR)128 (94–185)179 (97–191)0.03
Prior clinical rejection, n (%)13 (5)10 (8)0.26
Allograft histology
 Glomeruli, median (IQR)17 (12–25)18 (13–25)0.32
 Acute i-score, n with 0/1/2/3206/32/4/114/73/24/6<0.0001
 Acute t-score, n with 0/1/2/312/86/16/5<0.0001
 Acute v-score, n with 0/1/2/3103/15/1/0<0.0001
 Acute g-score, n with 0/1/2/3105/12/2/0<0.0001
 Acute ptc-score, n with 0/1/2/3105/13/0/1<0.0001
 Chronic i-score, n with 0/1/2/3141/84/13/564/41/12/20.41
 Chronic t-score, n with 0/1/2/3145/81/12/563/43/11/20.36
 Chronic v-score, n with 0/1/2/3153/74/15/170/39/7/30.31
 Chronic g-score, n with 0/1/2/3241/2/0/0118/0/1/00.22
 C4d positive, n (%)25 (10)20 (17)0.09
 BKV-viremia, n (%)11 (5)24 (20)<0.0001
Allograft function
 Serum creatinine, median (IQR)135 (111–163)134 (108–174)0.74
 eGFR, median (IQR)47 (39–58)46 (36–57)0.32
UTI, n (%)1 (0.4)5 (4.4)0.02
Urine protein
 Prot/creat ratio, median (IQR)13 (8–22)14 (9–21)0.34
 α1m/creat ratio, median (IQR)4.5 (2.6–8.1)4.9 (3.0–8.4)0.48
 CXCL10/creat ratio, median (IQR)0.72 (0.37–1.80)2.15 (0.70–6.72)<0.0001

Receiver–operator characteristic (ROC) analysis revealed an area under the curve (AUC) of 0.69 for urinary CXCL10 for detection of subclinical inflammation (p < 0.0001). By contrast, eGFR, total urine protein/creatinine ratio and urinary α1m/creatinine ratio had AUC ≤ 0.53 (p ≥ 0.33). The optimal cut-off for urinary CXCL10/creatinine ratio was 1.535 ng/mmol with a sensitivity of 61% and a specificity of 72%. Sensitivities, specificities and likelihood ratios of urinary CXCL10 at different cut-offs are summarized in Table 3. For each given cut-off we estimated the consequences if a urinary CXCL10-based strategy would have been used to guide surveillance biopsies. At the optimal cut-off of urinary CXCL10, only 140/362 surveillance biopsies (39%) would have been performed. Furthermore, only 67/243 surveillance biopsies (28%) demonstrating no relevant inflammation would have been performed (i.e. excess biopsies). However, 46/119 surveillance biopsies with allograft inflammation (39%) would have been missed, of which 37 were tubulitis t1 (80%), five were tubulitis t2–3 (11%) and four were isolated vascular compartment inflammations (9%). More specifically, 37/86 subclinical tubulitis t1 (43%), 5/21 subclinical tubulitis t2–3 (24%) and 4/12 subclinical isolated vascular compartment inflammations (33%) would have been missed (Table 3).

Table 3.  Diagnostic performance of urinary CXCL10/creatinine ratio for detection of subclinical inflammation at different cut-off levels. The last three columns indicate, how a urinary CXCL10-guided surveillance biopsy strategy would have (i) changed the frequency of performed surveillance biopsies, (ii) changed the frequency of “excess” surveillance biopsies (i.e. with no relevant inflammation) and (iii) missed subclinical pathologies
Cut-off (ng/mmol)SensSpecLR+ (95%CI)LR (95%CI)Performed surveillance biopsies, n of 362 (%)Performed biopsies with no relevant inflammation, n of 243 (%)Missed subclinical pathologies below cut-off
  1. Sens = sensitivity; Spec = specificity; LR+ (95% CI) = likelihood ratio of a positive test results with 95% confidence interval; LR (95% CI) = likelihood ratio of a negative test results with 95% confidence interval. For any prevalence of subclinical allograft inflammation (Prev; given as a value between 0 and 1) the positive posttest probability (PPP; numerically equal to the positive predictive value) can be calculated as follows: PPP = (Prev*LR+)/(Prev*LR+– Prev + 1). Accordingly, the negative posttest probability (NPP; numerically equal to (1 – the negative predictive value)) is: NPP = (Prev*LR)/(Prev*LR– Prev + 1).

0.581%37%1.3 (1.1–1.5)0.5 (0.3–0.8)249 (69)153 (63)Tubulitis t2–3 (2/21; 10%)Isolated vascular compartment inflammation (3/12; 25%)
1.067%59%1.7 (1.4–2.0)0.6 (0.4–0.7)179 (49)99 (41)Tubulitis t2–3 (4/21; 19%)Isolated vascular compartment inflammation (4/12; 33%)
1.53561%72%2.2 (1.7–2.9)0.5 (0.4–0.7)140 (39)67 (28)Tubulitis t2–3 (5/21; 24%)Isolated vascular compartment inflammation (4/12; 33%)
251%78%2.4 (1.7–3.2)0.6 (0.5–0.8)114 (31)53 (22)Tubulitis t2–3 (6/21; 29%)Isolated vascular compartment inflammation (7/12; 58%)
436%90%3.8 (2.4–6.0)0.7 (0.6–0.8)66 (18)23 (9)Tubulitis t2–3 (9/21; 43%)Isolated vascular compartment inflammation (9/12; 75%)
628%94%4.5 (2.5–7.9)0.8 (0.7–0.9)48 (13)15 (6)Tubulitis t2–3 (11/21; 52%)Isolated vascular compartment inflammation (11/12; 92%)

Ten of 37 biopsies with missed subclinical tubulitis t1 had additional vascular compartment inflammation (v1g0ptc0 C4d−[n = 3]; isolated g and/or ptc inflammation C4d−[n = 6]; v0g0ptc3 C4d+[n = 1]). Only one of five missed subclinical tubulitis t2–3 had additional vascular compartment inflammation (v1g0ptc0 C4d−). In total, 38/364 surveillance biopsies (15%) demonstrated vascular compartment inflammation (12 had isolated vascular compartment inflammation, 26 had additional tubulitis t1–3). Fifteen of these 38 biopsies with vascular compartment inflammation (39%) would have been missed (v1g0ptc0 Cd4−[n = 7]; isolated g and/or ptc inflammation C4d−[n = 7]; v0g0ptc3 C4d+[n = 1]).

Characteristics of indication biopsies

Demographic data of the 80 indication biopsies are summarized in Table 4. Among the five histologically defined groups, there were significant differences regarding the acute Banff scores and the time posttransplant, at which the biopsies were obtained. BKV-viremia, serum creatinine, eGFR, total urine protein/creatinine ratio and urinary α1m/creatinine ratio were not different across the five groups (p ≥ 0.17). The causes for allograft dysfunction in the 50 biopsies with an acute score zero or interstitial infiltrates only were as follows: drug-related toxicity (n = 25), acute tubular injury (n = 6), postrenal obstruction (n = 3), infection (n = 2), recurrent FSGS (n = 1), multiple etiologies (n = 6) and unknown (n = 7).

Table 4.  Characteristics of 80 indication biopsies obtained within the first year posttransplant, grouped according to histology results
 Acute Banff score zero (n = 40)Interstitial infiltrates only (n = 10)Tubulitis t1 + any i/g/v/ptc (n = 15)Tubulitis t2-3 + any i/g/v/ptc (n = 10)Isolated vascular compartment inflammation (n = 5)p Level
  1. Tac = tacrolimus; MMF = mycophenolate-mofetil; MPS = mycophenolate-sodium; mTOR = sirolimus or everolimus; P = prednisone; ATG = polyclonal anti-thymocyte globulin; IvIg = intravenous immunoglobulins; eGFR = estimated glomerular filtration rate calculated by the MDRD = formula.

Recipient
 Age, median (IQR)56 (40–64)51 (41–60)45 (30–60)57 (35–66)57 (36–59)0.31
 Female, n (%)10 (25)2 (20)3 (20)1 (10)4 (80)0.05
Primary disease
  ADPKD6 (15)2 (20)2 (13)3 (30)2 (40)0.96
  Diabetic5 (13)1 (10)2 (13)1 (20) 
  Vascular3 (7)1 (10)1 (7)1 (10) 
  Glomerulopathy10 (25)4 (40)6 (40)3 (30)1 (20) 
  Other16 (40)2 (20)4 (27)3 (30)1 (20) 
Immunological risk
  Normal risk, n (%)34 (85)9 (90)11 (73)9 (90)3 (60)0.50
  HLA-DSA, n (%)6 (15)1 (10)3 (20)1 (10)2 (40) 
  ABOi, n (%)1 (7) 
Induction therapy
  Basiliximab, n (%)33 (83)9 (90)12 (80)8 (80)3 (60)0.91
  ATG +/− IvIg, n (%)6 (15)1 (10)3 (20)2 (20)2 (40) 
  None, n (%)1 (2) 
Baseline IS
  Tac-MMF-P, n (%)14 (35)2 (20)6 (40)4 (40)2 (40)0.86
  Tac-MPS-mTOR, n (%)24 (60)8 (80)9 (60)5 (50)3 (60) 
  Other, n (%)2 (5)1 (10) 
Donor
 Age, median (IQR)52 (44–63)53 (39–57)48 (38–54)50 (46–60)58 (46–61)0.76
 Deceased donor, n (%)16 (40)2 (20)9 (60)4 (40)3 (60)0.32
 DGF, n (%)10 (25)1 (10)3 (20)1 (10)0.52
 HLA-A-B-DR MM, n with 0/1/2/3/4/5/62/1/6/6/12/7/60/0/1/2/2/5/01/0/2/3/5/3/10/1/3/2/3/0/10/0/2/0/1/2/00.76
BX days post-TRP, median (IQR)45 (23–105)18 (9–69)170 (40–277)223 (100–296)12 (9–27)0.0003
Allograft histology
 Glomeruli, median (IQR)18 (12–21)18 (13–20)15 (11–27)16 (11–17)21 (9–31)0.83
 Acute i-score, n with 0/1/2/340/0/0/00/9/1/02/1112100/2/3/53/2/0/0<0.0001
 Acute t-score, n with 0/1/2/340/0/0/010/0/0/00/15/0/00/0/6/45/0/0/0<0.0001
 Acute v-score, n with 0/1/2/340/0/0/010/0/0/012/1/2/06/3/1/01/4/0/0<0.0001
 Acute g-score, n with 0/1/2/340/0/0/010/0/0/013/1/1/08/2/0/03/2/0/00.008
 Acute ptc-score, n with 0/1/2/340/0/0/010/0/0/012/2/0/110/0/0/02/2/1/00.0001
 Chronic i-score, n with 0/1/2/332/7/1/08/1/1/07/6/1/15/4/1105/0/0/00.25
 Chronic t-score, n with 0/1/2/333/6/1/08/2/0/07/6/1/15/4/1/05/0/0/00.23
 Chronic v-score, n with 0/1/2/326/10/4/08/2/0/010/4/0/17/3/0/04/1/0/00.71
 Chronic g-score, n with 0/1/2/340/0/0/010/0/0/014/1/0/010/0/0/05/0/0/00.36
 C4d positive, n (%)3 (8)2 (20)5 (33)3 (30)3 (60)0.02
 BKV-viremia, n (%)1 (3)2 (13)2 (20)0.17
Allograft function
 Serum creatinine, median (IQR)210 (156–303)199 (130–326)174 (131–351)200 (156–267)215 (155–259)0.98
 eGFR, median (IQR)30 (20–38)33 (20–51)37 (20–52)32 (24–42)22 (21 -39)0.70
UTI, n (%)4 (10)1 (11)0.51
Urine protein
 Prot/creat ratio, median (IQR)24 (11–44)29 (19–129)24 (12–47)21 (11–59)23 (17–52)0.73
 α1m/creat ratio, median (IQR)9.1 (4.4–16.5)6.5 (4.6–15.0)9.0 (4.4–18.3)8.3 (6.2–17.4)8.3 (5.3–13.8)0.93
 CXCL10/creat ratio, median (IQR)1.00 (0.40–2.49)1.42 (0.73–2.41)2.59 (1.08–14.23)11.6 (5.16–29.84)1.30 (0.47–8.47)0.001

Urinary CXCL10 and clinical allograft inflammation

Urinary CXCL10/creatinine ratios were significantly different among the five groups (p = 0.001; Table 4). Median CXCL10/creatinine ratio was lowest in the acute score zero group (1.0 ng/mmol) with a stepwise increase to the interstitial infiltrates only group (1.42 ng/mmol; p = 0.37), the tubulitis t1 group (2.59 ng/mmol; p = 0.02), and the tubulitis t2–3 group (11.6 ng/mmol; p = 0.0001; Figure 1B). The isolated vascular compartment inflammation group had a similar median urinary CXCL10/creatinine ratio (1.30 ng/mmol) as the acute score zero and the interstitial infiltrates only groups (p ≥ 0.59; Figure 1B).

When biopsies with concomitant BKV-viremia and/or UTI were excluded from the analysis (n = 10), urinary CXCL10/creatinine ratios were still different across the five groups (p = 0.003). The acute score zero group had the lowest median urinary CXCL10/creatinine ratio (0.94 ng/mmol), with a stepwise increase to the interstitial infiltrates only group (1.33 ng/mmol; p = 0.48), the tubulitis t1 group (1.74 ng/mmol; p = 0.05) and the tubulitis t2–3 group (11.6 ng/mmol; p = 0.0003). The isolated vascular compartment inflammation group had a similar median urinary CXCL10/creatinine ratio (1.30 ng/mmol) as the acute score zero and interstitial infiltrates only groups (p ≥ 0.50).

Comparing clinical and subclinical pathologies (after exclusion of biopsies with concomitant BKV-viremia and/or UTI), urinary CXCL10/creatinine ratios were not different in the acute score zero (0.94 vs. 0.63; p = 0.14), the interstitial infiltrates only (1.33 vs. 1.12; p = 0.70), the tubulitis t1 (1.74 vs. 1.25; p = 0.22) and the isolated vascular compartment inflammation group (1.30 vs. 1.74; p = 0.92). However, urinary CXCL10/creatinine ratios were higher in clinical than in subclinical tubulitis t2–3 (11.6 vs. 1.92; p = 0.03).

Diagnostic characteristics of urinary CXCL10 for detection of clinical allograft inflammation

As for the analysis regarding subclinical allograft inflammation, the 80 clinical biopsies were divided into two categories (i.e. no relevant inflammation category [n = 50] and inflammation category [n = 30]). The prevalence of clinical allograft inflammation was therefore 37.5% (i.e. 30/80 indication biopsies). Serum creatinine, eGFR, total urine protein/creatinine ratio and urinary α1m/creatinine ratio were not different among the two categories (p ≥ 0.49), while urinary CXCL10/creatinine ratio was significantly higher in the inflammation category (median 3.86 ng/mmol vs. 1.08 ng/mmol; p = 0.0005).

ROC-analysis revealed an AUC of 0.74 for urinary CXCL10 for detection of clinical allograft inflammation (p = 0.01). By contrast, eGFR, total urine protein/creatinine ratio and urinary α1m/creatinine ratio had AUC ≤ 0.55 (p ≥ 0.27). Sensitivities, specificities and likelihood ratios of urinary CXCL10 at different cut-off levels are summarized in Table 5. For each given cut-off we estimated the consequences if a urinary CXCL10-based strategy would have been used to guide indication biopsies. At the optimal cut-off of urinary CXCL10, only 29/80 indication biopsies (36%) would have been performed. Furthermore, only 10/50 clinical biopsies (20%) demonstrating no relevant inflammation would have been performed (i.e. excess biopsies). However, 11/30 indication biopsies with allograft inflammation (37%) would have been missed, of which seven were tubulitis t1 (64%), one was a tubulitis t2–3 (9%) and three were isolated vascular compartment inflammations (27%). More specifically, 7/15 clinical tubulitis t1 (47%), 1/10 clinical tubulitis t2–3 (10%) and 3/5 clinical isolated vascular compartment inflammations (60%) would have been missed (Table 5).

Table 5.  Diagnostic performance of urinary CXCL10/creatinine ratio for detection of clinical inflammation at different cut-off levels. The last three columns indicate, how a urinary CXCL10-guided indication biopsy strategy would have (i) changed the frequency of performed indication biopsies, (ii) changed the frequency of “excess” indication biopsies (i.e. with no relevant inflammation) and (iii) missed clinical pathologies
Cut-off (ng/mmol)SensSpecLR+ (95% CI)LR (95% CI)Performed clinical biopsies, n of 80 (%)Performed biopsies with no relevant inflammation, n of 50 (%)Missed clinical pathologies below cut-off
  1. Sens = sensitivity; Spec = specificity; LR+ (95% CI) likelihood ratio of a positive test results with 95% confidence interval; LR (95% CI) likelihood ratio of a negative test results with 95% confidence interval. For any prevalence of clinical allograft inflammation (Prev; given as a value between 0 and 1) the positive posttest probability (PPP; numerically equal to the positive predictive value) can be calculated as follows: PPP = (Prev*LR+)/(Prev*LR+– Prev + 1). Accordingly, the negative posttest probability (NPP; numerically equal to (1 – the negative predictive value)) is: NPP = (Prev*LR)/(Prev*LR– Prev + 1).

0.590%22%1.2 (1.0–1.4)0.5 (0.1–1.5)66 (83)39 (78)Tubulitis t1 (2/15; 13%) Tubulitis t2–3 (0/10; 0%) Isolated vascular compartment inflammation (1/5; 20%)
183%48%1.6 (1.2–2.2)0.3 (0.1–0.8)51 (64)26 (52)Tubulitis t1 (3/15; 20%) Tubulitis t2–3 (0/10; 0%) Isolated vascular compartment inflammation (2/5; 40%)
2.58663%80%3.2 (1.7–5.9)0.5 (0.3–0.7)29 (36)10 (20)Tubulitis t1 (7/15; 47%) Tubulitis t2–3 (1/10; 10%) Isolated vascular compartment inflammation (3/5; 60%)
640%86%2.9 (1.3–6.5)0.7 (0.5–0.9)19 (24)7 (14)Tubulitis t1 (11/15; 73%) Tubulitis t2–3 (3/10; 30%) Isolated vascular compartment inflammation (4/5; 80%)
1037%90%3.7 (1.4–9.5)0.7 (0.5–0.9)16 (20)5 (10)Tubulitis t1 (11/15; 73%) Tubulitis t2–3 (4/10; 40%) Isolated vascular compartment inflammation (4/5; 80%)

Two of seven biopsies with missed tubulitis t1 had additional vascular compartment inflammation (v0g1tc0 C4d−; v2g0ptc0 C4d+). The only missed tubulitis t2–3 had no vascular compartment inflammation. In total, 16/80 indication biopsies (20%) demonstrated vascular compartment inflammation (5 had isolated vascular compartment inflammation, 11 had additional tubulitis t1–3). Five of these 16 biopsies with vascular compartment inflammation (31%) would have been missed (v2g0ptc0 Cd4+; v1g0ptc1 C4d−; v1g0ptc0 C4d+; v0g1ptc1 C4d+; v0g1ptc0 C4d−).

Discussion

The main observation in this study was that urinary CXCL10 levels correlate with the extent of (sub)clinical tubulo-interstitial inflammation in a real-life setting. This confirms previously reported results and provides for the first time more accurate and clinically applicable diagnostic characteristics of urinary CXCL10 for detection of (sub)clinical allograft inflammation.

The AUC, sensitivity and specificity of urinary CXCL10 for detection of (sub)clinical allograft inflammation is lower in this study than in previous reports (12,14–16). This can mainly be explained by three factors. First, in the current real-life study the prevalence of normal histology (i.e. no relevant inflammation) was around 65%, while this important control group represented less than a third in previous studies (12,14–16). This imbalance can have a critical influence on the calculation of the diagnostic characteristics in that the positive predictive value of urinary CXCL10 will be lower in patient populations with a lower prevalence of (sub)clinical allograft inflammation. Second, the distribution of the various rejection phenotypes was different and variably included borderline rejection. Third, this study included biopsies with isolated vascular compartment inflammation, which correlated poorly with urinary CXCL10 and therefore reduced its diagnostic value.

Even more important than calculation of diagnostic characteristics is how the assay would change performance of allograft biopsies and what pathologies would be missed. For subclinical pathologies, a urinary CXCL10-guided strategy using the optimal cut-off of 1.535 ng/mmol would have reduced surveillance biopsies at 3 and 6 months by 61% from 362 to 140. Most missed subclinical pathologies were tubulitis t1 (37/46; 80%), and only nine biopsies demonstrating tubulitis t2–3 or isolated vascular compartment inflammation would not be detected in the whole cohort of 213 consecutive patients. Clearly, it will be a centre-specific decision, how many missed subclinical pathologies are considered acceptable, balancing the costs and workload of surveillance biopsies and the clinical consequences of untreated subclinical pathology. For clinical pathologies, a urinary CXCL10-guided strategy using the optimal cut-off of 2.586 ng/mmol would have reduced indication biopsies by 64% from 80 to 29. As before, most missed pathologies were tubulitis t1 (7/11; 64%), and only four biopsies that demonstrated tubulitis t2–3 or isolated vascular compartment inflammation would not be detected. Although the number of missed clinical rejections can be further reduced using a lower urinary CXCL10 cut-off (see Table 5), a biopsy should always be considered in patients with allograft dysfunction and a high suspicion of rejection.

As mentioned earlier, most missed clinical and subclinical pathologies were tubulitis t1 (= equivalent to borderline rejection). Although the clinical relevance of borderline rejection is still largely unknown, a recent study showed that even subclinical borderline rejection is associated with subsequent deterioration of allograft function (9). Given the patchy nature of the rejection process and the inherent sampling error of allograft biopsies, incidentally detected borderline rejection might represent irrelevant inflammation, or reflect early relevant rejection progressing to more severe lesions. Furthermore, regulatory T-cells within the infiltrates might be able to critically influence the clinical impact of tubulo-interstitial allograft inflammation (23,24). Interestingly, using molecular analysis of biopsy samples, de Freitas et al. found that 67% of clinical borderline rejection could be reclassified as "nonrejection-like" (25). This is fairly consistent with our observation that 43–47% of tubulitis t1 were not considered as rejection by urinary CXCL10 levels. In addition, Matz et al. found that high urinary CXCL10 levels were associated with inferior allograft function even in the absence of clinical rejection, suggesting that urinary CXCL10 might be a prognostic biomarker beyond allograft histology (26). It will require further studies to evaluate whether urinary CXCL10 levels can help to better delineate the clinical relevance of borderline rejection.

Urinary CXCL10 levels in biopsies with isolated vascular compartment inflammation were in the range of the groups with interstitial infiltrates only and tubulitis t1. Accordingly, they would have been missed in about half of the cases. Although the number of biopsies with clinical and subclinical isolated vascular compartment inflammation was low (i.e. 5/80 [6%] and 12/362 [3%], respectively), the obtained data suggest that urinary CXCL10 does not sufficiently reflect inflammation in the vascular compartment, which is an important limitation of this biomarker. We cannot mechanistically explain this observation, but the most likely reason is that the inflammatory response is mainly restricted to the vascular compartment without a relevant "spill-over" into the urine. Interestingly, other studies investigating biomarkers for T-cell mediated rejection found lower urine levels of granzyme A/B and perforin mRNA in biopsies with predominant vascular inflammation (i.e. Banff II) compared to biopsies with tubulitis Banff Ia/b (27,28).

Previous publications and this study indicate that urinary CXCL10 is a biomarker for tubulo-interstitial inflammation irrespective of the etiology (12,14). Besides allograft rejection the most important disease leading to tubulo-interstitial inflammation is active BKV-infection. Urinary CXCL10 levels in biopsies with BKV-viremia were rather higher than expected from the extent of the associated tubulo-interstitial inflammation. Concomitant UTI might also be a confounder, but this was only observed in 11/442 biopsies (2.5%). Thus, in the absence of active BKV-infection (i.e. no urinary Decoy-cells or BKV-viremia) and UTI, elevated urinary CXCL10 can be reliably attributed to allo-reactive tubulo-interstitial inflammation.

To determine the diagnostic characteristics of urinary CXCL10, we defined the “no relevant inflammation” category very stringently (i.e. at most an interstitial infiltrate) and assigned all other groups including tubulitis t1 to the “inflammation” category. We believe that this is important because a potential noninvasive biomarker will be used in clinics for screening and should ideally separate absence/presence of any active allograft inflammation.

To the best of our knowledge, this is the first study investigating a potential noninvasive biomarker for renal allograft monitoring in a real-life setting. This allows determining the diagnostic characteristics more accurately than in highly selected patient populations. A limitation of the study is that the number of biopsies classified as clinical rejection (n = 30) is low, and they occurred at different time points posttransplant ranging from 5 days to 365 days. As the urinary CXCL10 cut-off was not calculated for specific time frames (e.g. first month posttransplant) due to low sample numbers, it is very likely not applicable to the first few weeks. In addition, urinary CXCL10 is also elevated in patients with acute tubular necrosis due to ischemia–reperfusion injury limiting its ability to predict allograft rejection in the early phase posttransplant (14,29).

In conclusion, we found that in a real-life setting, urinary CXCL10 had clinically useful diagnostic properties to detect tubulo-interstitial allograft inflammation making it a promising biomarker to guide performance of allograft biopsies, but not to replace them. Clearly, the clinical significance of such a noninvasive biomarker driven biopsy strategy has to be evaluated in a prospective study.

Acknowledgments

The authors thank the staff of the renal transplant unit and the histocompatibility laboratory for collection and processing of urine samples. PA is supported by the Swiss Kidney Foundation and the University of Basel. SS is supported by the Swiss National Foundation (grant 32473B_125482/1) and the Novartis foundation for biomedical research. PN holds the Flynn Family Chair in Renal Transplantation at the University of Manitoba. DR and PN are supported by operating grants from the Canadian Institutes of Health Research. JH is supported by the KRESCENT Young Investigator program and the Norman S. Coplon Satellite Healthcare Extramural Grant.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

Ancillary