Urinary miR‐155‐5p and CXCL10 as prognostic and predictive biomarkers of rejection, graft outcome and treatment response in kidney transplantation

Abstract

Aims

MicroRNAs (miRNAs) may be useful biomarkers of rejection and allograft outcome in kidney transplantation. Elevated urinary CXCL10 levels have been associated with acute rejection (AR) and may predict allograft failure. We examined the correlation of miRNA, CXCL10 levels and immunosuppressive drug exposure with AR and graft function in kidney transplant recipients.

Methods

Eighty de novo kidney transplant recipients were recruited from four European centres. All patients received tacrolimus, mycophenolate mofetil, and methylprednisolone. Urinary pellet expression of miR‐142‐3p, miR‐210‐3p and miR‐155‐5p was assessed by quantitative real‐time polymerase chain reaction and urinary CXCL10 levels by enzyme‐linked immunosorbent assay at the 1^st week and the 1^st, 2^nd, 3^rd and 6^th months post‐transplantation.

Results

Eight patients experienced AR. Before and during AR, patients showed a significant increase of urinary miR‐142‐3p, miR‐155‐5p and CXCL10 levels and a decrease of miR‐210‐3p levels. Receiver operating characteristic curve analysis showed that miR‐155‐5p (area under the curve = 0.875; P = 0.046) and CXCL10 (area under the curve = 0.865; P = 0.029) had excellent capacity to discriminate between rejectors and nonrejectors. The optimal cut‐off values for the prognosis of AR were 0.51, with 85% sensitivity and 86% specificity for miR‐155‐5p and 84.73 pg ml^–1, with 84% sensitivity and 80% specificity for CXCL10. miR‐155‐5p and CXCL10 levels correlated with glomerular filtration rate. Levels of both biomarkers normalized after recovery of graft function.

Conclusions

The regular early post‐transplantation monitoring of urinary miR‐155‐5p and CXCL10 can help in the prognosis of AR and graft dysfunction. Large prospective randomized multicentre trials are warranted to refine our cut‐off values and validate the clinical usefulness of these biomarkers.

What is Already Known about this Subject

• Immunosuppressive drug doses need to be adjusted based on side effects and drug levels, aiming for certain target concentrations.

• Long‐term outcome is less than optimal and there is a clear need for noninvasive prognostic markers of rejection and graft function.

• Several biomarkers have shown a promising association with acute rejection. Previous studies mainly analysed these biomarkers at the time of acute rejection, usually in allograft biopsies, and focused on their diagnostic capacity.

What this Study Adds

• A significant increase in urinary levels of both miR‐155‐5p and CXCL10 in the period leading up to the episode of acute rejection, with levels returning to normal after adjustment of the immunosuppressive drug dosage.

• Receiver operating characteristic curve analyses showed that miR‐155‐5p and CXCL10 had excellent capacity to discriminate between rejectors and nonrejectors.

• Post‐transplantation monitoring of miR‐155‐5p and CXCL10 levels could provide reliable prognostic markers of rejection and graft dysfunction and can be a useful tool for personalizing immunosuppressive treatment.

Introduction

In clinical transplantation, immunosuppressive drug dosages are adjusted in response to side effects in combination with measurement of drug levels, aiming for certain target concentrations. Despite good short‐term results, the long‐term outcome is still limited and there is a constant graft attrition rate after the first year, indicating a clear need for noninvasive molecular biomarkers to allow early assessment of the risk of rejection and graft clinical outcome. Several biomarkers have shown promising results in the prognosis of graft rejection and graft dysfunction, as well as in the prediction of personal response 1, 2, 3.

MicroRNAs (miRNAs) are noncoding RNA molecules that play a key role in the regulation of gene expression and can affect immune response 4. While miRNA expression has been mainly a diagnostic tool in transplantation 5, 6, miRNAs have recently been postulated as predictive and prognostic biomarkers of graft outcome 5, 7, 8. Overexpression of miR‐142‐5p in peripheral blood mononuclear cells (PBMCs) was associated with chronic rejection in kidney transplant recipients 9. In addition, miR‐210, a hypoxia‐induced miRNA 10 can modulate the expression of several cytokines, including the proinflammatory interferon‐γ and interleukin‐17 and the anti‐inflammatory transforming growth factor‐β and interleukin‐10 11. miR‐210 has also been identified as a reliable marker of acute rejection (AR) and long‐term graft function 12. miR‐155, a key player in the regulation of adaptive immunity and antibody‐related T‐cell response, is upregulated in multiple immune cell lineages by toll‐like receptor ligands and inflammatory cytokines 13, 14, 15 and is a mediator of inflammatory response 16. Local inflammation associated with rejection is tightly regulated by T helper (Th) balance, and miR‐155 controls differentiation of CD4⁺T cells into Th cells 17, 18 and also participates in the development of regulatory T cells 19. miR‐155 was overexpressed in renal allografts during AR 18, and a study of renal allograft biopsies and normal PBMCs found that several miRNAs, including miR‐155, overexpressed in AR biopsies were also highly expressed in PBMCs. Moreover, intragraft expression of miR‐155 and other miRNAs were reliable prognostic markers of AR and allograft function 20.

Chemokines play an important role in inflammation and immune response after transplantation 21, 22, and chemokine interferon‐inducible protein 10 (CXCL10) is a promising biomarker of short‐ and long‐term kidney graft function 23. Urinary CXCL10 concentrations showed higher sensitivity and specificity than serum creatinine (Cr) concentrations in differentiating between stable kidney transplant patients and those with AR and BK virus infections 24. In another study, urinary CXCL10 levels were linked to risk of AR, and moreover, CXCL10 levels during the first 4 weeks post‐transplantation were associated with graft function at 6 months 23. High urinary CXCL10 levels were also associated with AR from 1 month to 1 year post‐transplantation, and the CXCL10:Cr ratio correlated with AR risk 25.

To explore the clinical utility of miR‐142‐3p, miR‐210‐3p, miR‐155‐5p and CXCL10 further as early prognostic biomarkers, we have assessed their urinary pellet levels in adult kidney transplant recipients from the 1^st week to the 6^th month post‐transplantation and correlated our findings with AR, graft function and immunosuppressive drug exposure.

Methods

Study design and patients

In this prospective observational study, 80 adult kidney transplant patients were recruited from four European centres, two in Spain and two in Germany. Participants had to be de novo kidney transplant recipients from a deceased or living donor, with no other transplanted organs. Recipients older than 70 years or positive for hepatitis B or C or human immunodeficiency virus were excluded. The study was approved by the Ethics Committees of the participating centres and all patients provided their written informed consent (EudraCT number: 2013–001817‐33).

All patients received the same immunosuppressive therapy, adapted to the standard of each centre and the needs of each patient, of tacrolimus (TAC; Prograf; Astellas Pharma Inc, Tokyo, Japan), mycophenolate mofetil (MMF; Myfenax; Teva Pharmaceutical, Petach Tikva, Israel), and methylprednisolone. All patients received induction treatment of two 20‐mg doses of basiliximab. The initial daily TAC dose was 0.1 mg kg^–1 twice daily. The TAC dose was subsequently adjusted to achieve target trough concentrations between 7–10 and 5–8 ng ml^–1 during the first and third months, respectively. MMF was administered 1 g twice daily and dose reductions were allowed based on clinical needs. Treatment with methylprednisolone was initiated pretransplantation at a daily dose of 500 mg, tapered down to 40 mg on day 5, and progressively decreased to 20 mg on day 14, and 5 mg at month 2.

Patients had a total of five visits during the study: during the 1^st week and after the 1^st, 2^nd, 3^rd and 6^th months. Safety parameters were assessed at each visit, including graft function [Cr and glomerular filtration rate (GFR) by the modification of diet in renal disease method 26], blood counts, urinary protein, and routine chemistry. All participating centres used the same criteria for diagnosis of AR based on impairment of renal function as detected by Cr concentrations, Cr clearance rate and GFR. AR was always confirmed by histological evaluation of graft biopsies (biopsy proven acute rejection) according to the Banff 97 classification 27.

Pharmacokinetic monitoring

For each immunosuppressant, we analysed trough concentration, calculated area under the curve (AUC; 0 and 30 min, and 1, 1.5, 2, 3, 4, 6, 8 and 12 h after drug administration by linear trapezoidal rule) at 1^st week post‐transplantation and simplified AUC (predose and 1.5, 2 and 4 h postdose) at the 1^st, 2^nd, 3^rd and 6^th months post‐transplantation. Blood samples were collected in EDTA‐K3 tubes. Whole‐blood TAC concentrations were measured by liquid chromatography/tandem mass spectrometry and mycophenolic acid (MPA) plasma concentrations were measured by a validated method of high‐performance liquid chromatography with ultraviolet detector, as previously reported 28, 29, 30.

Biomarker analyses

First morning urine samples were collected from all patients at each of the five post‐transplantation visits and shipped to the Pharmacology Laboratory of the Biomedical Diagnostic Centre, Hospital Clinic (Barcelona, Spain) for centralized analysis of miRNA expression and CXCL10 production. All samples were analysed in a blinded fashion.

miRNA analysis

miR‐142‐3p, miR‐210‐3p and miR‐155‐5p expression was assessed in urinary cell pellets by quantitative real‐time polymerase chain reaction (qPCR). Urine specimens (50 ml) were collected in presence of EDTA RNAse at 4°C. The urine samples were centrifuged at 2000× g for 10 min at 4°C, the supernatant was discarded, the pellet was centrifuged at 2000× g for 2 min at 4°C, the supernatant was discarded, 1 ml of TRIzol (Life Technologies, Carlsbad, CA, USA) was added, and the sample was stored at –70°C. The sample was sent to the Pharmacology Laboratory of the Biomedical Diagnostic Centre, Hospital Clinic within 3 weeks.

RNA was extracted using TRIzol reagent, according to the manufacturer's instructions, and the concentration and quantity of total RNA were measured at 260 nm and 280 nm (A260/A280) with a NanoDrop (NanoDrop Technologies, Wilmington, DE, USA). Total RNA (100 ng) was reverse transcribed into cDNA. qPCR for miRNAs was performed using the miRCURY LNA Universal RT microRNA PCR, Polyadenylation and cDNA synthesis system (Exiqon, Denmark). All samples were processed in duplicate on 96 well plates and run on a Roche LightCycler 480 Real‐Time PCR System. Negative controls with 1 μg MS2 carrier RNA as mock template from the reverse transcription reaction were performed and profiled like the samples. The amplification curves were analysed using Roche LC Software for determination of Cq by the second derivative method. Average Cq values were normalized to the stably expressed reference miR‐103. First, the Cq values for all samples were determined and the ∆Cq was calculated as the difference in Cq values between the miRNA target and the reference control; relative expression levels of target miRNAs were then evaluated within a sample according to the formula 2^–∆Cq, where high values corresponded to higher expression.

CXCL10 production

For the analysis of CXCL10 production, urine samples were centrifuged at 2000× g during 10 min and supernatant was stored at –70°C and shipped to Pharmacology Laboratory of the Biomedical Diagnostic Centre, Hospital Clinic for batched analysis. Concentrations of CXCL10 (pg ml^–1) were measured by enzyme‐linked immunosorbent assay kit (R&D Systems, Minneapolis, MN, USA), according to the manufacturer's instructions. All samples were processed in duplicate. The minimum detectable dose of CXCL10 was 1.67 pg ml^–1. CXCL10:Cr ratios were also determined.

Statistical analyses

Demographic data and results of the molecular analyses were collected in a unified database. Samples were adjusted to a nonparametric distribution. Statistical differences between groups were assessed with the Mann–Whitney test and correlations between biomarkers and clinical events with Spearman's rho. We used the receiver operating characteristic (ROC) curve analysis to define the optimal cut‐offs for differentiating between patient groups with and without AR. Because, in our case, the purpose of these ROC curves was to propose prognostic biomarkers for the assessment of the risk of acute rejection during the first 6 months post‐transplantation, we included data points from nonrejectors and from rejectors before the rejection episode occurred. Optimal biomarker cut‐off points to discriminate between patients with and without AR were based on ROC curves and were calculated with the best Youden index (sensitivity + specificity – 1). Discriminatory capacity was defined by the AUC (0.7–0.8, acceptable; 0.8–0.9, excellent; >0.9, outstanding), with its 95% confidence interval (CI). All analyses were performed using SPSS 18.0 software (SPSS Inc, Chicago, IL, USA). All data are presented as the median ± standard deviation. A value of P ≤ 0.05 was considered statistically significant. To better evaluate not only the diagnostic capacity of the biomarkers evaluated in this study, but also their prognostic utility, we have included in the AR box‐plots data from the patients who rejected at this time plus the pre‐AR data of the patients who had not yet rejected at this time but who would do so in a later profile. We have not considered data from rejector patients, once the AR episode was solved, in the AR‐Box. The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology 31.

Results

Patients

Eight of the 80 patients (10%) experienced a total of 12 AR episodes, all of which were cellular rejections. Diagnosis of AR was based on clinical and laboratory findings and was confirmed by histological evaluation of graft biopsies. Table 1 displays the characteristics of rejectors and nonrejectors. Three episodes occurred during the 1^st week post‐transplantation, four during the 1^st month, and one during the 6^th month. Four patients had multiple AR episodes: one patient who rejected during the 1^st week post‐transplantation suffered a second episode at 3 months; three patients who rejected during the 1^st month post‐transplantation suffered second episodes – two during the 2^nd month and one during the 6^th month. In all patients who rejected, urine for biomarker analysis was always collected before the immunosuppressive treatment was modified to resolve the AR episode.

Table 1.

Characteristics of 80 kidney transplant recipients

		Kidney transplant recipients n = 80
		Nonrejectors n = 72	Rejectors n = 8
Recipient age (years)		49 ± 12.4a	48 ± 12.7a
Donor age (years)		53 ± 13a	57.5 ± 16.1a
Cold ischaemia time (h)		7.94 ± 5.8a	11 ± 6.33a
Type of donor	Living	45	1
	Deceased	27	7
Sex	Male	34	4
	Female	38	4
Immunosuppressive regimen	TAC + MMF+ methylprednisolone	72	8
Induction therapy	Basiliximab	72	8
Pretransplant disease	Polycystic kidney disease	17	2
	Glomerulonephritis	17	2
	IgA nephropathy	0	1
	Diabetes	7	0
	Single kidney with chronic pyelonephritis	0	1
	Chronic kidney disease	2	0
	Other	29	2
Type of acute rejection	Cellular	–	8
	Antibody‐mediated	–	0
Acute tubular necrosis		0	0
Post‐transplantation infections	Cytomegalovirus	13	2
	BK virus	8	1
	Bacterial	13	3

TAC, tacrolimus; MMF, mycophenolate mofetil

^amedian ± standard deviation

Table 2 summarizes the GFR (stratified by KDOQI‐CKD grades 32, and plasma Cr concentrations from the 1^st week to the 6^th month post‐transplantation in rejectors and nonrejectors. Patients who rejected had GFR < 30 ml min^–1 at the time of the AR episode and GFR < 40 ml min^–1 throughout the post‐transplantation period, indicating poor kidney function. Patients who did not reject had better kidney function during the entire period.

Table 2.

Glomerular filtration rate (GFR) and plasma creatinine (Cr) levels in nonrejectors and rejectors during the 6‐month post‐transplantation period

Nonrejectors (n = 72)
GFR	1st Week	1st Month	2nd Month	3rd Month	6th Month	Total
<15 ml min –1 a	10	2	1	0	0	13
15–29 ml min –1 a	22	4	1	4	1	32
30–59 ml min –1 a	35	49	38	36	36	194
60–89 ml min –1 a	4	14	29	29	32	108
>90 ml min –1 a	1	3	3	3	3	13
Plasma Cr (μmol l –1 ) b	195.36 ± 99.18	124.49 ± 63.54	127.23 ± 45.71	122.88 ± 59.67	117.57 ± 34.88

Rejectors (n = 8)
GFRc	1st Week	1st Month	2nd Month	3rd Month	6th Month	Total
<15 ml min –1 a	7	3	2	1	1	14
15–29 ml min –1 a	1	5	4	4	4	18
30–59 ml min –1 a	0	0	2	3	3	8
60–89 ml min –1 a	0	0	0	0	0	0
>90 ml min –1 a	0	0	0	0	0	0
Plasma Cr (μmol l –1 ) b	567.60 ± 108.74	271.04 ± 130.9	278.96 ± 102.01	224.40 ± 103.17	221.76 ± 65.38

^aNumbers indicate number of patients at each GFR level.

^bPlasma Cr concentrations are shown as median ± standard deviation. There was a significant difference in concentrations between nonrejectors vs. rejectors.

^cRejectors had GFR < 40 ml min^–1 throughout the post‐transplantation period.

Pharmacokinetics

Trough concentration and AUC values for TAC and MPA are summarized in Table 3. During the 1^st week post‐transplantation, TAC trough concentrations were 7–10 ng ml^–1 (within target range) in 26 (32.5%) patients, <7 ng ml^–1 in 22 (27.5%) patients, and >10 ng ml^–1 in 32 (40%) patients. At 1 month post‐transplantation, TAC trough concentrations were within target range in 33 (41.3%) patients, <7 ng ml^–1 in only four (5%) patients and >10 ng ml^–1 in 43 (53.7%) patients. Interestingly, three (43%) of the seven patients with an AR episode before the end of the 1^st month post‐transplantation had significantly lower TAC trough concentrations (<7 ng ml^–1) than the remaining four (57%; >7 ng ml^–1). During the 1^st week and 1^st month post‐transplantation, rejectors showed a clear tendency towards lower TAC trough concentrations and AUC than nonrejectors, but the difference was not significant, probably due to the relatively small number of rejectors.

Table 3.

Tacrolimus (TAC) and mycophenolic acid (MPA) trough concentration and area under the curve (AUC) values

1st Week	Nonrejectors n = 77	Rejectors n = 3
TAC dose (mg day –1 )	15.5 ± 4.7	13 ± 2.87
Cmin TAC (ng ml –1 ) *	9.13 ± 3.8	6.65 ± 3.3
TAC AUC (ng h ml –1 )	143.64 ± 71.54	96.71 ± 35.74
MPA dose (mg day –1 )	2000 ± 243	2000 ± 200
Cmin MPA (μg ml –1 )	2.72 ± 2.4	3.31 ± 3.03
MPA AUC (μg h ml –1 )	27.19 ± 19	44.07 ± 22.42

1st Month	Nonrejectors n = 76	Rejectors n = 4
TAC dose (mg day –1 )	12 ± 4.4	8.5 ± 6.78
Cmin TAC (ng ml –1 )	11.42 ± 4.7	8.95 ± 3.59
TAC AUC (ng h ml –1 )	82.13 ± 41.69	40.62 ± 22.45
MPA dose (mg day –1 )	1500 ± 390	2000 ± 250
Cmin MPA (μg ml –1 )	2.79 ± 2.0	5.24 ± 1.78
MPA AUC (μg h ml –1 )	19.33 ± 14.4	33.03 ± 26.8

2nd Month	Nonrejectors n = 76	Rejectors n = 2
TAC dose (mg day –1 )	9.00 ± 3.9	8.5 ± 5.1
Cmin TAC (ng ml –1 )	9.57 ± 3.7	8.1 ± 2.7
TAC AUC (ng h ml –1 )	81.46 ± 52.08	62.92 ± 53.8
MPA dose (mg day –1 )	2000 ± 500	2000 ± 390
Cmin MPA (μg ml –1 )	5.94 ± 2.1	3.4 ± 2.9
MPA AUC (μg h ml –1 )	17.29 ± 8.56	18.63 ± 10.12

3rd Month	Nonrejectors n = 79	Rejectors n = 1
TAC dose (mg day –1 )	7.0 ± 3.4	7.5
Cmin TAC (ng ml –1 )	9.6 ± 4.6	7.2
TAC AUC (ng h ml –1 )	75.26 ± 30.27	53.46
MPA dose (mg day –1 )	1500 ± 600	1500
Cmin MPA (μg ml –1 )	2.5 ± 1.7	4.4
MPA AUC (μg h ml –1 )	16.84 ± 9.64	35.41

6th Month	Nonrejectors n = 78	Rejectors n = 2
TAC dose (mg day –1 )	5.0 ± 2.8	8.25 ± 3.1
Cmin TAC (ng ml –1 )	8.4 ± 2.6	9.5 ± 2.7
TAC AUC (ng h ml –1 )	64.47 ± 26.99	72.24 ± 14.77
MPA dose (mg day –1 )	1000 ± 600	1500 ± 400
Cmin MPA (μg ml –1 ) *	2.1 ± 1.4	5.1 ± 2.8
MPA AUC (μg h ml –1 ) *	15.86 ± 11.02	45.60 ± 16.01

^* significant difference between nonrejectors and rejectors

From the 3^rd to the 6^th month post‐transplantation, only 10% of patients had TAC trough concentrations <5 ng ml^–1 and all three patients who rejected during this time had TAC trough concentrations of 7–10 ng ml^–1. No statistically significant differences between rejectors and nonrejectors were observed in trough concentrations for TAC or MPA during this period.

Patients with poorer kidney function, as indicated by GFR, did not have higher TAC concentrations and no correlation was observed between TAC trough concentration or AUC and GFR.

MPA trough concentrations were similar in nonrejectors and rejectors (Table 3). The need for MMF dose reduction was similar in rejectors and nonrejectors. In fact, MMF doses and MPA trough concentrations showed a tendency to be higher in rejectors than in nonrejectors.

miRNA expression and AR

Before and during an AR episode, rejectors had a significant gradual upregulation of miR‐142‐3p and miR‐155‐5p expression and lower miR‐210‐3p expression than nonrejectors (Figure 1A‐C).

Figure 1.

Correlation of post‐transplantation urinary miRNA expression and acute rejection (AR). Differences between rejectors (white boxes) and nonrejectors (grey boxes) in (A) miR‐142‐3p, (B) miR‐210‐3p, and (C) miR‐155‐5p expression during the first six months post‐transplantation. Eight of the 80 patients experienced a total of 12 AR episodes during the first 6 months post‐transplantation (n = 3 1^st week; n = 4 1^st month; n = 2 2^nd month; n = 1 3^rd month and n = 2 6^th month). AR box‐plots include data from the patients who rejected at this time plus the pre‐AR data of the patients who had not yet rejected at this time but who would do so in a later profile. Therefore, the number of samples that contributed to the data for both groups in each profile is the following: Nonrejectors 1^st week n = 68; 1^st month n = 71; 2^nd month n = 75; 3^rd month n = 77 and 6^th month n = 78 vs. Rejectors 1^st week n = 12; 1^st month n = 9; 2^nd month n = 5; 3^rd month n = 3 and 6^th month n = 2. Asterisks indicate significant differences between the two groups. (D) Receiver operating characteristic (ROC) curve analysis for discrimination between rejectors and nonrejectors. We have considered 31 data points for the rejector group (12 at 1^st week +9 at 1^st month +5 at 2^nd month +3 at 3^rd month and 2 at 6^th month post transplantation) and 369 data points for the nonrejector group (68 at 1^st week +71 at 1^st month +75 at 2^nd month +77 at 3^rd month and 78 at 6^th month post transplantation). High post‐transplantation miR‐155‐5p expression was associated with AR: area under the curve: 0.875 (95% CI: 0.784 to 0.966); cutoff = 0.51 with 85% sensitivity and 86% specificity

ROC curve analysis showed that miR‐155‐5p had excellent capacity to discriminate between rejectors and nonrejectors (AUC 0.875; 95% CI = 0.784–0.966; Figure 1D). The optimal cut‐off value for predicting AR based on the AUC of the ROC curve for miR‐155‐5p was 0.51, with 85% sensitivity, 86% specificity, 88% positive predictive value (PPV) and 100% negative predictive value (NPV). miR142‐3p (AUC 0.584) and miR‐210‐3p (AUC 0.185) showed less than acceptable discriminatory capacity and were thus not included in further analyses.

Figure 2A shows the evolution of miR‐155‐5p levels in the eight rejectors prior to, during and after the AR episodes. miR‐155‐5p expression progressively increased preceding the AR episode and reached maximum levels at the time of the episode. Once the immunosuppressive treatment was modified to resolve the AR episode, miR‐155‐5p levels decreased. At the time of the AR episode, all rejectors had miR‐155‐5p levels above the established cut‐off point of 0.51. Furthermore, in two patients (patients 5 and 13) with a second AR episode, miR‐155‐5p levels again increased prior to the second episode and reached levels similar to those attained during the first episode.

Figure 2.

Evolution of miR‐155‐5p expression and urinary CXCL10 levels in eight patients experiencing acute rejection (AR). The cutoff of 0.51 for miR‐155‐5p expression is shown on each graph of Figure 2A. Red arrows indicate the time of the AR episode. In three patients, the first AR episode occurred during the 1^st week after transplantation (patients 12, 13, 28), in four, it occurred during the 1^st month (patients 5, 6, 7, 22), and in one, it occurred during the 6^th month (patient 8). Four patients had a second AR episode: two at the 2^nd month (patients 7, 22), one at the 3^rd month (patient 13), and one at the 6^th month (patient 5). The cutoff of 84.73 pg ml^–1 for CXCL10 levels is shown on each graph of Figure 2B. Red arrows indicate the time of the AR episode. In three patients, the first AR episode occurred during the 1^st week after transplantation (patients 12, 13, 28), in four, it occurred during the 1^st month (patients 5, 6, 7, 22) and, in one, it occurred during the 6^th month (patient 8). Four patients had a second AR episode: two at the 2^nd month (patients 7, 22), one at the 3^rd month (patient 13), and one at the 6^th month (patient 5)

Urinary CXCL10 and serum creatinine levels and AR

Patients who experienced an AR episode had consistently higher CXCL10 levels than nonrejectors throughout the study period (Figure 3A). ROC curve analysis showed that CXCL10 levels had excellent capacity to discriminate between rejectors and nonrejectors (AUC 0.865; 95% CI 0.809–0.921). The optimal cut‐off for predicting AR based on the AUC of the ROC curve for CXCL10 levels was 84.73 pg ml^–1, with 84% sensitivity, 80% specificity, 90% PPV and 85% NPV (Figure 3D).

Figure 3.

Correlation of post‐transplantation urinary CXCL10, Creatinine and CXCL10:Cr levels and acute rejection (AR). (A) Differences in CXCL10 levels between rejectors (white boxes) and nonrejectors (grey boxes). Asterisks indicate significant differences between the two groups. Eight of the 80 patients experienced a total of 12 AR episodes during the first 6 months post‐transplantation (n = 3 1^st week; n = 4 1^st month; n = 2 2^nd month; n = 1 3^rd month and n = 2 6^th month). AR box‐plots include data from the patients who rejected at this time plus the pre‐AR data of the patients who had not yet rejected at this time but who would do so in a later profile. Therefore, the number of samples that contributed to the data for both groups in each profile is the following: Nonrejectors 1^st week n = 68; 1^st month n = 71; 2^nd month n = 75; 3^rd month n = 77 and 6^th month n = 78 vs. Rejectors 1^st week n = 12; 1^st month n = 9; 2^nd month n = 5; 3^rd month n = 3 and 6^th month n = 2. (D) Receiver operating characteristic curve analysis for discrimination between rejectors and nonrejectors. We have considered 31 data points for the rejector group (12 at 1^st week +9 at 1^st month +5 at 2^nd month +3 at 3^rd month and 2 at 6^th month post transplantation) and 369 data points for the nonrejector group (68 at 1^st week +71 at 1^st month +75 at 2^nd month +77 at 3^rd month and 78 at 6^th month post transplantation). High post‐transplantation CXCL10 levels were associated with AR [area under the curve (AUC) 0.865; 95% CI 0.809–0.921; cutoff = 84.73 with 84% sensitivity and 80% specificity]. (B) Differences in creatinine levels between rejectors (white boxes) and nonrejectors (grey boxes). Asterisks indicate significant differences between the two groups. (E) High post‐transplantation creatinine levels were associated with AR (AUC 0.784; 95% CI 0.718–0.850; cutoff = 431.5 with 46% sensitivity and 80% specificity). (C) Differences in CXCL10:Cr ratios between rejectors (white boxes) and nonrejectors (grey boxes). Asterisks indicate significant differences between the two groups. (F) High post‐transplantation CXCL10:Cr ratios were associated with AR (AUC 0.753; 95% CI 0.670–0.836; cutoff = 0.43 with 72% sensitivity and 73% specificity)

By contrast, the results showed that, pretransplantation, no significant differences were observed in serum creatinine levels between rejectors and nonrejectors in and that, post‐transplantation, creatinine levels were significantly higher in rejectors than in patients free of rejection (Figure 3B). ROC curve analysis showed that creatinine levels had an acceptable capacity to discriminate between rejectors and nonrejectors (AUC 0.784; 95% CI 0.718–0.850). The optimal cut‐off for predicting AR by creatinine levels was 431.5 μmol l^–1, with 46% sensitivity, 80% specificity, 63% PPV and 96% NPV (Figure 3E).

The CXCL10:Cr ratio was significantly higher in rejectors than in nonrejectors at the 1^st, 2^nd, 3^rd and 6^th month post‐transplantation (Figure 3C); however, no significant differences were observed at the 1^st week. ROC curve analysis showed that the CXCL:Cr ratio had acceptable capacity to discriminate between rejectors and nonrejectors (AUC 0.753; 95% CI 0.670–0.836). The optimal cut‐off for predicting AR based on the AUC of the ROC curve for the CXCL10:Cr ratio was 0.43, with 72% sensitivity, 73% specificity, 90% PPV and 96% NPV (Figure 3F).

Figures 2B and 4 show changes in CXCL‐10 and creatinine levels in the eight rejectors prior to, during and after the AR episodes. CXCL‐10 showed a similar pattern to that of miR‐155‐5p expression, with levels progressively increasing before the AR episode and reaching a maximum at the time of the episode. At the time of the AR episode, CXCL‐10 levels were above the established cut‐off point of 84.73 pg ml^–1 in all rejectors. Creatinine levels, however, did not progressively increase before the AR episode (no prognostic capacity) and AR did not always take place when creatinine reached maximum levels.

Figure 4.

Evolution of creatinine levels in eight patients experiencing acute rejection (AR). The cut‐off of 431.5 μmol l^–1 for creatinine levels is shown on each graph. Red arrows indicate the time of the AR episode. In three patients, the first AR episode occurred during the 1^st week after transplantation (patients 12, 13, 28), in four, it occurred during the 1^st month (patients 5, 6, 7, 22), and in one, it occurred during the 6^th month (patient 8). Four patients had a second AR episode: two at the 2^nd month (patients 7, 22), one at the 3^rd month (patient 13), and one at the 6^th month (patient 5)

miR‐155‐5p and CXCL10 levels and graft function

Figure 5 displays the GFRs, taken as an indication of kidney function, for rejectors and nonrejectors at each of the five time points of the study. Significant inverse correlations were observed between miR‐155‐5p expression and GFR (ρ = –754; Figure 5C) and between CXCL10 levels and GFR (ρ = –0.387; Figure 5D). Patients with GFR > 30 ml min^–1 had CXCL10 levels below the established cut‐off of 84.73 pg ml^–1. At the time of the AR episode, rejectors had GFR < 29 ml min^–1 and CXCL10 levels above the cut‐off of 84.73 pg ml^–1. No significant correlation was observed between the CXCL10:Cr ratio and GFR (ρ = 0.056).

Figure 5.

Glomerular filtration rate (GFR units ml min^-1) in rejectors and nonrejectors and correlation with miR‐155‐5p and CXCL10 levels. GFR was stratified based KDOQI CKD classification as: GF < 15 as an end‐stage kidney‐failure; a 15 < GF < 29 as severe decrease in renal function; 30 < GF < 59 as moderate decrease in kidney function and 60 < GF < 90 mild decreases in kidney function. (A) Differences in GFR between rejectors (white boxes) and nonrejectors (grey boxes). Eight of the 80 patients experienced a total of 12 AR episodes during the first 6 months post‐transplantation (n = 3 1^st week; n = 4 1^st month; n = 2 2^nd month; n = 1 3^rd month and n = 2 6^th month). AR box‐plots include data from the patients who rejected at this time plus the pre‐AR data of the patients who had not yet rejected at this time but who would do so in a later profile. Therefore, the number of samples that contributed to the data for both groups in each profile is the following: Nonrejectors 1^st week n = 68; 1^st month n = 71; 2^nd month n = 75; 3^rd month n = 77 and 6^th month n = 78 vs. Rejectors 1^st week n = 12; 1^st month n = 9; 2^nd month n = 5; 3^rd month n = 3 and 6^th month n = 2. Asterisks indicate significant differences between the two groups. (B) Differences in GFR at the time of the AR episode between rejectors and nonrejectors (n = 3 vs. n = 77 at 1^st week; n = 4 vs. n = 76 at 1^st month; n = 2 vs. n = 78 at 2^nd month; n = 1 vs. n = 79 at 3^rd month; n = 2 vs. n = 78 at 6^th month). Four patients had one AR episode and four had two. (C) miR‐155‐5p expression (transparent boxes) and GFR (grey boxes; ρ = –754). The dotted line indicates the cut‐off of 0.51 for miR‐155‐5p expression. (D) CXCL10 levels (white boxes) and GFR (grey boxes; ρ = –0.387). The dotted line represents the cut‐off of 84.73 pg ml^–1 for CXCL10 levels

miR‐155‐5p expression, drug concentrations and clinical outcome

As expected, a high interindividual variability in drug exposure was observed. According to the study design, TAC trough concentrations were targeted at 7–10 ng ml^–1 during the 1st month post‐transplantation. Of the 80 patients included, 24 had TAC trough concentrations <7 ng ml^–1 during this period: 20 at 1 week and four at 1 month post‐transplantation.

Six of the seven patients experiencing an AR episode during the 1^st month post‐transplantation had Cmin TAC < 9 ng ml^–1. Two had Cmin TAC < 7 ng ml^–1at 1 week post‐transplantation and four had Cmin TAC < 9 ng ml^–1 for the entire 1^st month (Figure 6A). Four of the patients with AR had Cmin MPA < 3.5 μg ml^–1 and three had Cmin MPA > 3.5 μg ml^–1 (Figure 6B). Despite having attained Cmin MPA ≈ 3.5 μg ml^–1, patients with AR generally had Cmin TAC < 9 ng ml^–1. An inverse correlation was observed between TAC trough concentrations and miR‐155‐5p expression but without reaching statistical significance (ρ = –0.088). Figure 6C shows data pairs for TAC – miR‐155‐5p and MPA – miR‐155‐5p for each patient during the 1^st month post‐transplantation.

Figure 6.

Correlation of tacrolimus (TAC) and mycophenolic acid (MPA) trough concentrations with miR‐155‐5p expression. Scatter diagrams showing the distribution of (A) Cmin TAC concentrations and (B) Cmin MPA concentrations, in relation to miR‐155‐5p expression levels, during the 1^st month post‐transplantation. Solid vertical lines indicate the standard clinical cut‐offs for (A) TAC trough concentrations (9 ng ml^–1) and (B) MPA trough concentrations (3.5 μg ml^–1). Dotted horizontal lines represent the cut‐off value (0.51) for miR‐155‐5p expression; patients with miR‐155‐5p > 0.51 were considered to be at high risk of acute rejection (AR). Each dot represents one patient. Red dots indicate patients with an AR episode. (C) Scatter diagram showing data pairs for TAC – miR155‐5p (red squares) and MPA – miR‐155‐5p (blue squares) for each patient during the 1^st month post‐transplantation. White squares indicate patients experiencing acute rejection. The solid vertical line indicates the cut‐off value (0.51) for miR‐155‐5p expression. Broken horizontal lines indicate the cut‐off values for TAC (9 ng ml^–1, red) and MPA (3.5 μg ml^–1, blue)

To evaluate the relationship between drug exposure biomarkers and the risk of AR, we performed additional pharmacodynamic–pharmacokinetic assessments. Patients with TAC AUC values >180 ng h ml^–1 (Figure 7A) and MPA AUC values >50 μg h ml^–1 (Figure 7B) before the end of the 1^st month post‐transplantation had miRNA‐155‐5p expression under the cut‐off established in this study for the risk of AR; all were free of AR (patients at low risk of AR). Moreover, all patients who experienced AR had TAC AUC values <100 ng h ml^–1 during the AR episode (Figure 7A). There were also eight patients with TAC AUC values of 100–180 ng h ml^–1 (Figure 7A) and miRNA‐155‐5p expression above the cut‐off established in this study for risk of AR. These patients could therefore be considered at high risk for AR (Figure 7A); however, they were rejection‐free during the entire study period, suggesting the importance of adequate TAC exposure in such patients. These patients also had better preserved kidney function (median GFR = 53 ml min^–1) in comparison with rejectors (GFR <29 ml min^–1). Five patients with MPA AUC values <30 μg h ml^–1 and two patients with MPA AUC values 40–50 μg h ml^–1 and high miR‐155‐5p levels (Figure 7B) experienced AR. Figure 7C shows data pairs for TAC AUC – miR‐155‐5p and MPA AUC – miR‐155‐5p for each patient during the 1^st month post‐transplantation. Rejectors had the lowest TAC and MPA AUC values and the highest miR‐155‐5p expression (above the cut off value). We also evaluated the possible relationship between drug concentrations and urinary CXCL‐10 production, but no correlation was observed between drug exposure and this biomarker.

Figure 7.

Correlation of miR‐155‐5p expression with tacrolimus (TAC) and mycophenolic acid (MPA) area under the curve (AUC) values. Scatter diagrams showing the distribution of AUC values for (A) TAC concentrations (7A) and (B) MPA concentrations, in relation to miR‐155‐5p expression levels, during the 1^st month post‐transplantation. Solid vertical lines indicate the selected cut‐offs for (A) TAC AUC (180 ng h ml^–1) and (B) MPA AUC (50 mg h ml^–1). Dotted horizontal lines represent the cut‐off value (0.51) for miR‐155‐5p expression; patients with miR‐155‐5p > 0.51 were considered to be at high risk of acute rejection (AR). Each dot represents one patient. Red dots indicate patients with an AR episode. (C) Scatter diagram showing data pairs for TAC AUC – miR155‐5p (red squares) and MPA AUC – miR‐155‐5p (blue squares) for each patient during the 1^st month post‐transplantation. White squares indicate patients experiencing acute rejection. The solid vertical line indicates the cut‐off value (0.51) for miR‐155‐5p expression. Broken horizontal lines indicate the cut‐off values for TAC AUC (180 ng h ml^–1, red) and MPA AUC (50 mg h ml^–1, blue)

Fifteen patients had cytomegalovirus infections, two of whom experienced an AR episode, and nine had BK virus infections, one of whom had an AR episode. In all three cases, the infections occurred after the AR episodes. There was no significant association between these infections and miRNA or CXCL10 levels.

Discussion

To the best of our knowledge, this European multicentre study is the first to demonstrate that urinary pellet miRNA expression levels could be useful prognostic and diagnostic biomarkers for AR during the early post‐transplantation period (from the 1^st week to the 6^th month). Specifically, we found that miR‐142‐3p and miR‐155‐5p levels were higher and miR‐210‐3p levels were lower in rejectors compared to nonrejectors. In addition, our results confirm previous findings that urinary CXCL10 levels could also be a useful prognostic biomarker of AR. Our study is also the first to examine a potential association between these biomarkers and immunosuppressive drug exposure.

Data on the prognostic impact of miR‐155‐5p and CXCL10 levels on the risk of rejection in kidney transplantation are limited. Previous studies 6, 20, 33 mainly analysed these biomarkers at the time of an AR episode, usually in allograft biopsies, and focused on their diagnostic capacity. Our results, in contrast, show an early significant increase in both urinary pellet miR‐155‐5p and urinary CXCL10 levels in the period leading up to the AR episode, clearly showing their potential utility as prognostic biomarkers, with levels returning to normal after adjustment of the immunosuppressive drug dosage to resolve the AR episode. In addition, ROC curve analyses showed that both miR‐155‐5p (AUC 0.875) and CXCL10 (AUC 0.865) levels had excellent capacity to discriminate between rejectors and nonrejectors. When we compared the performance of these biomarkers with measurement of serum creatinine concentrations, urinary pellet miR‐155 expression and urinary CXCL‐10 levels offered a strong advantage over creatinine levels in terms of early prognostic and diagnostic capacity for the assessment of risk of acute rejection in renal transplant recipients. Furthermore, in the early post‐transplant period, creatinine can be increased due of the inflammation/tissue injury associated to ischemia reperfusion process and as a consequence may be less sensible for the prognosis of the risk of acute rejection.

Our results are in line with previous reports on the potential of miRNAs as diagnostic biomarkers of AR 6, 20, 33, 34. However, the data reported so far on the expression of miR‐155 as a diagnostic biomarker of rejection were based on its assessment in PBMC or biopsies 6, 20 and not in urine. A recent study in 35 kidney transplant patients found that miR‐142‐5p, miR‐142‐3p, miR‐155 and miR‐223 in biopsy samples and miR‐142‐3p and miR‐223 in PBMCs could discriminate recipients with AR from those with normal allograft function 33. One of the few studies of urinary miRNA expression 12 evaluated miR‐10a, miR‐10b and miR‐210 levels in urine samples obtained at the time of allograft biopsy and found that only miR‐210 was able to discriminate between patients with AR and stable transplant patients or transplant patients before or after AR.

Several studies have found a robust association between CXCL10 levels and renal allograft outcome 23, 24, 25 participates in the recruitment of alloantigen primed T cells to the site of inflammation and during the induction of proinflammatory cytokines 35. It induces, maintains and amplifies both inflammatory and immune responses and plays a critical role in AR 36, 37. Our results are also in line with recent findings that early urinary CXCL10 levels and the CXCL10:Cr ratio in kidney recipients can be prognostic biomarkers of subsequent rejection 25.

In the present study, we found that both miR‐155‐5p and CXCL10 levels – but not the CXCL10:Cr ratio – were also indicative of graft function. In patients with good recovery of kidney function (GFR >45 ml min^–1), levels of both biomarkers were below the cut‐offs established for the risk of AR. Although an inverse correlation between urinary miR‐155 and GFR has previously been observed in nontransplant patients with nephrolithiasis 38, to the best of our knowledge, this association has not previously been examined in kidney transplantation.

As is to be expected in clinical transplantation, we observed a high interindividual variability in pharmacokinetics and pharmacodynamics among our patients. Although differences between rejectors and nonrejectors were not statistically significant, rejectors showed a clear tendency towards lower concentrations. Interestingly, our data show that patients with TAC AUC >180 ng h ml^–1 and MPA AUC >50 μg h ml^–1 – thus with well established immunosuppressive drug levels – and also miR‐155‐5p expression below the cut‐off established for the risk of AR were rejection‐free. In contrast, another group of patients, with TAC AUC 100–180 ng h ml^–1, MPA AUC 40–50 μg h ml^–1, and miR‐155‐5p expression above the cut‐off was also rejection‐free. These patients also had a more preserved kidney function (GFR = 53 ml min^–1) than those who had an AR episode. Even so, such patients warrant close monitoring as they may well have a high risk of AR.

miRNAs and other biomarkers of allograft function and risk of rejection have been tested in different biological samples, including allograft biopsies 6, 20, urine 12, PBMC 33, and serum 39. Noninvasive methods have clear advantages in all cases but urine samples are especially relevant in kidney transplantation, since not only are they easy and inexpensive to obtain but they are a direct product of the kidney allograft itself.

Our study has several limitations, including a relatively small sample size, the low incidence of AR in our population (10%), and the lack of an independent validation cohort. In addition, the inclusion of non‐Caucasian patients or those with viral infections or urinary tract infections would have enhanced the robustness of our study. Additional studies are warranted to evaluate the impact of confounding factors that might decrease the specificity of our biomarkers. Moreover, the clinical implementation of our findings depends on many factors, including the time‐ and cost‐effectiveness of biomarker analysis. In routine clinical practice, the analysis of CXCL10 is perhaps more feasible than that of miR‐155‐5p in that a commercial kit (enzyme‐linked immunosorbent assay) is available, sample manipulation requires only one centrifugation step, the analysis is not very labour‐intensive and turnaround time is minimal. In contrast, miRNA analysis requires more sample manipulation and a greater number of analytical procedures, which could lead to unintentional alterations of the sample; it is also more time‐consuming and requires specific equipment and qualified technicians, which could prove an obstacle to clinical implementation.

Nevertheless, our findings suggest that urinary pellet expression of miR‐155‐5p and CXCL10 levels may well be key prognostic biomarkers of AR and graft dysfunction. Regular monitoring of miR‐155‐5p and CXCL10 levels from as early as the 1^st week post‐transplantation, in combination with the monitoring of drug concentrations, can thus be a useful tool for personalizing immunosuppressive treatment. However, there is a need for prospective data from large independent cohorts of kidney transplant patients, as well as liver and heart transplant patients, in multicentre clinical trials to refine the preliminary cut‐off values identified in the present study and to evaluate the clinical usefulness of these biomarkers. If our findings are confirmed in a larger study, miR‐155‐5p and CXCL10 levels could be candidates for inclusion in an early, noninvasive prognostic biomarker panel to prevent rejection and to monitor the clinical outcome of transplant recipients 2.

Competing Interests

The authors declare no conflicts of interest.

The present study was partially supported by TEVA Pharmaceuticals and by a grant from Fundación Mutua Madrileña (AP153362014). The authors thank the physicians and patients from the four participating centres: Hospital Clinic, Barcelona; Fundació Puigvert, Barcelona; Charité‐Universitätsmedizin, Berlin; and University Hospital, Heidelberg.

Contributors

O.M. performed the miRNA and CXCL10 analyses, interpreted the data, performed the statistical analysis, and drafted the manuscript. K.B., C.S. and L.G. conceived and designed the study, included and followed patients, and revised the manuscript for important intellectual content. I.A. performed the miRNA and CXCL10 analyses. O.R. and M.M. prepared patient urine samples for analysis. B.B. included and followed patients. M.Z. followed up on patients. I.S. included patients. M.B. conceived, designed and coordinated the study, interpreted the data, and drafted the manuscript. All authors have approved the final version of the manuscript.

Millán, O. , Budde, K. , Sommerer, C. , Aliart, I. , Rissling, O. , Bardaji, B. , Matz, M. , Zeier, M. , Silva, I. , Guirado, L. , and Brunet, M. (2017) Urinary miR‐155‐5p and CXCL10 as prognostic and predictive biomarkers of rejection, graft outcome and treatment response in kidney transplantation. Br J Clin Pharmacol, 83: 2636–2650. doi: 10.1111/bcp.13399.