Investigation of novel biomarkers of drug-induced kidney injury in renal transplant recipients undergoing graft biopsy

Urinary and blood kidney biomarkers (BM) remain insufficient to detect early kidney injury. Our aim was to compare new kidney BM with histopathological data in kidney allograft recipients. Blood and urine samples of consecutive adult patients were collected just prior graft biopsy. All kidney samples were classified according to Banff 2007. The diagnostic performance (area under ROC curve, AUROC) of 16 new BM was compared to those of urinary proteins, blood urea nitrogen, eGFR and serum creatinine to identify histopathological groups. 223 patients were analyzed. Versus slightly modified renal parenchyma (SMRP), microalbuminuria and urinary proteins had the highest diagnostic performance toward glomerular injury. Urinary neutrophil gelatinase associated lipocalin (NGAL) had the best performance values for acute tubular necrosis (ATN) versus SMRP (AUROC 0.93). Others BM reached slightly lower AUROC reaching 0.89. When comparing ATN to acute rejection several new urinary BM (NGAL, cystatin C, and MCP1) and classical BM (eGFR, serum creatinine) reached similar AUROC values ranging from 0.80 to 0.85. Values of urinary NGAL were 10-fold higher in ATN compared to acute rejection (p = 0.0004). New BM did not outperform classical BM during renal transplantation. Urinary NGAL might be helpful to discriminate ATN and acute rejection.

biopsy as a routine evaluation at 3 month or because of an impaired renal function, and for whom blood and urine collection was possible, were considered for eligibility. Patients could not be included if they received nephrotoxic drugs other than those administered as standard care for renal transplant, within 1 month prior to screening until the last sample collection. Drugs known to alter tubular secretion of creatinine (e.g., cimetidine, amiodarone, dronedarone) but not cotrimoxazole, taken within 7 days prior to screening until the last sample collection, were additional exclusion criteria.
Inclusion started on July the 6 th 2011 and ended on March the 11 th 2015.

Methodology
This prospective, single-center, open label study was conducted in the nephrology department of Bicêtre University Hospital, APHP. It was part of the European Innovative

Medicines Initiative Joint Undertaking (IMI JU) Safer and Faster Evidence based Translation
(SAFE-T) Consortium aimed at the clinical biomarker qualification for drug-induced injury to kidney, liver and the vascular system. All the drugs and their doses were left to the treating nephrologists, using "real world" treatment options.
All the patients gave their written informed consent to participate. The duration of participation per patient (from screening to end-of-study) was 1 or 2 days depending on the time interval between blood and urine collection (Day 1), and the graft biopsy (Day 1 or Day 2).
Blood and urine samples were collected just prior graft biopsy.
Fasting blood samples were collected for plasma (on Li-Heparin and EDTA) and serum preparation (in a dry tube). The blood samples were centrifuged at +4°C at 2600 g for 10 min, within 30 min of collection for plasma samples, and after 30-60 min of clotting at +37ºC further 30 min at +4ºC for serum sample. Immediately following centrifugation, 3 aliquots (0.5 mL each) of the supernatant were transferred into 3 labeled polypropylene cryovials and immediately frozen at -80°C.
Fasting urine sample (spot urine) was collected in a dry container, after that he/she had emptied his/her bladder (overnight urine discarded) and immediately placed in an ice bath, then centrifuged at +4°C at 3000 g for 5 min, within 30 min of collection. Aliquots (1 mL each) of the supernatant were transferred into 12 labeled polypropylene cryovials, and they were immediately frozen at -80°C.
The same pathologist also classified patients in 11 groups based on his histological conclusion.

Statistical methods (see Additional file 2)
Descriptive statistics were presented as mean ± SD, median ± interquartile range.
All serum or blood renal BM were analyzed as raw data. All urinary renal biomarkers were analyzed as raw data corrected by urinary creatinine. For individual and combination of BM (when AUROC>0.7), bootstrap best BM's diagnostic threshold and associated 95%CIs were estimated by looking for the pair of estimated sensitivity/specificity closest to perfect sensitivity and specificity. Sensitivity and specificity bootstrap estimates and 95% CIs associated with this best BM's diagnostic threshold were also calculated.
Estimates were calculated using bootstrap methods in order to lighten the overfitting issue related to the search of the best biomarker among several ones.

Patients
Two hundred fifty-seven patients were included in this study from the nephrology department of Bicêtre University Hospital. Twenty-three included patients did not undergo a graft biopsy because of urinary infection or hemostatic abnormalities and were not retained in the present analysis, as well as 11 additional patients for whom no analyzable kidney tissue could be obtained. Thus, 223 patients were retained for the present BM study. Patient characteristics at baseline for BM population are presented in Table 1.
Patients were classified in 11 groups based on renal histology. Two patients could not be classified due to inadequate renal parenchyma ( Table 2).
Only groups containing strictly more than 5 patients were considered for further statistical analysis. Thus 5 groups were not included in this analysis due to their small numbers of patients, except when comparing to "all other patients".
Graft biopsies were also classified according to BANFF 2007 for the 5 main lesion types, for which severity was graded from 0 to 3 (Table 3).

Descriptive statistics for biomarkers
The range levels of BM normalized to urinary creatinine for all patients are presented in

BM performance for main histopathological diagnosis
Each main histopathological diagnostic group containing more than 5 patients was compared toSMRP as control group. improves, a recurring question arises concerning the possibility of an underlying graft rejection (ACR or AHR) or the normal recovery from an ATN. In order to assess the potential utility of these new BM to discriminate the 2 diagnosis (that might avoid a graft biopsy) we tested their diagnostic performance by comparing ATN to ACR, AHR or ACR+AHR. The only patient with mixed humoral and cellular rejection (AMR) was included in ACR and AHR groups.
When comparing ATN to ACR, the new BM did not improve the diagnostic performance of classical BM (Table 6). However, an association of BM including NGAL might increase the diagnostic performance with 90% sensitivity and 95% specificity (Additional file 5). These analyses of combinations of biomarkers are for exploratory purpose only.
When comparing ATN to AHR, the new BM did not improve the diagnostic performance of classical BM (

Discussion
This study aimed to assess diagnostic accuracy of selected urinary and blood renal BM reflecting lesions in different segments of the nephron compared to renal histopathological data. Our results revealed that when taken individually the new BM did not outperform the classical BM currently used in clinical practice. However, urinary NGAL seems to be potentially helpful to differentiate graft rejection from ATN recovery during early renal transplantation.

-BM association with each BANFF anatomical structure lesion:
When comparing the BM by histological lesion types and severity, total urinary proteins and microalbuminuria were the most performant BM to detect and quantify glomerular injury, as expected (25)