SEARCH

SEARCH BY CITATION

Keywords:

  • Delayed graft function;
  • donation after brain death;
  • gene expression;
  • prediction

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

We hypothesized that measurement of previously defined acute kidney injury-induced transcripts at the time of implantation would add a new dimension to existing methods based on donor factors, histology and recipient factors. We analyzed microarray results from implantation biopsies taken after reperfusion from 70 kidneys from 53 deceased donors. We used two definitions of early dysfunction: serum creatinine > 265 umol/L at day 7 posttransplant; and dialysis in the first week. The strongest correlate with early dysfunction was the mean expression of 30 injury transcripts. Older donor and recipient age were associated with early dysfunction, but histologic lesions were not. Prediction was best when the injury transcript expression was combined with donor or recipient age, particularly in standard criteria donors. In contrast, although extended criteria donor kidneys had a high risk of early dysfunction, no variables tested, including injury transcripts, predicted risk significantly, probably because these kidneys were allocated preferentially to old, high risk recipients. The injury transcripts did not predict late function, which was mainly associated with donor age. Thus, measurement of injury-induced transcripts at the time of implantation improves the prediction of early kidney dysfunction, but risk prediction may fail when old kidneys are transplanted into old recipients.


Abbreviations
AKI

acute kidney injury

DBD

donation after brain death

DGF

delayed graft function

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Kidneys from brain dead donors often manifest early dysfunction due to acute kidney injury (AKI) [1]. AKI can be incurred during brain injury and donor maintenance; organ removal, preservation and implantation; and postoperatively due to vasospasm and unstable recipient hemodynamics. Approximately 25% of kidneys from deceased donors with brain death (DBD) require dialysis in the first week after transplantation, defined as delayed graft function (DGF) [2]. DGF is associated with adverse effects, including longer hospitalization and higher costs, but usually recovers without long-term sequels. DGF is associated with reduced long-term outcomes in registry studies, but much of this is probably due to recipient comorbidities, which can contribute both to early dysfunction and late intercurrent illnesses and patient death. A common belief that DGF “causes” progressive late deterioration is not supported by phenotype data: recent biopsy studies of troubled transplants show that most late graft losses are attributable to definable entities such as antibody-mediated rejection (ABMR), nonadherence and recurrent disease [3, 4], and no phenotype of late unexplained deterioration associated with early DGF has been identified.

Both early and late dysfunctions are influenced by preexisting aging and age-related diseases in the donor. Older recipient age is also correlated with DGF [5], but this is complicated by the reluctance to transplant old, high-risk kidneys into young recipients, which ensures that the most compromised kidneys are concentrated in highest risk recipients. While measurement of biologic aging would be desirable, the molecular events associated with aging are not well-defined. Somatic cell senescence probably contributes to the poor performance of older tissues after injury [6-9], but cellsenescence is difficult to measure in a way that is more accurate than calendar age alone. Some recipient immune variables (PRA, transfusions, HLA mismatch, previous transplants) played a role in early graft dysfunction in earlier studies, probably reflecting undetected ABMR, but have become less important as cross-matching methods have improved [1, 2].

Better prediction of early kidney performance is an important unmet need. Concerns about acute injury and tissue quality cause thousands of kidneys to be discarded: 2641 of 14 409 (18%) in 2010 and 2645 of 14 787 (15%) in 2011 were discarded after being removed from the donor, and many others may have been declined before removal [10]. Improved assessment of donor organ quality would also be a useful for planning posttransplant management. Demographic predictive models such as the Kidney Donor Risk Index (KDRI) [11], Schold [12] and Irish scores [13, 14] all incorporate donor age, but their actual predictive ability is limited. Histologic assessment is commonly used to predict early dysfunction [15] but once donor age and brain death are taken into account, its predictive ability is poor, in part due to sampling error [16-20]. Histologic changes in donor biopsies can be misleading and may trigger excessive discards: in Europe, where biopsies are not used to predict kidney performance, many old kidneys are transplanted without biopsies, as in the European old-for-old program, that might be discarded in the United States [21, 22].

While the molecular assessment of tissue aging (quality) requires further refinement, recent progress in defining the molecular changes associated with AKI has opened the possibility of measuring AKI-induced molecules to add to the existing assessment of risk of early dysfunction based on demographics and histology [5, 23, 24]. We recently defined a set of AKI-induced transcripts—the injury-repair associated transcripts—that are expressed in indication biopsies of early kidney transplants with dysfunction but not in stable early transplants. These transcripts correlated with the functional disturbances, whereas histologic lesions designated as acute tubular injury did not [25]. In the present study, we hypothesized that expression of these well-defined AKI-induced transcripts as an AKI signal in DBD kidneys at the time of implantation would predict early functional impairment, adding to the predictive equations based on demographics or histology. We also examined their specific application in extended criteria donor (ECD) kidneys, and their possible relationship to long-term graft function.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Biopsy and clinical data collection

Renal allografts from DBD were biopsied intra-operatively within 1 h of revascularization upon consent from the transplant recipients (Health Research Ethics Board of the University of Alberta Issue # 5299) [5]. ECDs were defined as older than 60, or between 50–59 years of age with at least two of the following underlying risks: hypertension, cerebrovascular stroke death, or serum creatinine greater than 133 umol/L (1.5 mg/dL).

Histological assessment of biopsies

Biopsies were processed for histology (PAS, H&E and trichrome) and assessed following the updated 2009 Banff criteria [26]. Individual donor kidney histologic scores included the Remuzzi global kidney score [27, 28], the Maryland aggregate pathology index (MAPI) [29] and the percentage of sclerosed glomeruli [16].

Clinical assessment of kidney allografts

Early allograft dysfunction was defined in two ways: serum creatinine greater than 265 umol/L with or without hemodialysis on day 7 posttransplantation [30, 31]; or the need for dialysis within the first week after transplantation (the usual definition of DGF [30, 32]. The study investigators were not involved in the decision to initiate dialysis. Clinical variables for allograft dysfunction were defined based on predisposing factors to AKI and the ECD [33]. Renal function was analyzed as the reciprocal of serum creatinine (1/Cr) [34]. Clinical risk scores for evaluation of the early dysfunction of DBD kidneys included KDRI [11], the Web-based DGF risk calculator program by Irish et al. [14], and the equation developed by Schold et al. [12].

RNA extraction and microarrays

One additional 18-gauge biopsy core was collected for gene expression analysis. RNA extraction, quality control and HG_U133_Plus_2.0 GeneChip (Affymetrix Santa Clara, CA, USA) processing were described previously [35]. Detailed protocols are available in the Affymetrix Technical Manual (www.affymetrix.com). The RNA of each individual sample was run on an individual chip.

Microarray data analysis

Microarray data files (GSE37838) for 70 implantation biopsies and 8 nephrectomies were preprocessed using Robust Multiarray Analysis in Bioconductor [36]. The molecular characteristics of implant biopsies used previously identified pathogenesis-based transcript sets (PBT) that reflect biological processes of known relevance for the pathogenesis of renal inflammation and injury in transplants. Here, we applied the following PBTs to assess these processes: interferon-gamma (IFNG) effects on the tissue (GRIT) [37], kidney parenchymal transcripts (KT1s) and kidney solute carriers (KT2) [38], natural killer cell transcript burden (NKtb) [39], T cell transcript burden (TCtb) [39] and macrophage transcripts (QCMAT) [40]. The information about probe sets as well as the algorithms for PBT generation is available on our homepage (http://www.atagc.med.ualberta.ca/Research/GeneLists/Pages/default.aspx). We also studied the injury-repair response associated transcripts (IRRAT). The methodology to define IRRATs has been described [25]. The AKI signal was the summarized expression of top 30 transcripts. The summarized expression of transcripts for the PBTs sets was calculated as the geometric mean of the fold change values versus nephrectomies across all transcripts in a set.

Statistical analysis

Data analyses were performed using SPSS 19.0 statistical software package (SPSS Inc., Chicago, IL, USA), Bioconductor version 2.4 and R version 2.9.1. Differences between groups of patients were compared with chi-square tests for counts and with Student's t-test for continuous variables. Correlations between continuous variables (PBT scores, allograft function) and ordinal variables (Banff scores) were tested using Spearman's correlation. Multivariable logistic regression analysis was performed using the “RMS” package in “R” [41]. A bootstrapping procedure [41] was employed to correct for potential overfitting when combining donor or recipient age and AKI signal as a multivariable predictor and the product of multiply between variables was used as a interaction variable model. Area under the receiver operator characteristic curve (AUC ROC) was used to assess diagnostic accuracy. p-values ≤ 0.05 were considered statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Recipient and donor characteristics

We evaluated implantation biopsies taken after the opening of the vascular anastomosis in 70 kidneys from 53 DBD donors, transplanted between August 2004 and December 2010 at the University of Alberta Hospital (Table 1). Forty-two of these biopsies from 31 donors had also been included in an earlier analysis [5]. Both kidneys from 17 donors were included as “independent” observations; random exclusion of one of these pairs did not affect the conclusions. The mean follow-up time was 4 ± 1.3 years (median 4.2 years). No recipients experienced graft loss with return to dialysis in the follow-up period, rendering analysis of risk of failure impossible. There were 10 recipient deaths during this period, due to factors with no obvious relationship to kidney function: five sudden death with unknown causes, four infectious causes and one malignancy.

Table 1. Patient demographics and general characteristics of the implant biopsies
Recipient demographicsn = 70
Recipient gender (n, male/female)43/27
Recipient age (years, mean ± SD)52 ± 13
Recipient race (%) 
Caucasian77.2
Black1.4
Other (Asian, Hispanic, native Indian)15.7
Unknown5.7
Primary disease (%) 
Glomerulonephritis/vasculitis31.4
Polycystic kidney disease25.7
Diabetic nephropathy21.4
Interstitial nephritis/pyelonephritis10.0
Hypertension/large vessel disease7.1
Others4.3
Previously transplanted (%)6.0
Follow up time (years, mean ± SD)4.0 ± 1.1
Delayed graft function (%)17.1
Incident of acute rejection (%)10.0
Donor demographicsn = 53 (70 kidneys)
Donor gender (male/female)33/37
Donor age (years, mean ± SD)42 ± 17
Donor race (%) 
Caucasian80
Black0
Other (Asian, Hispanic, Native Indian)11.5
Unknown8.5
Cause of donor death (%) 
Traumatic brain injury40.0
Cerebrovascular stroke31.4
Brain anoxia11.4
Central nervous system infection7.2
Others (cavernous sinus thrombosis, brain tumor, sepsis, diabetic ketoacidosis, acute myocardial infraction)10.0
Last donor serum creatinine, umol/L (mean ± SD)66 ± 23
Cold ischemia time, hour (mean ± SD)16 ± 5.6

Among the recipients, mean age at time of transplantation was 52 ± 13 years; 39% were females; and 6% had undergone previous transplantation (Table 1). The mean donor age was 42 ± 17 years; 53% were females. The most frequent causes of donor death were brain trauma (40%) and cerebrovascular stroke (31%). The mean (final) donor serum creatinine was 66 ± 23 umol/L (0.75 ± 0.3 mg/dL). The mean cold ischemia time (CIT) was 16 ± 5.6 hours. Eleven kidneys (16%) were from ECDs.

During the first month posttransplant, seven kidneys also had indication biopsies. Six were in kidneys with DGF, showing borderline changes [3], T cell mediated rejection [2], or no major abnormalities [1]. The biopsy from one kidney without DGF showed no major abnormalities. No DGF was attributable to drug toxicity or surgical complications, and there was no hemodialysis done exclusively for hyperkalemia or volume overload.

Association between donor variables at the time of implantation and early function

Seventeen kidneys had early dysfunction by elevated serum creatinine at seven days, and 12 had DGF. We did not use GFR estimates because these are probably not valid in the dynamic early days when there is no steady state, and because all GFR estimate formulas include recipient factors that are variables of interest in the analyses.

The donor variables associated with early dysfunction were older donor age; higher risk scores that included donor age (KDRI and Schold); and the ECD phenotype, which is based on donor age [42] (Table 2). The contribution of donor age to the risk scores is reflected by their correlations with donor age: Irish r = 0.57; Schold score r = 0.78 and KDRI r = 0.75, all with p < 0.001.

Table 2. Clinical variables associated with DGF and dysfunction at day 7 posttransplant
 Mean ± SD or n High creatinine day 7Mean ± SD or n DGF
Clinical variablesGood function n = 53Poor function n = 17IGF n = 58DGF n = 12
  1. DGF = delayed graft function; KDRI = kidney donor risk index; BMI = body mass index; SD = standard deviation.

  2. 1There were no African American donors.

  3. 2ECD = Extended Criteria Donor defined by the following conditions:

  4. (1) Donor age more than or equal to 60 years or

  5. (2) Donor age 50–59 years, with at least two of the three following criteria: serum creatinine

  6. more than 1.5 mg/dL.

  7. 3Excluded pre-emptive kidney transplantation, n = 2.

  8. 4Donors who have underlying hypertension, dyslipidemia or diabetes and died from hemorrhagic or ischemic stroke.

  9. 5Correlation coefficient with donor age is 0.27, p = 0.02.

  10. 6Data split based on creatinine, so statistical inference is not valid.

  11. Parametric t-test was used for clinical variables. Significant differences are bolded. *p < 0.05, **p < 0.01, ***p < 0.001.

Donor variables1
Donor age (years)39 ± 1752 ± 13**40 ± 1753 ± 15*
Extended criteria donor (ECD)2 (n, %)5 (9.4%)6 (35.3%)*6 (10.3%)5 (41.7%)*
Irish score 2010319.7 ± 924.5 ± 12*19.2 ± 1025.8 ± 10*
Schold score0.75 ± 0.20.94 ± 0.2**0.77 ± 0.20.92 ± 0.2*
KDRI score1.20 ± 0.41.54 ± 0.7**1.22 ± 0.41.60 ± 0.8**
Female gender (n, %)26 (49.1%)11 (64.7%)31 (53.4%)6 (50%)
Donor BMI (kg/m2)25.3 ± 725.9 ± 625.5 ± 725.5 ± 6
Arteriosclerosis associated stroke death4 (n, %)17 (32.1%)5 (29.4%)18 (31%)4 (33.3%)
Last donor serum creatinine (umol/L)66.4 ± 2365.0 ± 2664.7 ± 2267.5 ± 30
Urine volume after brain death (L/day)3.8 ± 24.2 ± 33.7 ± 24.7 ± 4
Cardiac arrest (n, %)15 (28.3%)5 (29.4%)18 (31%)5 (41.7%)
Central diabetes insipidus (n, %)17 (32.1%)6 (35.3%)16/424/8
Cold ischemia time (hour)17 ± 615 ± 617 ± 615 ± 6
Recipient variables
Recipient Age (years)549.8 ± 1359.6 ± 10**49.9 ± 1363.1 ± 7***
Recipient BMI (kg/m2)27.6 ± 427.9 ± 427.8 ± 527.4 ± 3
HLA-A mismatch1.7 ± 0.51.6 ± 0.71.7 ± 0.51.5 ± 0.8
HLA-B mismatch1.8 ± 0.41.8 ± 0.41.8 ± 0.41.8 ± 0.4
HLA-DR mismatch1.9 ± 0.31.8 ± 0.411.9 ± 0.31.8 ± 0.5
 Panel reactive antibody (PRA) at transplant (%)11.9 ± 2312.2 ± 2911.0 ± 2216.6 ± 34
 Creatinine levels at day 7 posttransplantation (umol/L)6137.9 ± 43478.6 ± 346 plus 12 dialysis-dependent167.2 ± 139(all dialysis dependent)

Recipient age was higher in poorly functioning grafts (Table 2). Moreover, recipient age correlated with donor age (r = 0.273, p = 0.02), reflecting a deliberate allocation choice to avoid putting ECD kidneys into young recipients. Thus, nine of 11 ECD kidneys were allocated to old recipients (p = 0.009).

Well-functioning and poorly functioning grafts did not differ in other clinical variables, including HLA mismatches, panel reactive antibodies, CIT, terminal donor serum creatinine and death caused by stroke (Table 2). Acute rejection was not significantly increased in kidney with DGF, being diagnosed in two of 12 kidneys with DGF (16.7%) and in five of 58 with initial function (8.6%, p = 0.60).

Association between histologic lesions and early function

There was no significant difference between the grafts with poor or good function for any histologic assessment, including the Banff lesions, % glomerulosclerosis, or the histology-based Remuzzi score (Table 3). There were some numerical trends, indicating that some weak associations may have been missed by under-powering or by inadequate representation in some biopsies (e.g. arterial fibrous intimal thickening; % glomerulosclerosis). For example, the mean number of glomeruli per biopsy was 21.5, with median 11 and interquartile range 6–24.

Table 3. Histology of the donor kidneys at the time of implantation
 Histology at the time of implantation1 (Mean histopathologic score ± SD)
 High creatinine day 7Requirement of dialysis within first weekAssociation with donor age
Pathologic categoriesGood function (n = 53)Poor function (n = 17)IGF (n = 58)DGF (n = 12)Young donor (age < 55 yrs) (n = 54)Old donor (age ≥ 55 yrs) (n = 16)
  1. DGF = delayed graft function; IGF = initial graft function; SD = standard deviation.

  2. 1Three patients had inadequate tissue biopsies for evaluating ah Banff score; Four for i, ci, ct and mm Banff score; five cv Banff score; seven for glomerulosclerosis; nine for Remuzzi score and 12 for MAPI score.

  3. 2No tubulitis (t), intimal arteritis (v), early allograft glomerulitis (g), peritubular capillaritis (ptc) and allograft glomerulopathy (cg) lesions were found.

  4. Significance was tested by the non parametric Mann-Whitney test. Significant differences are bolded. *p < 0.05, **p < 0.01, ***p < 0.001.

Histology lesion scores (Banff)2
Mononuclear cell interstitial inflammation (i)0.08 ± 0.040.12 ± 0.080.09 ± 0.040.08 ± 0.080.04 ± 0.030.29 ± 0.13**
Interstitial fibrosis (ci)0.31 ± 0.070.41 ± 0.150.31 ± 0.060.42 ± 0.190.25 ± 0.060.64 ± 0.17*
Tubular atrophy (ct)0.39 ± 0.070.59 ± 0.150.39 ± 0.070.67 ± 0.190.33 ± 0.070.86 ± 0.14**
Vascular fibrous intimal thickening (cv)0.67 ± 0.081.00 ± 0.190.70 ± 0.091.00 ± 0.170.63 ± 0.081.23 ± 0.20**
Arteriolar hyaline thickening (ah)0.54 ± 0.100.82 ± 0.210.56 ± 0.090.83 ± 0.300.53 ± 0.100.93 ± 0.22
Mesangial matrix increase (mm)0.08 ± 0.040.29 ± 0.140.11 ± 0.040.25 ± 0.180.13 ± 0.060.14 ± 0.10
Histology risk evaluation
 Glomerulosclerosis (%)7.08 ± 1.524.54 ± 1.916.76 ± 1.425.04 ± 2.345.03 ± 1.1612.23 ± 3.88*
Remuzzi score1.96 ± 0.232.50 ± 0.571.96 ± 0.232.78 ± 0.661.65 ± 0.213.83 ± 0.46***
The Maryland Aggregate Pathology Index (MAPI)3.10 ± 0.392.46 ± 0.543.6 ± 0.372.60 ± 0.653.04 ± 0.362.75 ± 0.79

However, some histologic assessments were significantly associated with donor age, namely, the i-, ci-, ct- and cv-scores.

Molecular phenotype of donor kidneys at the time of implantation

We assessed the relationship between early dysfunction and the AKI signal and other PBTs in the implantation biopsies (Table 4). High AKI signal and IFNG effects (GRIT) in the implantation biopsy correlated with poorer early function at day 7 and with DGF. The degree of loss of kidney transcripts (KT1 and KT2) was also associated with DGF.

Table 4. Relationship of the molecular scores to early function in DBD kidneys
 High creatinine day sevenRequirement of dialysis within first week
 Set score (average fold change from normal kidney2)Set score (average fold change from normal kidney2)
Pathogenesis-based transcript sets1Good function (n = 53)Poor function (n = 17)IGF (n = 58)DGF (n = 12)
  1. DBD = donation after brain death donor; DGF = delayed graft function; IGF = initial graft function.

  2. 2Microarray data files of normal histology of renal cortical tissue from eight native nephrectomies performed for renal carcinoma served as controls.

  3. Parametric t-test was used for molecular variables. Significant differences are bolded. *p < 0.05, **p < 0.01, ***p < 0.001.

AKI signal (IRRAT30)1.822.63**1.862.79**
IFNG effects (GRIT)0.981.16*0.991.16*
Kidney transcripts (KT1)0.920.880.920.84*
Kidney solute carriers (KT2)0.750.670.760.60*
NK cell transcripts (NKtb)0.740.840.750.87
T cell transcripts (TCtb)0.830.910.840.90
Macrophage transcripts (QCMAT)0.931.110.941.14

In addition to the mean of the expression of the 30 AKI transcripts, 11 individual transcripts predicted DGF (Table 5), including some well known AKI biomarkers [43, 44].

Table 5. AKI genes (IRRAT30) that are significant in predicting DGF in DBD kidney biopsies after reperfusion
Gene symbolIGF (Fold change, IGF vs. normal kidneys)DGF (Fold change, DGF vs. normal kidneys)p-Value1 IGF vs. DGFAUC for DGF prediction
  1. DBD = donation after brain death donor; LD = living donor; DGF = delayed graft function; IGF = initial graft function; RTPCR = real-time polymerase chain reaction.

  2. Genes are sorted by the AUC value.

  3. 1Parametric t-test was used for clinical variables.

OSMR2.213.540.0100.74
SOD23.526.860.0100.74
NNMT2.704.590.0150.72
ITGB61.142.490.0180.72
CTSS0.651.200.0220.71
ITGB31.061.820.0210.71
CDH61.582.910.0220.71
LCN21.112.380.0220.71
SLPI1.984.260.0330.70
VCAN0.671.220.0290.70
OLFM41.853.960.0420.69

Independent predictors of early allograft function

A univariate logistic regression using donor variables showed that donor age and the molecular signal for AKI, IFNG effects, kidney transcripts and kidney solute carriers were all significant at p < 0.05. Donor gender, cause of death, CIT, donor creatinine and baseline histology were not significant by univariate analysis. When the significant variables were tested using backward stepwise logistic regression, only donor age (odds ratio [OR] = 1.06, 95% confidence interval [CI] = 1.01–1.13, p = 0.03) and the AKI signal (OR = 5.38, 95% CI = 1.57–18.38, p = 0.01) were retained as independent predictors of DGF among donor variables.

A univariate logistic regression using recipient variables showed that recipient age also predicted DGF (OR = 1.15, 95% CI = 1.05–1.26, p = 0.003).

Recipient age and the AKI signal at implantation were retained as independent predictors among all variables of DGF in the overall population (Table 6).

Table 6. Logistic regression analysis of relevant clinical, histology and molecular variables for predicting DGF in the overall DBD population
  Univariate analysis for DGF predictionMultivariate analysis for DGF prediction1
   95% CI  95% CI 
Variables ORLowerUpperp-ValueORLowerUpperp-Value
  1. DGF = delayed graft function; OR = odds ratio for DGF; 95% CI = 95% confidence interval.

  2. 1p-Value of final model prediction ≤ 0.001.

  3. Significant p-values are bolded.

ClinicalRecipient age1.151.051.260.0031.171.051.320.006
 variablesDonor age1.0611.110.03Drops out of model
 Cold ischemia time1110.32    
 Donor creatinine1.010.981.030.71    
 (before procurement)        
 Cause of death1.110.34.170.88    
 Donor gender0.870.253.020.83    
HistologicvariablesMononuclear cell interstitial inflammation0.890.098.410.92    
 (i Banff)        
 Interstitial fibrosis1.460.454.780.53    
 (ci Banff)        
 Tubular atrophy2.640.818.610.11    
 (ct Banff)        
 Vascular fibrous intimal thickening2.060.775.50.15    
 (cv Banff)        
 Arteriolar hyaline thickening1.590.713.540.26    
 (ah Banff)        
 Mesangial matrix increase2.130.548.380.28    
 (mm Banff)        
 Glomerulosclerosis score0.980.91.060.63    
MolecularInterferon gamma effect6.261.0138.860.05Drops out of model
variablesKidney transcripts0.0100.510.02Drops out of model
 Kidney solute carriers0.180.040.780.02Drops out of model
 AKI signal4.921.5815.350.017.441.5535.860.012

Predictive models for DGF

To build a combination/interaction predictive model for DGF, only significant variables in a univariate logistic regression shown in Table 6 were used. We did not include the molecular variables, GRIT, KT1 and KT2 because they are highly correlated with the AKI signal. (Analyses were performed on DGF rather than on creatinine at day 7 because the equations were only derived for DGF.) We also evaluated the predictive ability in the subgroups: standard criteria donor (SCD) and ECD; Table 7). We compared the predictive capability of donor age, recipient age, AKI and the combination of AKI with both, to published equations for predicting DGF (KDRI, Schold and Irish scores), using ROC curve analysis.

Table 7. Predictors of delayed graft function
 DGF prediction for general DBD (IGF = 58, DGF = 12)DGF prediction for standard criteria donor (IGF = 52, DGF = 7)DGF prediction for extended criteria donor (IGF = 6, DGF = 5)
Predictor variablesBest cut pointAUCPPVNPVBest cut pointAUCPPV1NPV1Best cut pointAUC
  1. ROC = receiver operating characteristic; AUC = area under the ROC curve; DBD = donation after brain death donor; DGF = delayed graft function; IGF = initial graft function; PPV = positive predictive value; NPV = negative predictive value; KDRI = kidney donor risk index; AKI = acute kidney injury.

  2. 1PPV and NPV shown only for comparisons where the AUC is >0.5 and the p value (AUC vs. 0.5) is < 0.05.

  3. 2The PPV and NPV of traditional indices correspond to the best discriminative cut point of present study.

  4. 3Recipient variables in Irish score are PRA, duration of dialysis and recipient BMI (no recipient age) (14).

  5. 4Low score of Irish (≤ 29) showed AUC for IGF prediction; AUC = 0.82, p = 0.02, PPV = 80, NPV = 83.3.

  6. 5Bias-adjusted to correct for overfitting a multivariable model.

  7. Significant values are bolded. *p < 0.05, **p < 0.01, ***p < 0.001.

The predictive ability of donor variables of early allograft function
 Donor age48.00.67*30.394.648.00.60  66.00.47
 KDRI21.500.71*41.290.61.350.67  1.590.47
 Schold20.900.69*32.192.90.850.67  1.220.35
 Irish1,220.00.68*29.091.927.00.73*36.493.729.00.184
 AKI signal (log2 score)0.930.76**32.394.91.140.84**35.397.60.790.63
 Donor age, AKI signal0.210.78A***47.494.10.100.815***26.91000.420.425
 Donor age × AKI signal54.60.81***45.094.059.70.83**50.095.954.30.63
The predictive ability of recipient variables of early allograft function
 Recipient age59.00.83***42.993.956.00.83**30.097.467.00.67
 Recipient age, AKI signal0.190.87A***52.497.90.080.93A***46.71000.470.47A
 Recipient age × AKI signal63.70.84***50.096.077.20.91***60.09852.00.63

In overall population and particularly in SCD group, the AKI signal, recipient age and the combinations of AKI signal with donor age or recipient age were the best predictors of DGF (Table 7). The combined AKI signal and the recipient age was the best model for predicting DGF in general DBD and in SCD population (Figure 1).

image

Figure 1. Prediction of delayed graft function by recipient age, AKI signal and the combination of recipient age and AKI signal in the overall (A) and standard criteria donor (SCD) population (B). Areas under the receiver operating curves (AUC) for predicting DGF are shown in the inset. Combination of the recipient age with the AKI signal has the largest AUC.

Download figure to PowerPoint

Surprisingly, no variables or models showing significant predictive value in ECD population (Table 7).

Since 70 kidneys from 53 donors were included in the full analysis, we tested whether the inclusion of paired kidneys biased the results. When we repeated the analysis using only one kidney per donor, the results were similar, indicating that inclusion of paired kidneys did not bias the results (Supplementary Table 1).

Interaction between donor and recipient age and early dysfunction

We studied relationships the determinants of allograft function with SCD and ECD kidneys, taking into account the striking association of ECD kidneys with older recipients, both for DGF (black symbols in Figure 2A) and for elevated creatinine day 7 (black symbols in Figure 2B). ECD are shown as triangles. This demonstrates the strong relationships among donor age, recipient age and impaired initial function.

image

Figure 2. The relationship between the donor and recipient age and graft dysfunction. Graft dysfunction was defined by DGF (A) or high creatinine day 7 (>265 μmol/L) (B). Transplants from old donors into old recipients are at a high risk for early dysfunction. However, interpretation of early dysfunction is confounded by a relationship between the donor and recipient age.

Legend; ○ = SCD with early graft function, ● = SCD with graft dysfunction, ∆ = ECD with early graft function, ▲ = ECD with graft dysfunction. SCD – standard criteria donors, ECD – extended criteria donors

Download figure to PowerPoint

Rejection was diagnosed more frequently in ECD kidneys (2/11) compared to SCD kidenys (5/59) but this was not significant.

Implantation risk factors and late allograft function

We examined the relationship between risk factors and late allograft function at 1 and 3 years posttransplant (Table 8). The reciprocal of creatinine (1/cr) was used for representing allograft function instead of conventional formulas (i.e. Cockcroft–Gault and MDRD equations) to avoid the artifact of having recipient age in the GFR equations when recipient age was the variable being studied. Risk factors were grouped into four categories; AKI signal, clinical variables (donor age, recipient age) histology (Banff score ci, ct, Remuzzi and glomerulosclerosis score) and predictive risk variables (KDRI, Irish and Schold).

Table 8. Correlation between the pre-implantation risk factors and late allograft function (1/cr)
 Correlation2 with late allograft function (1/cr) All (n = 70)
Variable related to implantation biopsy1One yearThree years
  1. 1No correlation between long-term allograft function and CIT, i cg, ah, mm, cv-Banff, MAPI score, GRIT, KT1 and KT2 was found.

  2. 2Spearman correlation co-efficient tested relationship between risk variables and allograft function.

  3. Significant correlation co-efficient are bolded *p < 0.05, **p < 0.01, ***p < 0.001.

Implantation molecular injury:  
AKI signal (IRRAT)0.080.05
Clinical variables  
Donor age−0.120.25*
Recipient age−0.14−0.07
Histology at implantation  
Interstitial fibrosis (ci)−0.15−0.18
Tubular atrophy (ct)−0.110.32**
Glomerulosclerosis−0.12−0.18
Remuzzi score−0.160.22*
Predictive models  
KDRI−0.28*−0.28**
Irish−0.24*−0.34**
Schold−0.10−0.19

The AKI signal at the time of implantation did not predict late renal function. Donor age and donor age-dependent models such as KDRI and Irish correlated weakly with late kidney function, as did some histology features that are associated with donor age. Recipient age was not associated with late function. Thus AKI, recipient age and donor age all contribute to early dysfunction, but donor age and related histology features are the main determinants of late function.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

The present study examined whether the previously defined molecular AKI signal, when measured in biopsies from kidneys of brain dead donors at the time of implantation, would add insight into early kidney dysfunction beyond the usual predictors based on histology, donor age and recipient age. We confirmed that donor age (or indices based on donor age) and recipient age correlated with early dysfunction, but age-related histologic lesions had did not have any independent ability to predict early function once age was considered. The molecular AKI signal predicted early dysfunction independently from donor and recipient age, across all donors and within the SCD group. However, for ECDs, prediction of early dysfunction is confounded by the interaction between ECD and old recipient age. This reflects the practice of avoid transplantation of ECD kidneys into young recipients, which creates a powerful association of ECD and old high-risk recipients. Long-term function is poorer with kidneys from old donors, and thus age-related histology lesions have weak association with late GFR measurements, but recipient age was not significant. The results show the potential value of injury molecules for predicting early outcomes, but remind us of the complexity of predicting outcomes for ECD kidneys in old recipients.

The AKI signal reflects the kidney injury-repair response that is triggered by parenchymal injury, including transcripts from epithelial cells, stromal cells and inflammatory cells related to wound repair [45]. The 30 AKI signal transcripts, originally identified in indication biopsies from transplanted kidneys in the weeks after transplant [25], include some biomarkers for AKI, for example, LCN2 and OSMR [43, 46]. The AKI signal in implantation biopsies probably reflects injury induced in the donor during the period from brain injury to kidney removal and cooling, not injury induced during preservation or after opening the vascular anastomoses. At least 10 of the AKI transcripts have previously been shown to be induced in biopsies taken before implantation [24, 44, 47, 48], where they were either increased in DBD versus living donor kidneys or increased in kidneys which subsequently displayed DGF versus IGF (Table 9). Transcription of the AKI mRNAs is energy, temperature and time dependent. With exception of a small number of “immediate early genes” such as JUN and FOS that appear within minutes, most “effector” transcripts such as those involved in cellular responses in vitro are late, in some cases requiring a wave of protein synthesis [49, 50]. The time at 37°C required for injury from preservation and implantation to induce AKI transcripts as the kidney rewarms is uncertain. Thus, the AKI signal in DBD kidneys at the time of implantation probably accurately reflects the extent of parenchymal injury during the hours after brain injury and in donor maintenance, but the extent to which it reflects very recent injury when there has been little time at 37°C (e.g. donation after cardiac death, cold preservation) remains to be determined.

Table 9. AKI genes (IRRAT30) identified in published studies of DBD kidneys before reperfusion
Gene symbolGene nameComparisonPlatform typeReferences
  1. DBD = donation after brain death donor; LD = living donor; DGF = delayed graft function; IGF = initial graft function.

  2. 1Transcripts significantly higher in DBD than LD biopsies at the time of transplantation.

  3. 2Transcripts in DBD biopsies at the time of transplantation that were significantly higher in DGF than IGF.

ADAMTS1Disintegrin-like and metalloprotease with thrombospondin t1 mot1DBD vs. LD1microarray(24)
CDH6Cadherin 6, type 2, K-cadherin (fetal kidney)DBD vs. LD1microarray(24)
LCN2Lipocalin 2 (Neutrophil gelatinase-associated lipocalin; NGAL)DGF vs. IGF2RTPCR(44)
LTFLactotransferrinDBD vs. LD1microarray(24,47)
NNMTNicotinamide N-methyltransferaseDBD vs. LD1and also in DGF vs. IGF2microarray(24,48)
PTPRCProtein tyrosine phosphatase, receptor type, CDGF vs. IGF2microarray(48)
RARRES1Retinoic acid receptor responder (tazaroteneDBD vs. LD1 and also in  
  induced) 1DGF vs. IGF2microarray(24,47,48)
SERPINA3Serine proteinase inhibitor, clade A, member 3DBD vs. LD1microarray(24,48)
SLPISecretory leukocyte protease inhibitor (antileukoproteinase)DBD vs. LD1microarray(24)
VCANChondroitin sulfate proteoglycan 2 (versican)DGF vs. IGF2microarray(48)

The weak association of histologic lesions with early outcomes is probably due to their association with donor age, which is the real predictor. We found independent effects of older donor or recipient age and the AKI signal, but not for histologic lesions. The histologic lesions that characterize old donor kidneys—atrophy, scarring, glomerulosclerosis, arterial fibrous intimal thickening, arteriolar hyalinosis—were associated with donor age as expected but their associations with early dysfunction were not as strong as donor age itself. The literature on histologic assessment has never supported the belief that lesions such as % glomerulosclerosis were strongly predictive of early function [18, 19, 51]. Moreover, studies of the correlation of age-related histologic lesions with late outcomes are complicated by the allocation paradox: high-risk old kidneys are given to high risk old recipients. The finding in registries that old kidneys have high late failure rates is probably driven at least to some extent by recipient comorbidities. The old kidneys in this study manifested increased early dysfunction, followed by reduced but stable GFR. However, kidney loss was rare except for recipient death, reflecting the strong association of ECD kidneys with old recipients, who have high comorbidities. Suboptimal GFR is usually better than dialysis, and the kidneys from older donors in our study continue to provide clinically stable and acceptable function after recovery from DGF.

Thus, recipient age emerges as an important factor because co-morbidities increase not only the risk of patient death but of graft failure during intercurrentillnesses. Multiple regression models alone cannot deal with systematic link between donor age/tissue quality issues and recipient age/medical risk issues. Thus, it is important to define these survival issues in terms of actual phenotypes of failure, by studies attributing causality to observed kidney failures. In our attribution analysis [52], we found some kidneys that fail in the context of intercurrent medical illnesses, possibly explaining the impact of recipient age (and age-related comorbidities) on death censored late kidney loss as well as recipient death [52].

Although the AKI signal predicts early dysfunction, a molecular risk assessment tool at the time of organ removal should incorporate estimates of biologic aging and somatic cell senescence if it is to guide decisions at the time of organ allocation and avoid discarding potentially useful DBD kidneys because of uncertainty. Knowledge of AKI and the extent of biological aging could also direct postoperative management, for example, monitoring in the ICU for a period of time to prevent hemodynamic instability in recipients with impaired left ventricular function. Thus, a risk assessment equation should incorporate the AKI signal, a measure of the extent of biological aging and a weighted version of donor age, reflecting the u-shaped effect of donor age that recognizes that the young adult is the best donor source, as in equations such as Irish, Schold and KDRI. The equation should also predict performance in old recipients.

Although molecular testing of biopsies using RTPCR or other technologies with rapid processing time (e.g. under 3 h) is feasible, such testing must demonstrate independent predictive power in high-risk ECD kidneys, because there the costs would be more than offset by increasing the number of transplants and potentially guiding their management. This will require larger studies with more power that specifically target ECD.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

This research has been supported by funding and/or resources from Novartis Pharma AG, and in the past by Genome Canada, the University Of Alberta Hospital Foundation, Roche Molecular Systems, Hoffmann-La Roche Canada Ltd., the Alberta Ministry of Advanced Education and Technology, the Roche Organ Transplant Research Foundation and Astellas. Dr. Halloran held a Canada Research Chair in Transplant Immunology until 2008 and currently holds the Muttart Chair in Clinical Immunology.

Disclosure

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

P. F. H. holds shares in Transcriptome Sciences Inc., a company with an interest in molecular diagnostics. Beyond that, no author has any conflict of interest to disclose as described by the American Journal of Transplantation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Disclaimer: Supplementary materials have been peer-reviewed but not copyedited.

FilenameFormatSizeDescription
ajt12043-sup-0001-TableS1.xlsx13KTable S1: Area under the ROC curve for discriminating day seven graft function (comparing pairs and single pair analysis)

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.