SEARCH

SEARCH BY CITATION

Keywords:

  • Banff lesions;
  • microarrays;
  • pathology of renal transplantation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Emerging molecular analysis can be used as an objective and independent assessment of histopathological scoring systems. We compared the existing Banff i-score to the total inflammation (total i-) score for assessing the molecular phenotype in 129 renal allograft biopsies for cause. The total i-score showed stronger correlations with microarray-based gene sets representing major biological processes during allograft rejection. Receiver operating characteristic curves showed that total-i was superior (areas under the curves 0.85 vs. 0.73 for Banff i-score, p = 0.012) at assessing an abnormal cytotoxic T-cell burden, because it identified molecular disturbances in biopsies with advanced scarring. The total-i score was also a better predictor of graft survival than the Banff i-score and essentially all current diagnostic Banff categories. The exception was antibody-mediated rejection which is able to predict graft loss with greater specificity (96%) but at low sensitivity (38%) due to the fact that it only applies to cases with this diagnosis. The total i-score is able to achieve moderate sensitivities (60–80%) with losses in specificity (60–80%) across the whole population. Thus, the total i-score is superior to the current Banff i-score and most diagnostic Banff categories in predicting outcome and assessing the molecular phenotype of renal allografts.


Abbreviations: 
ABMR

antibody-mediated rejection

BATs

B-cell-associated transcripts

BFC

biopsy for cause

CTL

cytotoxic T lymphocytes

CATs

cytotoxic T-cell-associated transcripts

GRITs

interferon-γ and rejection-induced transcripts

IFTA

interstitial fibrosis and tubular atrophy

IGTs

immunoglobulin-associated transcripts

IRITs

injury and repair-induced transcripts

KTs

kidney transcripts

PAM

predictive analysis of microarrays

PBT

pathogenesis-based transcript set

QCATs

quantitative cytotoxic T-cell-associated transcripts

ROC

receiver operating characteristic

TCMR

T-cell-mediated rejection.

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The diagnosis of allograft rejection in biopsies is based on morphological lesions that are qualitatively identified and semiquantitatively graded by arbitrary thresholds. These lesions were specified empirically because there was no independent external standard (1–5), and their clinical relevance was then demonstrated by studies showing associations between the features and thresholds on the one hand and response to antirejection treatment, allograft function and outcome on the other (6). In renal allografts, the lesions selected for scoring were interstitial infiltration and tubulitis, lesions which are believed to reflect the presence and severity of T-cell-mediated rejection (TCMR) (5) (The scoring of arteritis is also important but is dealt with in another report.). However, tubulitis and interstitial inflammation are not specific for TCMR; they can be observed in acute tubular necrosis (7), interstitial nephritis (5) and cases with antibody-mediated rejection (ABMR) (8). To avoid overdiagnosis and overtreatment, arbitrary minimum thresholds for the diagnosis of TCMR were introduced (4,5). At least 25% of the nonscarred cortex must have interstitial infiltrates and at least moderate tubulitis must be found to call a case TCMR. Inflammation in the immediate subcapsular renal cortex, in areas with interstitial fibrosis and tubular atrophy (IFTA) and in the adventitia of large vessels were empirically considered nonspecific and excluded from the interstitial inflammation (i-) score. Similarly, tubulitis in atrophic tubules was regarded as nonspecific and excluded from the tubulitis (t-) score (4). Immunohistochemistry for phenotyping of the heterogeneous interstitial infiltrate is not part of the routine diagnostic work up of an allograft biopsy according to the Banff classification, but might have the potential to discriminate between specific/harmful and nonspecific/harmless infiltrates (9–11). Although such semiquantitative thresholds and arbitrary scoring rules by light microscopy inevitably result in limited intra- and interobserver reproducibility (12,13), these features have been useful in numerous multicenter drug trials employing biopsy-proven rejection as the endpoint (6,14,15).

Not surprisingly, however, these historical thresholds may no longer be optimal in the current era of low rejection rates. Recent data from protocol biopsies indicate that infiltrates regarded as nonspecific as well as infiltrates quantitatively below current diagnostic thresholds are present in 80% of all renal allografts and correlate with reduced long-term function (16–18). Furthermore, we were able to show that biopsies with infiltrates in areas of IFTA are associated with a worse prognosis compared to those biopsies with IFTA lacking significant inflammation within the scarred areas (19). Thus, reporting infiltrates in IFTA has potential to be of clinical relevance. At the 2007 Banff meeting, a score that acknowledged all cortical inflammation, including currently ignored types of infiltrates, especially those in IFTA, was proposed (4,20). This new total i-score requires validation and remains a provisional feature.

In this study, we used microarray-based expression data as an external reference point to compare the provisional total i-score to the current Banff i-score and t-score in terms of their prognostic value and their capability to reflect the molecular phenotype of renal allografts, that is the molecular burden of inflammation and tissue injury. Microarray data were expressed as pathogenesis-based transcript sets (PBTs), which summarize large groups of related molecules (21–27). These gene sets were derived from experimental models and represent a priori defined major biological processes in renal allografts. They include cytotoxic T-cell-associated (CATs) (21,22), interferon-γ-induced (GRITs) (24), and injury and repair-induced (IRITs) (25) transcripts, loss of epithelial transcripts (KTs) (23) and infiltration by B cells (BATs) or plasma cell-associated transcripts (IGTs) (28). CATs and GRITs correlate with interstitial inflammatory cell scores in human renal allograft biopsies (26), whereas the interstitial inflammatory T-cell burden can be assessed by a refined set of quantitative cytotoxic T-cell-associated transcripts (QCATs) (27). By the PBT approach, large-scale and cumbersome microarray gene expression results are collapsed into single PBT scores representing a measurement of the respective biological/pathological process in the tissue. Furthermore, the PBT annotation of a probeset acts as a rapid way of understanding the biological process represented by detected changes of a specific transcript. We hypothesized that microarray-based gene expression data can serve as an independent, external standard against which current histologic thresholds and features of renal allograft pathology can be assessed and refined.

Material and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Biopsies and clinical follow-up

The study was approved by the University of Alberta Health Research Ethics Board (Issue 5299). After receiving written informed consent from 104 patients undergoing a clinically indicated transplant biopsy for cause (BFC) between January 01, 2004 and October 10, 2006, 129 renal allograft biopsies obtained between 1 week and 20 years posttransplant (median 19 months) were included. In terms of immunosuppression, at the time of biopsy only five patients received no steroids and 4 no calcineurin inhibitor (65 were on cyclosporine A and 64 on tacrolimus). Seventeen patients were on azathioprine and 101 on mycophenolate mophetil, while only 11 received an mTOR inhibitor. For allograft survival, patients were followed after biopsy for a mean of 24.6 ± 8.7 months (3–41 months).

Forty-four living donor and 41 deceased donor implant biopsies, taken approximately 1 h after anastomosis, and eight native kidney samples taken from histopathologically unaffected areas of the cortex of tumor nephrectomies, served as controls.

Scoring interstitial inflammation by histopathology

Histopathological reevaluation of the 129 BFCs was done by one observer (MM) to exclude reproducibility issues as a confounding factor in this study. Paraffin embedded biopsies were stained and graded according to the updated Banff criteria, and frozen sections were stained for C4d as described previously (29). All samples fulfilled the minimal Banff criteria for adequacy (4,30). Besides the established Banff scores (e.g. i-score, t-score, etc….), the provisional total i-score, representing all interstitial inflammation, was assessed. In addition to the standard ordinal values (i.e. 0–3), both the i-score and total i-score were assessed as the continuous percentage of involved cortex. For the i-score, only nonscarred cortex was taken into account, whereas for the total i-score, the whole cortex was read. Hence, in accordance with the recent update of the Banff classification, the total i-score included infiltrates in areas of nonscarred tubulointerstitium, in areas of interstitial fibrosis and tubular atrophy (IFTA), nodular infiltrates, perivascular infiltrates and subcapsular infiltrates (20).

Microarray experiments

After sufficient tissue was obtained for histopathology, an additional 18-gauge biopsy core was collected for gene expression analysis. The tissue was placed immediately in RNALater and stored at -20°C. RNA extraction, labeling and hybridization to HG_U133_Plus_2.0 GeneChips (Affymetrix, Santa Clara, CA) were carried out according to protocols published at http://www.affymetrix.com and as previously described (26). Microarray gene expression results for each of the 129 biopsies were collapsed into PBT scores: the geometric mean of fold changes across all probe sets annotated for that PBT. Fold changes were defined as the ratio of expression values in each BFC versus the average value from the eight native kidney control samples. Probe sets in each PBT are available on our homepage (http://transplants.med.ualberta.ca/).

Statistical analysis

To analyse how Banff i-score and total i-score reflect the molecular T-cell burden and probability of rejection, receiver operating characteristic (ROC) curves were generated using the Bioconductor package ’ROCR‘ (31). P-values for comparing areas under the curves (AUCs) were determined through a permutation test. Continuous variables were correlated by Pearson correlation or Spearman correlation if not normally distributed. Not normally distributed data (i.e. Banff lesions scores) were compared using Mann–Whitney U-test. For allograft survival analysis, patients were either event-censored or censored for end of the follow-up. An event was defined as either graft loss or persistently low estimated glomerular filtration rate (eGFR by the Cockcroft–Gault formula (32)) defined as at least 3 months of eGFR <30 mL/min. In the latter case, time to event was defined as the time to the end of the 3 months of low eGFR. In patients with multiple biopsies, survival analysis was calculated using the last biopsy from each patient (n = 104) using Kaplan–Meier analysis with log-rank testing (SPSS 15.0 software; SPSS Inc., Chicago, IL).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Histopathological diagnosis

Demographics for the 104 patients providing the 129 biopsies were previously published as part of a larger study (26). From this previous study, cases with BK nephropathy (n = 6), inadequate histology (n = 6) and missing slides (n = 2) were excluded. Evaluating the remaining 129 BFCs according to the current Banff criteria (20) revealed that 20 biopsies were suspicious for rejection (i.e. borderline); 19 had TCMR; 14 antibody-mediated rejection (ABMR), that is C4d positive and had circulating anti-HLA antibodies, and met the histological criteria (2 acute and 12 chronic-active ABMR); 76 did not have sufficient histologic criteria for rejection; 9 biopsies showed transplant glomerulopathy without C4d in peritubular capillaries; 14 had acute tubular necrosis/injury, 15 had signs of calcineurin inhibitor toxicity, 17 had de novo/recurrent glomerulonephritis, 10 had IFTA NOS and 11 had other findings, for example vascular changes due to hypertension or neutrophil casts.

Correlation with pathogenesis-based transcript sets

We correlated the t-scores, the i-scores and total i-scores with PBT scores representing the molecular phenotypes of major biological processes in renal allografts (Table 1). The i-score and the t-score showed similar correlations with the molecular features: CATs, GRITs, KTs, IRITs, BATs and IGTs. However, the total i-score showed much stronger correlations with all of these molecular features/phenotypes, the strongest correlation being with the QCATs (r = 0.74, p < 0.001). Restricting the analysis to early biopsies taken ≤12 months posttransplantation revealed similar results with always stronger correlations for total i-score with the PBTs than observed for the i-score, and again the strongest correlation being with the QCATs (r = 0.77, p < 0.001). This confirmed our previous finding that the QCATs represent a quantitative molecular measurement of the cytotoxic T-cell burden in the tissue (27). Thus we elected to use the QCATs as an independent molecular standard for further comparisons between the total i-score and the current i-score.

Table 1.  Correlations between pathogenesis-based transcripts sets and Banff scores
Gene sets#i-scoret-scoreTotal i-score
  1. # Spearman correlation.

  2. *p < 0.05; **p < 0.01; ***p < 0.001.

T-cell-associated transcripts0.53***0.48***0.74***
y-lnterferon-induced transcripts0.53***0.44***0.70***
Kidney parenchyma transcripts−0.30*** −0.30*** −0.54*** 
Injury and repair-associated transcripts0.38***0.36***0.65***
Immunoglobulin transcripts0.17*  0.15   0.46***
B-cell-associated transcripts0.28** 0.28** 0.66***

Establishing a molecular threshold for pathological inflammation using the QCATs

We used living donor implant biopsies (n = 44, all within normal limits by histopathology) as a reference point to estimate a cut-off for abnormal/pathological inflammation in allografts. The threshold for giving a biopsy a ’pathological T-cell burden by transcriptome‘ call was defined as one living donor QCAT standard deviation above the maximum QCAT value seen within the living donors. This approach ensures that no BFC is assessed as having an abnormal level of infiltrating T cells unless its QCAT score is distinctly above the highest score found in the implant biopsies from living donors. This threshold (red dotted line in Figure 1) was then further assessed in deceased donor implant biopsies and nephrectomy samples, both of which occasionally show focal inflammation. In the distribution curves (Figure 1), the QCATs showed a large overlap of peak densities in the controls, that is implant biopsies (living and deceased) and nephrectomies. In BFCs, a wide range of values was found for QCAT scores: one-third of the BFCs were within the range of controls and two thirds showed a distinct separation from the control kidneys. Of interest, three control kidneys showed increased QCAT scores: two deceased donor implant biopsies (2/41) and one nephrectomy specimen (1/8). Histopathological analysis of corresponding sections of these three cases did not detect abnormal i-scores, illustrating how focal abnormalities can be missed by microscopic sections (the microarray samples a larger tissue mass than microscopic sections).

image

Figure 1. Distribution of cytotoxic T-cell-associated transcript set scores in different types of biopsies. Distribution of QCAT scores in 44 living donor implant biopsies (LD), 8 nephrectomies (Neph), 41 deceased donor implant biopsies (DD) and 129 biopsies for cause (BFC). The cytotoxic T-cell-associated transcripts (QCATs) show relatively little overlap between LD and BFC and simultaneously a wide distribution over the whole spectrum of QCAT scores in the BFCs. A threshold for calling a case pathologically inflamed according to the QCATs (red dashed line) was defined as follows: highest QCAT score in a living donor (LD) implant biopsy +1 standard deviation of the QCAT score distribution in the 44 LD biopsies.

Download figure to PowerPoint

Comparing the ability of the Banff i-score and the total i-score to reflect the molecular burden of cytotoxic T-cell infiltration

We compared how the histological inflammation scores assessed an abnormal T-cell burden as defined by high QCAT expression. We calculated ROC curves for each scoring strategy for the 129 biopsies for cause (Figure 2). ROC curves for the two infiltrate scores were compared for all cases independent of the tubulitis score, when tubulitis was scored as absent (t = 0), and when tubulitis was present (Figure 2A–D).

image

Figure 2. Comparing Banff i-score to total i-score at different degrees of tubulitis for the prediction of high CTL-associated transcript (QCAT) scores. Receiver operating characteristic (ROC) curves showing that the area under the total i-score curve (AUC = 0.85) better reflects the cytotoxic T-cell burden than does the i-score (AUC = 0.73) if tubulitis is ignored (Figure 2A, p = 0.012). If only cases without detectable tubulitis, that is with a Banff t-score = 0 are analyzed, again the total i-score shows the greater AUC (0.82) than the i-score (0.58) in terms of reflecting the QCAT burden in the tissue (Figure 2B, p = 0.001). No difference between the total i-score and the i-score is seen if the samples are restricted to those with t-scores above either the Banff borderline (t > 0, Figure 2C) or TCMR (t > 1, Figure 2D) thresholds.

Download figure to PowerPoint

Independent of the t-score, the area under the curve (AUC) was greater using the total i-score (0.85) compared to the i-score (0.73) (Figure 2A, p = 0.012). When tubulitis was present, either at t >0 or at the current TCMR threshold t>1, the total i-score and the Banff i-score were similar in predicting an abnormal T-cell burden (Figure 2C and Figure 2D). However, in cases where the t-score = 0, the total i-score outperformed the Banff i-score at predicting an abnormal T-cell burden (AUC 0.82 vs. 0.58, p = 0.001) (Figure 2B). This was also the case if the analysis was restricted to early biopsies (≤12 months posttransplantation) with a t-score = 0 (total i-score AUC 0.83 vs. 0.6 for i-score, p = 0.022).

We analyzed in detail those biopsies where the current Banff criteria (i.e. only considering nonatrophic tubules) assigned a t-score of 0 (n = 68). Biopsies with t = 0 and high QCATs compared to those t = 0 cases with low QCAT expression had similar i-scores and vasculopathy scores (cv-scores). However, biopsies with t = 0 and high QCATs had higher lesion scores for total i-score and for scarring (ci-score and ct-score), as well as g-score, cg-score, v-score and ptc-score (Figure 3). Thus in t = 0 biopsies, a molecular disturbance (high QCAT score) that is significantly associated with a number of serious pathologic lesions is better detected by the total i-score than by the i-score.

image

Figure 3. Comparing Banff to cases with high versus low expression of CTL-associated transcripts (QCATs). Biopsies with assigned Banff t-scores of 0 (i.e. only nonatrophic tubules were taken into account) were separated into those with high and low QCAT scores according to the threshold defined in Figure 1. Those biopsies with high QCAT scores show significantly (all p > 0.05, Mann–Whitney U-test) higher mean lesion scores for total i, ci, ct, v, g, cg and ptc.

Download figure to PowerPoint

Infiltrate scores and allograft survival

The above findings indicate that the current Banff i and t strategy fails in biopsies with advanced scarring. Thus, we assessed both strategies in scarred and unscarred biopsies for their ability to predict which kidneys were at risk of failure. Ignoring tubulitis as a diagnostic criterion, and applying the current infiltrate threshold for Banff TCMR (i.e. 25% inflamed cortex), Kaplan–Meier analyses showed that an elevated Banff i-score predicted a borderline significant increase in graft survival (p = 0.058, Figure 4A). In contrast, when the total i-score was used, a highly significant difference was found (94.3% vs. 66.7% allograft survival, p < 0.0001; Figure 4B).

image

Figure 4. A-D Banff i- and total i-score and allograft survival. Kaplan–Meier curves for the i-score and the i-total score. Analysing all patients and applying the current TCMR threshold for interstitial infiltrates (i.e. 25%) shows for the current i-score (Figure 4A) borderline significance and for the i-total score (Figure 4B). In patients with at least grade I IFTA (≥ ci1, ct1), that is those with more infiltrates in areas only considered by the total i-score, the i-score definition of TCMR (Figure 5C) has no prognostic significance while the total i-score definition (Figure 5D) still has.

Download figure to PowerPoint

Since the i-score by definition excludes areas of interstitial fibrosis and tubular atrophy, we conducted a subanalysis of scarred biopsies (showing at least IFTA grade I, that is >5% of cortex with IFTA). In biopsies with IFTA (n = 88, mean time posttransplantation 38 months), the current Banff i-score with a 25% threshold did not predict risk of graft loss (p = 0.599, Figure 4C) whereas the total i-score strongly predicted graft loss (p = 0.002, Figure 4D).

Scoring tubulitis in addition to the extent of interstitial infiltration did not improve the predictive value of the Banff i-score for graft loss, because all cases with >25% interstitial infiltrates have at least moderate tubulitis (t ≥ 2). The total i-score had prognostic significance regardless of the tubulitis score, because its superiority to the current i-score lies in its performance with biopsies where tubulitis cannot be scored because of advanced scarring.

Comparing infiltrate scores to diagnostic Banff categories

Figure 5 shows the relationship between interstitial infiltrates (i.e. i-score and total i-score) and diagnostic Banff categories: significantly greater mean total i-scores than i-scores can be found in cases with ABMR, borderline rejection, calcineurin inhibitor toxicity, recurrent/de novo glomerulonephritis, TCMR and transplant glomerulopathy (all p < 0.05). The remaining diagnostic categories (acute tubular necrosis, IFTA NOS and others) showed as well a trend toward greater total i-scores but failed statistical significance. Thus all disease categories in kidney allografts show interstitial inflammation with the total i-score always exceeding the i-score.

image

Figure 5. Mean extent for i-score and total i-score in diagnostic Banff categories. The relationship between interstitial infiltrates (i.e. i-score and total i-score) and diagnostic Banff categories: significantly greater total i-score than i-scores can be found in cases with antibody-mediated rejection (ABMR), borderline rejection, calcineurin inhibitor toxicity (CNIT), recurrent/de novo glomerulonephritis, T-cell-mediated rejection (TCMR) and transplant glomerulopathy (TG).

Download figure to PowerPoint

ROC curves (Figure 6) confirmed that the total i-score better predicts outcome than the i-score. In addition, with the exception of ABMR all diagnostic Banff categories show lower or equal (transplant glomerulopathy) sensitivity and specificity for predicting outcome compared to the total i-score. ABMR was able to predict graft loss with 38% sensitivity and 96% specificity, whereas the total i-score is able to achieve moderate sensitivities (60–80%) with losses in specificity (60–80%) across the whole population.

image

Figure 6. Comparing i-score and total i-score to diagnostic Banff categories in terms of predicting death-censored allograft survival. ROC curves confirming that the total i-score better predicts outcome than the i-score. In addition, with the exception of ABMR all diagnostic Banff categories show lower or equal (TG) sensitivity and specificity for predicting outcome compared to the total i-score. ABMR was able to predict graft loss with 38% sensitivity and 96% specificity, whereas the total i-score is able to achieve moderate sensitivities (60–80%) with acceptable losses in specificity (60-80%).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The emergence of genome-wide microarray expression data provides an opportunity to use objective transcriptome measurements to compare different approaches in histology for assessing disease states in biopsies (26). The current Banff consensus has defined certain rules that may miss relevant pathology, particularly if a biopsy is heavily scarred (3,20). We compared the ability of a proposed new score—the total i-score—to the existing Banff i-score. We found that the new total i-score uniformly outperformed the current Banff i-score and t-score in terms of correlations with changes in transcript sets (PBTs) representing inflammation and injury in renal allografts. ROC curves comparing the proposed total i-score and the current Banff i-score to the molecular cytotoxic T-cell burden of biopsies for cause showed greater AUCs for the total i-score, mainly because the total i-score also includes infiltrates in scarred cortical areas. The fact that these abnormalities detected by the total i-score were clinically significant was underlined by the superiority of the total i-score in predicting an increased risk of graft loss compared to the current i-score. In addition, with the exception of ABMR, the total i-score also outperformed all diagnostic categories of the current Banff classification. The results indicate that the total i-score could add valuable prognostic information, which cannot be provided by the current Banff score, especially in late biopsies with advanced IFTA and showing other diagnosis than chronic–active ABMR. Moreover, this study underscores the advantages of using objective microarray data to guide the refinement of histopathology criteria for clinically important features in biopsies.

Unlike the current Banff i-score and t-score used for assessing rejection in renal allografts, which miss cases with clinically significant inflammation, the total i-score allows the pathologist to define biopsies with abnormal inflammation even if they are scarred. Several recent studies suggest that clinically significant inflammation is not adequately represented by the current Banff criteria, that is that ‘nonspecific’ inflammation (e.g. infiltrates in IFTA) are associated with the risk of allograft failure (16–19,26,33). Recently, we were able to show that the inflammation in IFTA is the relevant feature in terms of prognosis (19). Biopsies with severely inflamed IFTA had a worse prognosis than those with IFTA lacking inflammation. The initial decision in 1991 to ignore inflammation in areas of IFTA may have been appropriate in an era when severe acute rejection episodes in the early phase after transplantation were the major diagnostic challenge. However, the total i-score emerged as robust predictor of the molecular T-cell burden, and death-censored allograft survival, mostly because it allows the pathologist to assess biopsies with advanced IFTA as well as those without IFTA (i.e. early biopsies). This does not ignore the fact that IFTA has different inflammation elements than do non-IFTA (i.e. a shift toward more B cells, plasma cells and mast cells (19). By Banff, infiltrates in areas of IFTA are considered nonspecific and hence currently ignored. But under modern immunosuppression more allografts survive longer and show advanced IFTA. Thus biopsies for cause, that is due to clinical dysfunction, from late transplants are expected to present more frequently with extensive infiltrates in IFTA. However, the question whether infiltrates in IFTA are amenable to therapy and if so, whether this would improve allograft survival has not been properly addressed yet. However, our data and those from protocol biopsies might stimulate the design of respective randomized clinical trials addressing this important question.

Tubulitis, currently a crucial diagnostic criterion for assessing renal allograft biopsies, may add little to the assessment of rejection once infiltration has been assessed, and needs to be revisited. Tubulitis has developed a special status in the minds of clinicians and immunologists that it may not deserve. Pathologists always indicated that tubulitis was not specific and could be seen in a variety of non transplant diseases (5). In our experimental models, tubulitis reflects deterioration of the epithelium in general, with loss of its ability to exclude inflammatory cells (21). The additional diagnostic information provided by the self-fulfilling criteria of tubulitis according to its current definition can be questioned. Tubulitis is only to be scored in nonatrophic tubules and hence is highly correlated with the extent of inflammation in non-IFTA areas (r = 0.85, p < 0.0001) but grading of tubulitis is more inaccurate compared to estimating the extent of the interstitial infiltrate (13). Our previous studies indicate that the t1 to t2 interface is essentially unreliable by the standard of predicting the molecular features of the biopsy (26). Thus, reproducibility of histopathology might improve if simplified histological criteria (total i-score, no grading of tubulitis) were applied (13). In addition, in our study, cases where according to the current Banff rule a t-score of zero had to be assigned showed significant greater total i-scores. It may be conceivable that these cases actually represent TCMR episodes but could not be called TCMR by histology because no nonatrophic tubules were left to make a sufficient t-score. However, this does not mean that reporting the pure presence and nature of tubulitis is completely redundant; it can indicate as a surrogate that the epithelium has experienced severe injury. Especially, severe tubulitis with rupture of the tubular basement membrane might have specificity in terms of diagnosis (TCMR vs. ABMR) and whether a nephron can recover from injury (34). Thus the total i-score opens the possibility of diminishing reliance on tubulitis, a welcome development given the troublesome features of this lesion in clinical practice.

Future diagnostic systems should aim to capture the power of transcriptomic assessment. Gene sets such as the QCATs, assessed by standardized automated platforms, provide a robust, objective and highly reproducible measurement of the effector T-cell burden in the tissue (26,35). Furthermore, summarizing high-dimensional microarray expression data into a single score (i.e. QCAT score) represents a clinically practicable approach. It gives the clinician one number rather than an uninterpretable spreadsheet with countless single-gene expression values. The QCATs (n = 25), for example were derived from T-cell cultures and include cytotoxic molecules (GNLY, GZMA, PRF1, GZMK and GZMB), signaling molecules (CD3D, CD8A, LCK, ITK and STAT4), the NK receptor NKG2D/KLRK1 as well as the effector cytokine interferon-γ (27) (for the complete gene list go to http://transplants.med.ualberta.ca/). The association of expression of most of these genes with rejection has been shown by many groups (25–42). Therefore, it is reassuring that the QCATs are the gene set showing the highest correlation with the total i-score. Furthermore, using gene set values rather than individual markers such as GZMB, may be an advantage in that random variation is reduced by averaging over the gene set (43).

Having established the prognostic value of the total i-score, we must question its diagnostic specificity. We showed that the total i-score and thus the QCAT burden represents not just TCMR but T cells entering the allograft as mediators of inflammation due to other underlying disease processes like ABMR (as was the case in numerous of our cases with advance IFTA and inflammation, that is chronic–active ABMR), drug toxicity, or recurrent glomerulonephritis. Immunohistochemistry might help to further characterize the interstitial infiltrate and provide specificity in terms of underlying disease causing the interstitial inflammation. However, as a key recent development it has been the demonstration that effector memory T cells have transcripts similar to effector T cells (44). Thus the assumption that markers such as granzyme B and granulysin can represent a cognate T-cell response is not valid. This forces us to acknowledge that the presence of T cells in a tissue (independently whether it shows IFTA or not) may simply reflect injury or damage, rather than allo-specific T-cell recognition and further challenges the specificity of the diagnostic lesions for TCMR, that is the i- and t-score. Thus the features that actually distinguish real cognate TCMR from ABMR and from nonspecific inflammation remain elusive. Nevertheless, the ultimate goal remains that an allograft should have minimal inflammation. Under current protocols more than 80% of all renal allografts show an abnormal degree of inflammation, frequently correlating with inferior outcomes (16–18). Thus, the assessment and characterization of inflammation will remain an important task, both for histopathology and transcriptomics.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

This research has been supported by funding and/or resources from Genome Canada, Genome Alberta, the University of Alberta, Alberta Health Service Edmonton Area, the University of Alberta Hospital Foundation, Roche Molecular Systems, Hoffmann-La Roche Canada Ltd., Alberta Advanced Education and Technology, the Roche Organ Transplant Research Foundation, the Kidney Foundation of Canada and Astellas Canada. Dr. Halloran also holds a Canada Research Chair in Transplant Immunology and the Muttart Chair in Clinical Immunology.

None of the authors has to declare any competing financial conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References