SEARCH

SEARCH BY CITATION

Keywords:

  • Cytotoxic T cells;
  • kidney transplantation;
  • transcriptome

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Having defined CTL-associated transcripts (CATs) in CTL in vitro, we used microarrays to quantify the burden of CAT sets compared with individual transcripts in human renal transplant biopsies with T-cell mediated rejection (TCMR). CAT sets in TCMR resembled diluted CTL RNA, maintaining overall hierarchy of expression relative to CTL in vitro. NK selective sets were not detected in TCMR, indicating the CATs mainly reflect T cells. We selected 25 highly expressed CATs that diluted quantitatively in kidney RNA (QCATs) and remained detectable after 32-fold dilution. QCAT burden in 14 kidneys with TCMR was 3 to 15% of CTL RNA, correlating with infiltration. One biopsy diagnosed as TCMR only by endothelialitis had little interstitial infiltrate and lowest CAT burden. CAT sets were more consistent than individual CATs such as perforin or granzyme B, which showed heterogeneity. In luster and principal component analysis, QCATs grouped biopsies with TCMR together, in close relationship to in vitro CTL. Thus QCAT sets robustly measure the burden of CTL and effector memory T cells in biopsies as %CTL RNA, in a manner not achieved by measurement of individual transcripts.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

T-cell-mediated rejection (TCMR) of organ allografts is characterized by interstitial infiltration by T cells (1–3), which display effector or effector memory (EM) phenotypes. CD8+ T cells are more common than CD4+, while naïve T cells play little role since they cannot enter inflamed sites (4). Antigen-specific T cells are probably a minor component, compared to EM T cells recruited by the inflammatory process (5). NK cells are infrequent in the infiltrate of TCMR (6,7). Diagnosis of TCMR relies on histologic assessment of the infiltrate, as well as the presence of tubulitis and/or endothelialitis. Despite an international consensus process (8), semi-quantitative histologic grading of TCMR is subjective and limited by poor inter-observer agreement (9,10). Immunohistochemical phenotyping of the interstitial infiltrate has been proposed to improve diagnostic accuracy (3), but this method is also observer dependent and severely limited by technical issues and inter-laboratory variation (11).

Analysis of mRNA transcript expression can provide an objective measurement of events in the graft. Transcript expression is highly altered in kidneys with TCMR (12–14), in particular by the appearance of cytotoxic T lymphocyte (CTL)-associated transcripts (CATs) such as perforin, granzyme B and Fas ligand (15–17). In a mouse kidney allograft model of TCMR that mimics the human histologic lesions, we recently showed that CAT expression increased with interstitial infiltration in the allograft, reaching a maximum of about 15% of the signal in a pure CTL culture. Thus CATs could be used for estimating the effector T-cell burden (14,16). Moreover, human orthologs of mouse-derived CATs correlated strongly with the histologic diagnosis of rejection (14,16), indicating that CAT expression has diagnostic value. As the relationships of transcript expression to transplantation biology are understood, diagnosis based on transcript expression and histology is feasible. Given the increased accuracy for a quantitative result based on transcript expression in biopsy samples we propose that transcript expression supplement histological assessment of biopsies to increase accuracy.

In a recent analysis of human allostimulated CTL (see accompanying manuscript), we observed that CATs were shared extensively between CD4+ and CD8+ CTL, as well as NK cells and EM T cells. We now studied how CATs behaved in renal transplant biopsies. The goals of the present study were to estimate the burden of CATs in rejecting biopsies; to assess whether they preserved the general hierarchy of transcript expression relative to CTL in vitro; and to define whether individual transcripts predict the CTL burden in renal allografts. Our hypothesis was that the burden of CTL and EM T cells in a tissue (4) can be estimated by expression of CATs in that tissue and that CAT sets would more consistently estimate the T-cell burden than individual transcripts.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Isolation and generation of cell populations

Purified cell populations were isolated from whole blood of healthy volunteers and treated as described previously (see accompanying manuscript).

Human kidney transplant biopsies

We included human renal allograft biopsies from the database of the Alberta Transplant Applied Genomics centre, which contains histologic, clinical and mRNA microarray data on 143 renal allograft biopsies performed for clinical indication as standard of care between January 2004 and October 2006. Details of this study are published elsewhere (14). We selected biopsies histologically diagnosed as TCMR using the current Banff criteria (8) (n = 17) and biopsies with histological features of ATN (18) (12 biopsies from 10 patients). Kidney tissue from macroscopically and histologically unaffected areas of the cortex of native nephrectomies performed for renal carcinoma served as controls (n = 8). Sixteen TCMR cases were confirmed retrospectively as a clinical rejection ‘episode’, defined by clinically apparent functional changes and/or response to therapy in the absence of alternative explanations (e.g. obstruction, calcineurin inhibitor toxicity). Three patients received anti-rejection treatment before the biopsy was performed. One biopsy was diagnosed as TCMR based on the presence of endothelialitis (v1 lesion) but lacked significant tubulointerstitial changes (t1, i1), with persistent dysfunction with no clear episode. This uncommon ‘isolated v’ lesion is classified as rejection by convention, but whether it is actually TCMR is not known (8). Analysis of the biopsy material was approved by the University of Alberta Health Research Ethics Board (Issue # 5299), and all included patients consented to this study.

Microarrays and RNA preparation

In addition to the tissue cores obtained for conventional diagnosis, one 18-gauge biopsy core was collected for gene expression analysis. The tissue was placed immediately in RNA. Later, kept at 4°C for 4–24 h, then stored at −20°C. RNA extraction, labeling and hybridization to HG_U133_Plus_2.0 GeneChip (Affymetrix, Santa Clara, CA) were carried out according to the protocols published at http://www.affymetrix.com.

Microarray data preprocessing and selection of transcript sets

Data files were preprocessed using robust multi-chip averaging and subjected to variance-based filtering (19), as described previously (20). Preprocessed data was imported into GeneSpring(tm) GX 7.3 (Agilent, Palo Alto, CA) for further analyses.

As previously done for mouse CATs (16), we defined human CATs highly expressed in human CTL (see accompanying manuscript). We selected transcripts preferentially expressed in human CD8+ CTL compared to nephrectomy, B cells and monocytes [significant at a false discovery rate of 0.01 in cells (21,20)] (Supplementary Figure S1A), and removed transcripts inducible more than 2-fold (p < 0.05) in monocytes treated with IFN-γ. In addition, transcripts with signal intensity >200 in nephrectomy, B cells and untreated or IFN-γ-treated monocytes were deleted. Note that this algorithm identifies CATs but not CTL-specific transcripts.

The NK-selective transcript set was derived from the overlap of probe sets showing increased expression (p < 0.01) in NK cells compared to purified B cells, monocytes and nephrectomy samples. All transcripts with signal intensity >200 in nephrectomy samples, B cells, monocytes, IFN-γ treated monocytes, CD4+ or CD8+ CTL were removed (Supplementary Figure S1B).

Gene expression was analyzed as the raw signal values for each probe set. Averaged fold changes were calculated as the geometric means, unless stated otherwise. All transcript abbreviations use Entrez Gene nomenclature: http://www.ncbi.nlm.nih.gov/entrez.

Dilution experiment

To simulate dilution of CTL RNA by kidney RNA in rejecting kidneys, we diluted labeled cRNA from CD8+ CTL with kidney cRNA in increasing ratios (1:1; 1:4; 1:8; 1:16 and 1:32) to create a standard curve of CAT expression.

Immunostaining

We evaluated the number of CD8+ cells in tissue sections by staining for CD8 in those biopsies with a diagnosis of TCMR for which tissue was available (n = 12). Antigen retrieval of formalin-fixed and paraffin-embedded tissue sections was performed with an EDTA-based buffer on a Ventana Benchmark® XT. Tissue sections were stained with mouse monoclonal anti-human CD8 primary antibody (DAKO clone C8/144B; 1:50, 30 min at room temperature), detected by Alexa Fluor 546 labeled goat anti mouse secondary antibody (Invitrogen, Carlsbad, CA; 1:10, 1 h at room temperature). Staining was evaluated by counting the absolute number of positive cells in three high power fields with the highest density of infiltrating cells.

Real-time RT-PCR

Expression of selected transcripts (CD4, CD8A, CD8B1, EOMES, FASL, PRF1 and GZMB) was confirmed by real-time RT-PCR using TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA). Detailed methods and ABI gene expression IDs are available as supplementary material.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

CAT expression in TCMR

We previously showed that CAT expression is increased during TCMR in human and mouse kidney transplants (14,16). In mouse transplants, the signals for the highly expressed CATs plateaued at 15% of the signal in CTL in vitro, reflecting dilution by mRNA from kidney cells and other cell types (16). Because of this dilution, only CATs with relatively high expression in CTL were detectable above the background of the microarray in rejecting mouse kidney allografts. We explored how CAT expression could be used to quantify the CTL burden in human biopsies, incorporating our recent analyses of CAT expression in allospecific CTL (see accompanying manuscript).

We organized CATs derived from CD8+CTL (n = 205) on the basis of the geometric mean of their expression levels in allospecific human CD8+ CTL (Supplementary Table S1, Supplementary Figure S1A) and arbitrarily divided them into subsets of 10 transcripts each, arranged by decreasing signal intensity in CD8+ CTL. We selected the top four and the bottom four subsets for analysis in human kidney transplant biopsies. The top four subsets were increased in TCMR, and each set gave the same estimate of the CAT burden in each biopsy as measured by the fold increase in geometric mean signal of each sample over the geometric mean of the nephrectomy samples (Figure 1A). In contrast, the four CAT subsets with lowest expression in CD8+ CTL were not consistently elevated in TCMR biopsies, indicating an inability of these transcripts to be detected above the background of the microarray system. Despite variability between TCMR cases, the order in which each subset was expressed in TCMR biopsies resembled that of their expression in CD8+ CTL in vitro. Thus CAT expression in infiltrating cells in a biopsy preserves its relationship to gene expression in CTL in vitro and behaves like a dilution of in vitro-stimulated CD8+ CTL. The CAT burden—and by implication the CTL burden—in a biopsy can thus be measured by any of several independent CAT sets as a fraction of the expression in CTL (%CTL mRNA), provided they have high enough expression to be detected above the microarray background.

image

Figure 1. Expression of CATs and NK-selective transcript in kidney TCMR compared to CD8+ CTL and NK cells. (A) Linegraph depicts CAT subsets. A total of eight subsets, four depicting the top 40 CATs and four the bottom 40 CATs were arranged by highest expression in CD8+ CTL. Expression of these subsets was assessed in nephrectomy, TCMR, CD8+ CTL and NK cell samples. Values displayed are the fold-increase in signal normalized to the mean nephrectomy score. (B) Linegraph depicts NK-selective transcript subsets. A total of four subsets of 10 transcripts each were generated, depicting the top 40 NK-selective transcript arranged by highest expression in NK cells. NK-selective transcript subset geometric mean values for individual nephrectomy, TCMR and NK cell samples are shown. Values displayed are normalized to the mean nephrectomy score. Lines connecting data points are placed solely to allow visual separation of the CAT subsets.

Download figure to PowerPoint

NK-selective transcripts in TCMR

CATs are expressed in CD4+ and CD8+ T CTL and in EM T cells, but many are also expressed in NK cells (Figure 1A). To determine the contribution of NK cells to the CAT signal in biopsies with TCMR, we analyzed expression of an NK-selective transcript set (n = 322) during TCMR (Supplementary Table S2, Supplementary Figure S1B). Similar to CATs, we divided NK-selective transcripts into subsets based on their expression in NK cells and examined expression of the top four subsets in TCMR biopsies (Figure 1B). Unlike CATs, NK-selective transcripts showed little increase in TCMR biopsies, indicating low numbers of NK cells, compatible with previous immunostaining estimates of NK cell numbers in TCMR (6,7). The small increases that were observed are within the range expected from their low expression level in CTL.

Generation of quantitative CATs (QCATs)

To select a set of CATs that would be detected quantitatively when diluted in biopsies (QCATs), we performed serial dilutions of CD8+ CTL RNA in kidney RNA. QCATs were selected by signal intensity >1000 in a CTL/Nephrectomy dilution sample (1:1) and a Pearson correlation coefficient ≥0.98 between signal intensity and dilution ratio. This algorithm identified 28 transcripts. Three of these showed signal intensities below 50 in all TCMR biopsies and were thus considered unreliable for quantitation and excluded from the set.

The remaining QCATs (n = 25) included cytolytic 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 IFNG (Table 1). The standard curve of the QCAT geometric mean signal is shown in Figure 2A (Pearson correlation coefficient r = 0.99).

Table 1.  Expression of QCATs in nephrectomies (Neph), TCMR, treated TCMR and ATN cases
Gene symbolNephStdErrTCMRStdErrTreated TCMRStdErrATNStdErr
  1. Geometric mean signal values for the indicated sample group and transcript ± SE are shown.

CD2 6010.038955.5114 9.4 93 8.7
CD3D 98 7.248178.514013.2108 9.5
CD8A 63 8.336263.0129 8.2 64 4.4
CST7 50 5.116619.0 8012.4 48 5.9
CXCR6 30 2.1 62 6.8 41 3.1 21 0.9
GNLY 20 1.2 9024.1 4312.7 36 8.5
GZMA 25 5.823236.4 9329.4 40 5.4
GZMB 19 1.417740.7 4115.3 22 2.3
GZMK 53 7.438456.113735.27211.3
GPR171 21 3.612321.2 38 3.5 19 2.2
HOP10129.527144.9 98 8.625274.8
IFNG 18 1.6 7916.8 28 3.4 11 0.4
IL2RB 77 6.429145.3118 2.1 96 7.6
ITK 16 2.210217.7 25 3.8 22 1.8
KIAA0101 37 7.116524.5 8011.410118.2
KLRK1 64 7.828531.114126.6 90 6.9
LCK 36 8.122035.5 57 2.8 51 4.9
NELL2 56 7.210816.3 38 1.6 47 6.9
NKG7 30 1.915425.9 6510.3 47 6.2
NUSAP1 52 7.010315.5 52 7.3 81 9.9
PRF1 84 7.931259.913930.7101 8.4
RRM2 34 7.312022.7 6010.3 6713.4
STAT4 35 4.2 9211.8 34 6.1 41 2.9
TRA@10417.341254.515915.712811.7
TRGV9 44 7.715023.4 43 6.6 45 6.2
image

Figure 2. Generation and characterization of QCATs. (A) Geometric mean of the QCAT signal values plotted against the dilution ratio of CD8+ CTL mRNA. Trendline equation and calculated Pearson correlation value are displayed. (B) QCAT expression was characterized in three effector lymphocyte populations. QCAT signal geometric mean was calculated in CD4+ and CD8+ CTL and NK cell samples. Raw signal values are shown. Graph depicts the geometric mean of the transcript set +/− SD.

Download figure to PowerPoint

The geometric mean expression of QCATs was highest in CD8+ CTL but reached signal intensities >1000 in CD4+ CTL and NK cells (Figure 2B). Thus QCATs potentially represent the burden of all of these cell types and probably also of CD8+ EM T cells (see accompanying manuscript).

Estimation of CTL burden in renal allografts

We analyzed the CAT burden in renal allograft biopsies by calculating the %CTL mRNA in biopsies with TCMR, biopsies with acute tubular necrosis (ATN) and control kidneys. The calculated %CTL mRNA was higher in TCMR compared to treated TCMR and ATN cases (p < 0.0005) (Figure 3A). The individual QCAT scores in untreated TCMR were always above the range observed in transplant kidneys with ATN and control kidneys, except for one case in which the QCAT score was low (Figure 3B). This biopsy had ‘v1’ with little inflammation and tubulitis below the diagnostic Banff threshold for TCMR (‘isolated v’), and was classed as TCMR exclusively because it had endothelial arteritis. The low %CTL mRNA reflects the low inflammatory burden in the biopsy. Two other cases (4th and 14th TCMR samples on the graph) also showed low QCAT score as would be expected given the scores of i2 and i1, respectively, for these samples. It is important to note that two ATN cases show QCAT scores similar to TCMR cases with low QCAT scores.

image

Figure 3. Calculation of %CTL mRNA using QCATs in renal allografts undergoing TCMR. CD8+ CTL, nephrectomy, TCMR, treated TCMR and ATN samples were analyzed using QCATs. (A)%CTL mRNA was calculated in groups of kidney biopsies from nephrectomy samples, untreated and treated TCMR cases and ATN cases. Graph depicts the calculated %CTL mRNA +/− SD. *p < 0.0005, Student's t-test using the geometric mean signal for QCATs in the specified groups. (B)%CTL mRNA was calculated in individual kidney biopsies from nephrectomy samples, ATN cases, treated and untreated TCMR cases. Graph depicts the calculated %CTL mRNA in individual samples. (C) Average QCAT geometric mean signal for all samples based on interstitial inflammation (i) score. Graph shows average QCAT geometric mean +/− SD.

Download figure to PowerPoint

Cases read as TCMR but treated before the biopsy all had low QCAT scores. In cases of antibody-mediated rejection (ABMR), QCAT expression is also increased, reflecting the strong inflammatory response that is also observed in ABMR (data not shown, manuscript in preparation) (14).

In agreement with a quantitative transcript set, the calculated %CTL mRNA correlated with the degree of interstitial inflammation as determined by the histologic i score (r = 0.677) (Figure 3C) and also correlated with the burden of CD8+ T cells in the graft estimated by immunostaining (Spearman rank correlation coefficient r = 0.69, p = 0.01). This agreement is good, given that i scores are only semi-quantitative and are observer dependent.

In a hierarchical cluster analysis, samples were clustered in groups by their QCAT expression as expected: kidney samples formed one cluster, CTL samples formed a second group (Figure 4a). With the exception of the ‘isolated v’ case, biopsies with TCMR formed a distinct subgroup within the kidney cluster. QCAT expression in the TCMR biopsies more closely resembled CTL lines than did other biopsies. The remaining kidney samples separated into three subgroups: control kidneys clustered together, and cases diagnosed as ATN clustered in two groups according to their QCAT burden (low or intermediate). Treated TCMR cases had intermediate QCAT burdens and grouped with the ATN samples. The one ‘isolated v’ case of TCMR with low %CTL mRNA grouped with control kidneys and ATN samples with low QCAT expression, reflecting its paucity of interstitial inflammation and tubulitis.

image

Figure 4. Principal component (PC) analysis of QCATs in biopsy samples. Individual biopsies and CD8+ CTL were analyzed using QCATs. (A) Clustering (condition tree and gene tree) was based on distance. Branch lengths represent the degree of similarity between individual samples or transcripts. Black, green and red tiles indicate no change, decrease in expression and increase in expression, respectively. Colors for top branch lines and categories at bottom of heatmap correspond to the groups of the samples: Turquoise, blue, orange, yellow and red color indicates nephrectomy, ATN, TCMR, treated TCMR and CD8+ CTL samples, respectively. (B) PC analysis of nephrectomies, biopsies with ATN, treated TCMR or TCMR and CTL. (C) PC analysis of individual QCATs based on their expression in nephrectomies, biopsies with ATN, treated TCMR or TCMR and CTL. Circles highlight the main subgroups formed by this analysis.

Download figure to PowerPoint

These observations were confirmed by displaying relationships among measurements in samples by principal component (PC) analysis. Samples were grouped by PC1 according to their similarity to CD8+ CTL: control kidneys and ATN samples on the left, treated TCMR and TCMR intermediate and CD8+ CTL on the right (Figure 4B). The ‘isolated v’ TCMR case with low inflammatory burden and the treated TCMR cases clustered with ATN kidneys, separate from the other TCMR cases. ATN kidneys clustered in two groups based on PC1, one group close to control kidneys, and one group of ATN samples forming a separate cluster. This separation of samples by PC2 was absent in control kidneys and the cases with TCMR, indicating that the QCAT expression in some ATN cases differed not only quantitatively, but also qualitatively from TCMR cases.

Heterogeneity in QCAT expression in renal allografts

The gene tree in the hierarchical cluster analysis (Figure 4A) revealed that QCATs are heterogeneous and grouped by their expression in the TCMR cases. PC analysis on individual QCATs (Figure 4C) confirmed these different behaviors of individual QCATs. One QCAT subset contained mainly membrane-associated molecules (CD2, CD3D, CD8, IL2RB). Interestingly PRF1 and GZMK clustered in this group, unlike the other cytotoxic molecules (GZMA, GZMB, GNLY), which form a second cluster, suggesting differential regulation. GZMA, GZMB and GNLY were the transcripts with the highest expression level in CTL and had the highest contribution to PC1 in Figure 4B.

To further evaluate the heterogeneity of individual QCATs, we examined the relationship of individual QCATs to the mean QCAT score by correlating the individual transcript expression values with the QCAT mean (Table 2). Some transcripts encoding cell surface receptors including IL2RB, GPR171 and CD3D correlated well with the mean value of the QCATs (Figure 5A). Others such as IFNG, NELL2 and HOP showed low correlations (Figure 5B). The QCATs PRF1 and GZMB are of particular interest (17). Although expression levels of these transcripts were similar in individual TCMR samples, many samples showed lower %CTL mRNA for these two transcripts than for the QCAT set (Figure 5C).

Table 2.  Correlation of calculated%CTL mRNA for TCMR cases based on individual QCATs with the CTL mRNA ratios calculated using the entire QCAT list
Transcript nameCorrelation with QCAT%CTL mRNA
IL2RB0.942
GPR1710.937
CD3D0.933
CD20.927
CST70.906
LCK0.897
TRA@0.876
NKG70.867
CXCR60.840
PRF10.836
KLRK10.833
GZMK0.829
CD8A0.821
ITK0.820
GZMA0.818
NUSAP10.746
GZMB0.740
GNLY0.728
RRM20.728
KIAA01010.722
TRGV90.687
STAT40.659
IFNG0.651
NELL20.501
HOP0.199
image

Figure 5. Differences in individual QCAT expression within TCMR biopsies. The %CTL mRNA values were calculated for TCMR biopsies using the QCAT transcript set (solid line) or individual QCATs (dotted line). The three individual QCATs with highest correlation (A), the three with lowest correlation (B), and cytotoxic molecule transcripts PRF1 and GZMB (C) are shown.

Download figure to PowerPoint

Thus the QCAT transcript set expressed as %CTL mRNA values is not accurately predicted for an individual sample by selected CTL transcripts. Thus individual transcripts have information not carried by the QCAT mean, and vice versa.

Confirmation by RT-PCR

Expression of selected QCATs (CD8A, GZMB, IFNG, GZMK and PRF1) was confirmed by real time RT-PCR on TCMR cases. Pearson correlation values for this PCR subset of 5 genes was 0.902 (p < 0.001) compared to the microarray values from the QCAT transcript set (Table 3). High correlation (>0.744) of individual transcripts by RT-PCR with QCAT transcript set microarray results was also observed.

Table 3.  Pearson correlation of QCAT microarray geometric mean signal with selected QCATs assayed by RT-qPCR (PCR QCAT) (p < 0.001 for all correlations)
 log QCAT (microarray)PCR QCAT
log QCAT (microarray)1.0000.902
PCR QCAT0.9021.000
CD8A PCR0.8890.960
GZMB PCR0.7540.883
IFNG PCR0.8540.961
GZMK PCR0.8100.854
PRF1 PCR0.7440.710

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The goal of these experiments was to develop a quantitative estimate of the burden of CATs in rejecting biopsies. Sets of CATs highly expressed in human CTL gave consistent estimates of the CAT burden in individual human renal allograft biopsies, while preserving their overall hierarchy compared to CTL in vitro. Although CATs are expressed by CD4+ and CD8+ CTL, EM T cells and NK cells (see accompanying manuscript), in renal biopsies they probably reflect mainly the burden of CTL and EM lymphocytes, because NK cell-selective transcripts were not detectable in TCMR, and NK cells are uncommon by immunohistochemistry in TCMR (6,7). In the analyzed biopsies, sets of CATs with high expression in CD8+ CTL were consistently increased in TCMR samples compared to ATN and control kidneys, with the exception of one unusual ‘isolated v’ case. QCAT expression in biopsies was 3–15% of that in CTL in vitro, the expression pattern resembling a dilution of in vitro CD8+ CTL. QCAT expression correlated strongly with the T-cell burden (CD8+ cells) by immunohistochemistry. Hierarchical clustering and PCA using QCATs grouped biopsies according to their similarity to CD8+ CTL, the TCMR biopsies being most similar to CTL, followed by ATN and nephrectomies. The QCAT burden most probably reflects the CTL and EM T-cell burden in the tissue, thus providing a composite measurement of all T cells that correlates with interstitial T-cell infiltration by immunohistochemistry.

A high QCAT burden likely indicates TCMR, but the relationship of QCAT burden to TCMR cannot be resolved until we define the elusive underlying mechanisms of TCMR. T cells with cognate recognition of graft allo-antigens are probably infrequent, with the majority of T cells in the graft reflecting homing to the inflammatory compartment created by the cognate events. Orosz et al. concluded that in TCMR only 0.2% of allograft-infiltrating cells are reactive to donor antigens (22), suggesting that the majority of the T cells are being recruited to the inflammatory site by antigen-independent mechanisms. The need for only a small number of allo-specific T cells may explain how rejection episodes occur following profound T-cell depletion (23). This may also explain the ‘isolated v’ case of TCMR, with a low %CTL mRNA, minimal interstitial inflammation and tubulitis. Such cases are diagnosed as TCMR by convention, but their nature is not established (8). The significance of ‘isolated v-lesions’ is currently under study in the Banff consensus process. Any T-cell infiltrate must have some level of QCATs because of recruitment of EM cells, which express many cytotoxic molecule transcripts. Nephrectomies and transplants with ATN showed minor CAT expression that probably reflects focal infiltration by EM T cells in injured or old tissue. Focal mononuclear infiltrates are frequent in older human kidneys, kidneys with ATN, or even during normal aging (3). Thus for diagnostic purposes the QCAT score as an estimate of the T-cell burden must be interpreted in the context of established diagnostic systems, but the ultimate test of TCMR will require that we define the underlying cognate events and mechanisms.

Interstitial inflammation in an allograft is one parameter used for diagnosis, and increasing Banff grades of interstitial inflammation in TCMR correspond to inferior prognosis (24,25). Since the QCATs measure the T-cell interstitial inflammatory burden, we can assume that they would relate to prognosis the same way the histological difference between borderline and Banff grade I TCMR does. These two Banff categories are discriminated by quantitative differences in interstitial infiltrates and extent of tubulitis, which correlates with the extent of interstitial inflammation (26,27). Therefore, evidence suggests that QCATs have prognostic value but this must be confirmed in larger patient cohort with respective follow-up available. The ability to estimate the T-cell burden in transplant biopsies using mRNA transcript expression would provide an objective quantitation method that is not observer dependent. We plan for QCAT measurements to compliment the histologic assessment of interstitial inflammation thus increasing the ability for an accurate diagnosis.

While expression of CAT sets in allografts preserves the overall hierarchy of gene expression in CTL in vitro, individual transcripts can vary from the QCAT mean. It would be surprising if the pattern of individual transcripts expressed in allografts did not change between in vitro and in vivo conditions, given the arbitrary conditions in vitro, and among biopsies in vivo. But such individual variations are neutralized when sets of 10 CATs are averaged, so that at the gene set level graft-infiltrating T cells simulate a dilution of in vitro-generated CTL in kidney RNA. Hierarchical clustering and PC analysis of QCATs in TCMR also indicate that QCAT set expression in TCMR is related to expression in CD8+ CTL in vitro, but that individual CATs carry unique additional information. The heterogeneity of the PCA results for individual QCATs is encouraging that individual transcripts carry information not conveyed by the QCAT mean. Interesting CD8+ CTL transcripts increased in TCMR are highlighted in this study, such as GZMK, EOMES and NKG2D/KLRK1, which are less well characterized in TCMR (28,29). Thus quantitation of individual transcripts, including well-known examples (GZMB, PRF1) (17) and lesser known examples such as GZMK, may provide more detailed mechanistic, diagnostic or prognostic information, particularly when normalized by overall QCATs expression.

QCATs represent a refinement of our recently published larger pathogenesis-based transcript set (PBT) based on CATs. These were originally derived from mouse transplants and cell culture experiments and showed potential for diagnostic application to human transplants (14). QCATs are improved by being derived from human-allostimulated CD8+ CTL, with a more stringent selection against transcripts expressed in other cell types (see accompanying manuscript). This increases sensitivity and robustness of interpretation in TCMR, and could serve to measure CTL burden in other T-cell-rich inflammatory processes. However, despite the high correlation between (i) scores and QCAT expression, cases with intermediate to high amounts of infiltrate (i2/i3) showed high variability in QCAT expression that may reflect variability in histologic assessment or differential T-cell activation states in individual patients. QCAT expression will represent effector/EM T cells recruited into an inflamed site and thus non-rejection cases with low degrees of inflammation will show low QCAT expression whereas rejection cases with high inflammation will show higher QCAT expression. It is however important to note that QCAT expression is only part of the overall transcriptional changes observed in TCMR (14). Although not proven as diagnostic in this study due to the limited patient population, we believe QCATs will show their diagnostic potential in future studies.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The authors would like to thank Sujatha Guttikonda for technical support and Drs. Colin Anderson and Troy Baldwin for critical reading of the manuscript.

This research has been supported by funding and/or resources from Genome Canada, Genome Alberta, the University of Alberta, 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. Michael Mengel received a research stipend from the Dr. Werner Jackstädt Kidney foundation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Figure S1. Algorithms to identify transcripts associated with CD8+ CTL (CATCD8s) and NK cell selective transcripts. A) CATCD8s were derived from the overlap of probesets showing increased expression (p < 0.01) in CD8+ CTL (CATCD8s) compared to purified B cells (> B cell), monocytes (> Monoc), and Nephrectomy samples (> Neph). In addition, all transcripts with signal intensity &ft;200 in nephrectomy, B cells, monocytes, and IFN-γ treated monocytes were removed. B) The NK-selective transcript set was derived from the overlap of probesets showing increased expression (p < 0.01) in NK cells compared to purified B cells (> B cell), monocytes (> Monoc), and Nephrectomy samples (> Neph). All transcripts with signal intensity > 200 in nephrectomy samples, B cells, monocytes, IFN-g treated monocytes, CD4+ or CD8+ CTL were removed. The number of transcripts composing each transcript set is indicated.

Figure S2. Algorithms to identify transcripts associated with CD8+ CTL (CATCD8s) and NK cell selective transcripts. Relationship between PBT scores, histopathologic lesions, histopathologic and clinical diagnosis, and classifier predictions. Biopsies for cause (n = 143) were sorted based on the QCAT score (from lowest to highest). According to this order, scores for all PBTs (CAT1, CAT2, GRIT1, GRIT2, KT1, KT2, described in Mueller et. al., AJT 7-2712, 2007) are illustrated for each individual biopsy for cause. The panel above the graph illustrates the relationship of the PBT scores to the presence of Acute Tubular Necrosis (ATN), the degree of interstitial infiltrate (i score), tubulitis (t score), intimal arteritis (v score), histopathology diagnosis, retrospective clinical-pathologic diagnosis and the probability of rejection (%), predicted from the classifiers.

Table S1. Expression of CATCD8s (CD8+CTL associated transcripts) in CD8+CTL, NK cells, nephrectomy samples, and biopsies with T cell mediated rejection (TCMR)Numbers in bold represent fold changes compared to nephrectomy samples; "signal" indicates signal intensity of the probeset in nephrectomy samples.

Table S2. Expression of NK selective transcripts in CD8+CTL, NK cells, nephrectomy samples, and biopsies with T cell mediated rejection (TCMR)Numbers in bold represent fold changes compared to nephrectomy samples; "signal" indicates signal intensity of the probeset in nephrectomy samples.

Methods RNA Processing for Real-time RT-PCR

Please note: Blackwell Publishing are not responsible for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
AJT_2129_sm_FiguresS1-S2.ppt233KSupporting info item
AJT_2129_sm_Methods.doc35KSupporting info item
AJT_2129_sm_TableS1.pdf233KSupporting info item
AJT_2129_sm_TableS2.pdf364KSupporting info item

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.