SEARCH

SEARCH BY CITATION

Keywords:

  • Antibody mediated rejection;
  • C4d;
  • endothelial cells;
  • gene expression;
  • microarrays;
  • transplantation

Abstract

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

Anti-HLA alloantibody is a risk factor for graft loss, but does not indicate which kidneys are experiencing antibody-mediated rejection (ABMR). C4d staining in biopsies is specific for ABMR but insensitive. We hypothesized that altered expression of endothelial genes due to alloantibody acting on the microcirculation would be sensitive indicator of ABMR. We identified 119 endothelial-associated transcripts (ENDATs) from literature, and studied their expression by microarrays in 173 renal allograft biopsies for cause. Mean ENDAT expression was increased in all rejection but was higher in ABMR than in T-cell-mediated rejection and correlated with histopathologic lesions of ABMR, and alloantibody. Many individual ENDATs were increased in ABMR and predicted graft loss. Kidneys with high ENDATs and antibody showed increased lesions of ABMR and worse prognosis in comparison to controls. Only 40% of kidneys with high ENDAT expression and chronic ABMR or graft loss were diagnosed by C4d positivity. High ENDAT expression with antibody predicts graft loss with higher sensitivity (77% vs. 31%) and slightly lower specificity (71% vs. 94%) than C4d. The results were validated in independent set of 82 kidneys. High renal endothelial transcript expression in patients with alloantibody is indicator of active antibody-mediated allograft damage and poor graft outcome.


Abbreviations: 
ABMR

Antibody-mediated rejection

ENDAT

Endothelial cell-associated transcripts

FDR

False discovery rate

HLA

Human leukocyte antigen

TCMR

T-cell-mediated rejection

Introduction

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

In patients with kidney transplants, presence of alloantibody is associated with an increased rate of late graft loss. The potent destructive power of alloantibody was first recognized as hyperacute rejection due to preformed antibody against HLA class I or ABO blood group antigens, at time of transplantation (1). The problem of antibody-mediated rejection (ABMR) at later times was recognized in kidney transplants (2) and is also reported in heart, pancreas and lung allografts (3,4). ABMR is dominated by endothelial damage in microcirculation (5,6), often associated with inflammatory cells in capillaries (7–11). Endothelial damage may be mediated by complement (12–15) and leukocytes recruited via Fc and complement receptors (2,16–18).

Predicting which renal transplants are at risk for antibody-mediated graft deterioration is an unmet need. Circulating HLA antibodies develop in 25% of renal transplant recipients and are associated with increased graft failure (19), but have low predictive value for graft loss. The diagnosis of ABMR in biopsies was facilitated by the recognition that many cases with ABMR features have diffuse deposition of C4d, a breakdown product of the complement component C4, in peritubular capillaries (20,21). C4d+ ABMR can occur early or late, and can be acute or insidious. Features of late C4d+ ABMR are variable, including transplant glomerulopathy (22–25), peritubular capillary basement membrane multilayering, and scarring (interstitial fibrosis and tubular atrophy) (2,22,23,25). Diagnosis of ABMR currently requires diffuse C4d staining plus alloantibodies (10,25). However, many cases of transplant glomerulopathy with alloantibodies are C4d negative (22,26,27), suggesting that C4d is specific but not sensitive.

If C4d staining misses some cases of antibody-mediated injury, and the presence of alloantibody does not identify which grafts are undergoing antibody-mediated damage, we need a new method for identifying which kidneys are being damaged by alloantibody. Recognizing the key role of endothelial changes in ABMR, we hypothesized that altered expression of endothelial genes in biopsies from patients with alloantibody would identify kidneys incurring antibody-mediated damage and at risk for graft loss, whether they were C4d+ or negative. The present study explored whether expression of endothelial genes was increased in biopsies manifesting antibody-mediated graft injury, and whether such changes could be seen in C4d negative as well as C4d positive biopsies.

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

Patients and sample collection

The study was approved by the institutional review board of the University of Alberta (issue # 5299) and by the University of Illinois, Chicago (protocol # 2006–0544). Written informed consent was obtained from all study patients. All consenting renal transplant patients undergoing a biopsy for cause (acute or stable impaired function, proteinuria) as standard of care between 09/2004 and 05/2008 were included. The study consisted of 212 biopsies from 169 patients at University of Alberta (n = 208 biopsies) or at University of Illinois (n = 4 biopsies). Normal kidney tissue was obtained from 7 native nephrectomies with renal carcinoma. Biopsies were obtained under ultrasound guidance by spring-loaded needles (ASAP Automatic Biopsy, Microvasive, Watertown, MA). In addition to the cores for pathology, we collected one additional 18-gauge biopsy core for gene expression analysis and placed it immediately in RNALater, kept at 4°C for 4–24 hrs, then stored at −20°C.

Histopathology and clinical data collection

Biopsies were graded by a pathologist (BS) according to Banff classification (10,25,28). C4d staining was performed on frozen sections using a monoclonal anti-C4d antibody (Quidel, San Diego, CA, USA). All samples were adequate with the exception of 5 biopsies lacking arteries. Clinical data from the time of transplantation to data analysis was entered into a Laboratory Information Management System developed in our laboratory. Diagnostic classifications, methods for anti-HLA antibody testing, RNA preparation, microarray processing, and Real-Time PCR are detailed in supplementary material.

Analysis of the transcriptome and clinical data

Original and test set microarrays (Affymetrix HG U133 Plus 2.0 Array) were pooled into two normalization batches and preprocessed using robust multi-chip averaging and implemented in GeneSpring (GX) 7.3.1 and BRB-Array Tools. Gene expression is given as fold change versus controls. Inter-quartile range (range between the third and first quartiles) non-specific filtering was then used to remove probe sets that have low variability across the entire dataset. Hierarchical clustering and principal component analysis were used to discover classes within the dataset.

We previously derived gene sets on mouse and human transplants and cultured cells reflecting major biologic events in kidney allograft rejection–cytotoxic T-cell infiltration (29–31) and interferon-γ effects (32). We identified a literature-based endothelial cell-associated transcript (ENDAT) set (323 probe-sets; 119 unique genes), in which genes were identified based on their selective expression in cultured human endothelial cells when compared with non-endothelial cells or by serial analysis gene expression libraries comparing tag frequencies in endothelial versus non-endothelial libraries (33,34). Gene expression within each gene set was summarized as a score representing the geometric mean of fold changes across all probe sets. The probe sets are available at http://transplants.med.ualberta.ca/.

Welch's t-test was used to find genes whose expression differed between C4d+ABMR and TCMR. All corrected p-values refer to false discovery rates (FDR), e.g. a corrected p-value of 0.05 signifies that 5% of the probe sets identified as significant at the 0.05 level will be false positives. We used a semi-supervised principal component method to predict graft survival from gene expression data (35). Death-censored graft survival analyses were performed after last biopsy with Kaplan–Meier method using log rank test. End point for graft failure was defined as return to dialysis.

Results

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

Patient demographics and diagnostic classifications

We studied 173 consecutive transplant biopsies for cause (renal dysfunction and/or proteinuria) from 130 patients by microarrays as well as histopathology, with no exclusions or technical failures. Biopsies were taken between 6 days and 31 years post-transplant (median 16.4 months) (Supporting Table S1). We included 43 repeat biopsies from 33 patients because they were performed for separate clinical indications. Biopsies were read without knowledge of the molecular findings.

Rejection was diagnosed in 41 biopsies from 32 recipients. C4d+ ABMR was diagnosed in 15 biopsies (2 acute, 13 chronic) according to Banff criteria including diffuse C4d staining, presence of alloantibodies, and morphologic evidence of tissue injury. All had HLA antibodies and dysfunction (rapid onset (n = 8) or slow onset (n = 5)) or proteinuria (n = 2). T-cell-mediated rejection (TCMR) was diagnosed in 23 biopsies with clinical rejection episode, either as typical TCMR (n = 20) or borderline TCMR (n = 3). Three C4d+ biopsies with TCMR lesions (i.e. mixed ABMR and TCMR) were excluded from ABMR versus TCMR class comparisons (Supporting Table S1).

Mean endothelial cell-associated transcript expression is increased in rejection but with higher expression in C4d+ ABMR

Our hypothesis was that endothelial gene expression would be a feature of both TCMR and C4d+ ABMR, but would be greater in ABMR. We defined 119 endothelial cell-associated transcripts (ENDATs) from the literature, and measured their mean expression score by microarrays (Table 1). ENDAT scores were higher in biopsies with diffuse C4d staining compared to biopsies with focal or no C4d staining. ENDAT score was increased in both TCMR and C4d+ ABMR compared to normal controls, but was higher in C4d+ ABMR compared to TCMR.

Table 1.  Comparison of average gene set scores in C4d+ (diffuse+) renal allograft biopsies versus C4d focal and C4d negative biopsies, and in C4d+ antibody-mediated rejection versus T-cell-mediated rejection biopsies
NNormal Nephrectomies 7C4d+1 17C4d focal 6C4d negative 150p2
  1. ABMR = antibody-mediated rejection; TCMR = T-cell-mediated rejection; NS = not significant.

  2. For gene sets, numbers indicate geometric mean of fold change over normals ± SD.

  3. 1Diffuse positive by Banff criteria (10).

  4. 2C4d+ vs combined C4d focal and C4d negative, Independent Samples T test.

  5. 3No clinical rejection group does not include any C4d+ cases.

  6. 4ABMR vs. TCMR, Independent Samples T test.

  7. 5mixed ABMR and TCMR cases were excluded from ABMR vs. TCMR comparisons.

Endothelial cell-associated transcripts1.0 ± 0.11.3 ± 0.11.0 ± 0.21.1 ± 0.1<0.001
CTL associated-transcripts1.0 ± 0.12.8 ± 1.01.6 ± 1.02.2 ± 1.5 0.02 
Interferon-γ induced transcripts1.0 ± 0.12.3 ± 0.41.4 ± 0.41.7 ± 0.5<0.001
 
NNormal Nephrectomies 7No clinical rejection3 132C4d+ ABMR 155TCMR 23p4
 
Endothelial cell-associated transcripts1.0 ± 0.11.1 ± 0.11.3 ± 0.11.1 ± 0.10.001
CTL associated-transcripts1.0 ± 0.11.8 ± 0.82.7 ± 1.14.5 ± 2.50.01 
Interferon-γ induced transcripts1.0 ± 0.11.6 ± 0.42.3 ± 0.42.5 ± 0.6NS

Principal component analysis using 119 ENDATs showed that both TCMR and C4d+ ABMR samples were different from controls on principal component 1, but principal component 2 discriminated between C4d+ ABMR and TCMR (Figure 1A).

image

Figure 1. (A) Principal component analysis (PCA) using literature selected endothelial cell-associated transcripts (n = 119) showed a clear discrimination between C4d+ ABMR and C4d- TCMR samples, especially based on PCA component 2. Both ABMR and TCMR samples were different from controls on PCA component 1. (B). Hierarchical clustering of 165 renal allograft biopsies with available anti-HLA antibody test at time of biopsy using 25 endothelial genes that are changed C4d+ABMR vs. C4d negative TCMR (Pearson correlation, Average linkage). The cluster with high expression of endothelial-transcripts (red box) included 22 C4d negative Ab+ biopsies clustered together with all 18 C4d+ABMR biopsies indicating that both groups had similar degree of changes in endothelial genes. This cluster also included 19 biopsies with no antibody or C4d with elevated expression of endothelial genes.

Download figure to PowerPoint

ENDAT scores were similar in acute (n = 2) and chronic (n = 13) C4d+ ABMR (1.4 ± 0.2 vs. 1.3 ± 0.1, respectively). ENDAT-scores did not differ in C4d positive biopsies between patients who had been treated (3 of 18) vs. non-treated patients, although treatment options were variable.

Cytotoxic T-cell-associated transcripts and interferon-γ induced transcripts were increased in C4d+ ABMR compared to C4d negative biopsies, and, as previously reported (31), in both TCMR and C4d+ ABMR compared to biopsies without rejection (Table 1). Cytotoxic T-cell-associated transcripts were higher in TCMR than ABMR.

ENDATs correlated (in the mild to moderate range) with lesions of C4d+ ABMR: peritubular capillaritis, glomerulitis, transplant glomerulopathy, peritubular capillary basement membrane multilayering, interstitial inflammation, and scarring (Table 2). In contrast, ENDATs did not correlate with tubulitis, the characteristic lesion of TCMR. In multivariate analysis, C4d positivity, interstitial inflammation, and scarring were independently associated with elevated ENDATs. Thus mean ENDAT expression was increased in both TCMR and ABMR, but with selectivity for ABMR.

Table 2.  Association between endothelial gene set score and pathologic lesions of antibody-mediated rejection in 173 renal allograft biopsies
Univariate analysis LesionsENDAT score
Correlation coefficientp
  1. ENDAT = Endothelial cell-associated transcripts; NS = not significant.

  2. 1Peritubular capillary basement membrane multilayering was indicated as number of basement membrane layers assessed by electron microscopy.

  3. Other lesions were scored according to Banff criteria. Univariate correlates of p < 0.3 were entered into the multivariate model. Because ci and ct were tightly correlated (r = 0.9) only ci was entered into the model.

C4d deposition0.376<0.001
Peritubular capillaritis (ptc)0.252 0.002
Peritubular capillary basement membrane multilayering10.266 0.004
Glomerulitis (g)0.248 0.001
Interstitial inflammation (i)0.358<0.001
Tubulitis (t)0.135NS
Intimal arteritis (v)0.092NS
Glomerulopathy (cg)0.261 0.001
Mesangial matrix increase (mm)0.1730.02
Interstitial fibrosis (ci)0.330<0.001
Tubular atrophy (ct)0.286<0.001
Fibrous intimal thickening (cv)0.014NS
Arteriolar hyalinosis (ah)−0.050 NS
 
Multivariate analysis LesionsR2= 0.41 ENDAT score p
 
C4d deposition<0.001
Interstitial inflammation (i)<0.001
Interstitial fibrosis (ci)<0.001

Mean ENDAT expression correlates with presence of circulating alloantibodies

We studied whether biopsies from patients with alloantibodies displayed higher mean ENDAT expression. Serum was available for HLA antibody testing in 165/173 biopsies. Biopsies with panel reactive (n = 82) or donor specific HLA antibodies (n = 54) had elevated ENDATs (Figure 2). Antibodies against class II antigens were associated with elevated ENDATs, but antibodies against class I were not.

image

Figure 2. Correlation of endothelial cell-associated transcripts (ENDATs) to anti-HLA alloantibodies. (A,B) Panel reactive antibodies (PRA). (C,D) Donor specific antibodies (DSA).

Download figure to PowerPoint

Individual ENDATs are more increased in C4d+ ABMR than in TCMR

Since mean ENDAT expression was elevated in rejection but higher in C4d+ ABMR than TCMR, we studied whether individual ENDATs increased by rejection were also more elevated in C4d+ ABMR. Twenty-five individual ENDATs were differentially expressed in C4d+ ABMR versus TCMR (FDR 0.05) (Supporting Table S2). Seventeen were more increased in C4d+ ABMR, the highest being von Willebrand's factor (VWF). Others included CAV1 (caveolin 1), MCAM (melanoma cell adhesion molecule), PECAM1 (platelet/endothelial cell adhesion molecule), SELE (selectin E), PALMD (palmdelphin), CD34, and cadherin 5.

Surveying transcripts more elevated in C4d+ABMR than in TCMR

We surveyed all transcripts changed in C4d+ ABMR versus TCMR, to see whether this approach would demonstrate endothelial changes as more characteristic of C4d+ ABMR than other changes. We studied the expression of 9718 interquartile-range filtered probe-sets to identify all different transcripts in C4d+ ABMR versus TCMR. This class comparison found 98 transcripts differentially expressed in C4d+ ABMR compared to TCMR (FDR 0.05). Of these, 45 were increased in both TCMR and ABMR, and 5 of 45 (11%) were ENDATs and were higher in ABMR versus TCMR (VWF, CDH5, RHOJ, PALMD, CAV1). Other transcripts more increased in ABMR versus TCMR were also expressed by endothelium even though they were not included in our strict criteria for ENDATs e.g. cadherin 13, Duffy blood group, SOX7, THBD (thrombomodulin), and MALL. Thus endothelial transcripts are over-represented among the transcripts increased in ABMR compared to TCMR.

Validation of microarray expression data by RT-PCR

The increased expression of endothelial genes in ABMR (that were arbitrarily selected from those elevated by microarrays) was confirmed by RT-PCR. The mean expression of VWF, EDN1, CAV1, CDH5, CDH13, PALMD, PECAM1, and SELE was higher in ABMR than in TCMR biopsies (Figure 3).

image

Figure 3. RT-PCR confirmation of increased expression of endothelial transcripts in antibody-mediated rejection (ABMR). The expression of endothelial cell-associated transcripts, arbitrarily chosen from those increased by microarrays, was increased in ABMR vs. T-cell-mediated rejection (TCMR) measured by RT-PCR.

Download figure to PowerPoint

Expression of individual ENDATs is associated with risk of graft loss

We assessed the association of individual ENDATs with death-censored graft survival in 124 kidney transplants (repeat biopsies were excluded) with a semi-supervised principal component method (35) using leave-one-out cross-validation. Ten of 119 ENDATs predicted graft loss (Figure 4). Ranked by p-value, VWF was most strongly associated with graft loss.

image

Figure 4. Survival risk prediction analysis using 119 endothelial-transcripts with Cox proportional hazards model with the semi-supervised principal component method with a leave-one-out cross-validation.

Download figure to PowerPoint

Clustering with ENDATs shows all C4d+ ABMR and some C4d negative biopsies have endothelial stress

We performed hierarchical clustering on 165 biopsies using the ENDATs differentially expressed in C4d+ABMR versus TCMR. We focused on the cluster (see the box in Figure 1B) with high ENDAT expression (colored in red in Figure 1B) that included all C4d+ biopsies (n = 18) (15 C4d+ ABMR and 3 mixed ABMR and TCMR). This cluster also included 22 C4d negative biopsies associated with panel reactive HLA antibody, as well as 19 C4d negative with no HLA antibody. This cluster was not enriched in TCMR: 12 TCMR and 22 with no TCMR or borderline changes were in the cluster versus 17 TCMR and 60 with no TCMR or borderline changes outside the cluster (p = 0.16). Thus the high ENDAT cluster that included all C4d+ ABMR identified a set of C4d negative biopsies from patients with antibody and with endothelial abnormalities, and did not select most cases of TCMR.

Defining training and validation sets for predicting ABMR and graft survival

Exclusion of repeat biopsies and inclusion of new biopsies (n = 39) created a population of 163 biopsies from 163 kidney transplant patients. We split the data into a training and test set using stratified sampling (36), by matching time post-transplant, presence of panel reactive HLA antibodies, C4d status, cg-scores (transplant glomerulopathy), and follow up time after biopsy (p > 0.05): this created a training set of 81 biopsies and a test set of 82 biopsies. Mean time post-transplant at biopsy was 73 and 58 months in the training set and validation set, respectively. Mean follow up time after biopsy was 25.2 and 25.8 months in the training set and validation set, respectively.

Prediction of C4d+ ABMR in an independent validation set using ENDAT expression

We defined a cutoff for ENDAT expression to detect C4d+ ABMR with a high sensitivity in the training set. ROC curve analysis specified the cutoff as > 1.14 fold over controls (Figure 5). This cutoff for ENDAT expression identified C4d+ ABMR cases with 88% sensitivity, 56% specificity, 18% positive predictive value, and 98% negative predictive value (Area under curve (AUC): 0.781) (p = 0.009).

image

Figure 5. Prediction of C4d+ antibody-mediated rejection (ABMR) by using ENDAT expression: (A) Training set. (B) Validation set: Prediction of C4d+ABMR was validated by applying the training set derived cutoff value to an independent cohort of 82 biopsies.

Download figure to PowerPoint

Prediction of C4d+ABMR was validated by applying the training set derived cutoff value to an independent cohort of 82 biopsies: High ENDAT expression predicted C4d+ ABMR with a 86% sensitivity, 65% specificity, 19% positive predictive value, and 98% negative predictive value ((AUC: 0.739) (p = 0.03).

The specificity of ENDAT expression to predict C4d+ABMR was not perfect due to the flawed gold standard (C4d staining). As we previously reported (22), C4d misses many kidneys with late ABMR: these kidneys had circulating antibodies and chronic ABMR pathology, but were C4d negative.

High ENDAT expression detects C4d negative chronic ABMR

Having identified ENDAT elevation in C4d+ ABMR, we studied whether ENDATs would be elevated in C4d negative biopsies with antibody-mediated injury. We reasoned that ENDAT changes in C4d+ ABMR may be present in some biopsies that have alloantibody but are C4d negative.

We used the training set derived ENDAT expression cutoff (to detect C4d+ ABMR, see above section) to classify which biopsies had endothelial abnormalities. We used two features to classify all biopsies: high ENDAT expression (E) and circulating panel reactive HLA antibody (Ab) to classify 163 kidneys into three groups: Kidneys with no Ab, kidneys with Ab and no E, and kidneys with Ab and E.

In the training set, kidneys with Ab and E showed increased histologic lesions of ABMR including more peritubular capillaritis, transplant glomerulopathy, and scarring in comparison to kidneys with Ab and no E or kidneys with no Ab (Table 3 and Figure 6A). However, C4d was positive only in 38% of kidneys with Ab and E that developed chronic ABMR pathology (transplant glomerulopathy) (Figure 6C). In other words, C4d missed 62% of chronic ABMR cases with endothelial alterations. Kidneys with Ab and E also showed higher expression of interferon-γ induced transcripts than kidneys with Ab and no E or kidneys with no Ab. These observations were validated in independent set of 82 biopsies (Figure 6B and D, Table 3).

Table 3.  Gene set scores and histopathology in 163 kidneys grouped by high ENDAT expression (E) and antibody (Ab) status
NTraining set (n = 81)Validation set (n = 82)
Kidneys with no Ab 30Kidneys with Ab and no E 21Kidneys with Ab and E 30Kidneys with no Ab 31Kidneys with Ab and no E 31Kidneys with Ab and E 20
  1. Numbers are given as mean ± SD unless noted otherwise.

  2. 1Comparisons were done using log10 transformed values for time post-transplant.

  3. 2Significant difference between kidneys with Ab and E versus kidneys with Ab and no E (p < 0.05).

  4. 3Significant difference between kidneys with Ab and E versus kidneys with no Ab (p < 0.05).

  5. Comparisons were done by independent samples T test or Mann-Whitney U test, where applicable.

Post-transplant time to biopsy (months)142 ± 6868 ± 8873 ± 62342 ± 6471 ± 8464 ± 563
Average gene set scores
 Endothelial cell-associated transcripts1.1 ± 1.11.0 ± 1.1  1.3 ± 1.12,31.1 ± 1.21.0 ± 1.1  1.3 ± 1.12,3
 Interferon-γ induced transcripts1.6 ± 1.41.4 ± 1.3  2.0 ± 1.32,31.5 ± 1.41.5 ± 1.3  2.1 ± 1.22,3
 CTL-associated transcripts2.0 ± 1.81.5 ± 1.52.5 ± 1.721.6 ± 1.71.6 ± 1.4  2.6 ± 1.62,3
Pathology
 Glomerulitis (g-score)0.2 ± 0.70.3 ± 0.80.6 ± 0.830.0 ± 0.20.2 ± 0.6  0.8 ± 0.92,3
 Interstitial inflammation (i-score)1.1 ± 0.91.0 ± 0.81.5 ± 0.920.9 ± 0.91.2 ± 0.91.6 ± 0.93
 Tubulitis (t-score)0.9 ± 1.00.7 ± 0.70.8 ± 1.0 0.5 ± 0.90.8 ± 0.90.8 ± 0.9 
 Vasculitis (v-score)0.1 ± 0.60.1 ± 0.50.2 ± 0.6 0.0 ± 0.20.1 ± 0.50.2 ± 0.4 
 Peritubular capillaritis (ptc-score)0.2 ± 0.50.3 ± 0.7  1.0 ± 1.22,30.1 ± 0.30.6 ± 0.9  1.3 ± 1.22,3
 Transplant glomerulopathy (cg-score)0.3 ± 0.60.4 ± 0.7  1.2 ± 1.22,30.3 ± 0.60.6 ± 1.0  1.1 ± 1.02,3
 Mesangial matrix increase (mm-score)0.5 ± 0.80.5 ± 0.7  1.4 ± 1.12,30.5 ± 0.80.8 ± 0.81.3 ± 1.03
 Interstitial fibrosis (ci-score)1.2 ± 1.01.1 ± 0.7  1.8 ± 1.02,31.1 ± 0.91.1 ± 0.7  2.0 ± 0.82,3
 Tubular atrophy (ct-score)1.3 ± 1.01.1 ± 0.7  1.9 ± 0.92,31.1 ± 0.91.2 ± 0.7  2.0 ± 0.82,3
 Fibrous intimal thickening (cv-score)1.5 ± 0.91.2 ± 0.81.5 ± 0.9 1.4 ± 1.11.3 ± 0.71.4 ± 1.1 
 Arteriolar hyalinosis (ah-score)0.9 ± 1.11.6 ± 1.01.8 ± 1.131.1 ± 1.11.2 ± 1.01.5 ± 1.2 
 Peritubular capillary basement membrane multilayering (number of layers)3.4 ± 2.54.4 ± 2.85.2 ± 3.4 2.9 ± 2.34.4 ± 4.44.8 ± 3.03
image

Figure 6. Transplant glomerulopathy is increased in kidneys with high ENDAT expression (E) and circulating antibodies (Ab) in comparison to controls. A–B: Mean transplant glomerulopathy scores are shown in the training set (A) and validation set (B). C–D: Incidence of transplant glomerulopathy with or without C4d in the training set (C) and validation set (D). C4d was positive only in approximately 40% of kidneys with Ab and E, that developed chronic ABMR pathology (transplant glomerulopathy).

Download figure to PowerPoint

High ENDAT expression with alloantibody predicts poor kidney graft survival

Kidneys with Ab and E showed significantly decreased death-censored graft survival in comparison to kidneys with Ab and no E or kidneys with no Ab (Figure 7A). Of 13 graft losses in the training set, 10 (77%) had alloantibody and high renal ENDAT expression in biopsy. Of these 10, only 4 (40%) were C4d positive (Figure 8A). Thus many grafts failed due to ABMR were C4d negative. The prognostic value of ENDATs was confirmed in the validation set (Figures 7B and 8B). Thus high renal ENDAT expression is a predictor for poor graft outcome in patients with alloantibody. Furthermore, high ENDAT expression with alloantibody predicts graft loss with higher sensitivity (77% vs. 31%) and slightly lower specificity (71% vs. 94%) than C4d.

image

Figure 7. Death censored graft survival in patients grouped according to the presence of panel reactive HLA antibody (Ab), and high renal ENDAT expression (E) (described in results). (A): Training set of 81 patients. (B) Validation set of 82 patients.

Download figure to PowerPoint

image

Figure 8. Percentage of graft loss in kidneys grouped by HLA antibody (Ab) and high ENDAT expression (E), and incidence of C4d in failed kidneys. Kidneys with Ab and E showed increased graft loss in comparison to other groups. Approximately 60% of grafts failed due to ABMR were C4d negative. (A) Training set of 81 patients. (B) Validation set of 82 patients.

Download figure to PowerPoint

High ENDAT expression with no alloantibody is not related to poor survival

Of 41 early biopsies before 3-month post-transplant, 16 (none showed donor specific antibodies) showed high ENDAT expression including 6 acute tubular necrosis, 2 borderline, 1 acute bacterial infection, 2 glomerulonephritis, 1 TCMR, and 4 minor changes. Of 6 biopsies with BK virus nephropathy, 2 showed increased ENDAT expression (one of these had class I antibodies). We did not have any CMV cases, but one case with acute bacterial infection showed high ENDAT expression. Thus, as expected, high ENDAT expression can be seen in the absence of anti-HLA antibodies and rejection, i.e. in some cases of ischemic injury and infections.

However, high ENDAT expression in patients with no alloantibody is not an indicator of graft damage and graft loss. We grouped biopsies according to increased ENDAT score (E), HLA antibody positivity (A), and C4d positivity (C). This divided 161 biopsies (with available C4d and antibody testing) into five groups: 13 AEC, 37 AE, 21 E only, 50 A only, 40 no AEC. AEC and AE biopsies shared the phenotype with increased transplant glomerulopathy, scarring, and poor survival. Survival was not different between AEC vs. AE (p = 0.197), thus C4d positivity did not have additional adverse prognostic impact over high ENDAT with antibody (Figure 9). However, biopsies with only E or only A did not show increased transplant glomerulopathy, scarring (data not shown) or graft loss (Figure 9), compared to biopsies with no AEC.

image

Figure 9. Death censored graft survival in patients grouped according to the presence of panel reactive HLA antibody (A), C4d staining (C) and high renal ENDAT expression (E) (described in results). This divided 161 biopsies (with available C4d and antibody testing) into five groups: 13 AEC (Ab+ENDAT+C4d+), 37 AE (Ab+ENDAT+C4d-), 21 E only (Ab-ENDAT+C4d-), 50 A only (Ab+ ENDAT-C4d-), 40 no AEC (Ab-ENDAT-C4d-).

Download figure to PowerPoint

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 earlier finding that some C4d-negative kidneys share features of antibody-mediated injury with biopsies that fulfill the criteria for C4d+ ABMR suggested that C4d positivity may be only the extreme of the ABMR spectrum. We postulated that changes in endothelial transcripts might identify kidneys experiencing antibody-mediated injury, including some C4d negative biopsies. Our strategy was to develop a biopsy cohort including all kidney transplant patients, representing the complete spectrum of clinical problems requiring biopsy at any time in their course. The mean ENDAT expression was elevated in C4d+ biopsies and correlated with ABMR lesions but not tubulitis, the hallmark lesion of TCMR. Mean ENDAT expression was higher in biopsies with HLA antibodies, particularly class II, consistent with previous associations of class II antibodies with transplant glomerulopathy and graft loss (22,37). Using ENDAT score, we developed a cutoff for predicting C4d+ ABMR. ENDAT changes identified a considerable number of C4d negative biopsies that resembled C4d+ ABMR biopsies. High ENDAT expression in biopsies from patients with antibody was associated with transplant glomerulopathy, scarring, and increased graft loss, even in C4d negative biopsies. Thus C4d negative cases that shared endothelial changes and antibody with C4d+ ABMR displayed more transplant glomerulopathy and scarring than controls. We conclude that in patients with alloantibodies, abnormalities in expression of endothelial genes identify not only C4d+ ABMR but some kidney transplants developing antibody associated graft injury despite negative C4d staining. Diffuse C4d staining is only the ‘tip of the iceberg’ of late ABMR, identifying approximately 40% of biopsies with ABMR, but missing some cases of C4d negative ABMR at high risk for graft loss. Moreover, based on this study, the combined burden of C4d+ and C4d negative ABMR accounts for the majority of graft losses in kidney transplants biopsied for clinical indications (17 of 26, 65%).

The elevation of the ENDATs is of value in determining which biopsies for cause in patients with antibody may have antibody-mediated injury, even when they are C4d negative. The poor prognosis associated with elevated ENDATs in patients with HLA antibody probably does not reflect the inherent impact of endothelial changes, but our current inability to interrupt antibody production, with the result that the endothelial stress is sustained indefinitely. In this respect it is instructive to note that some ENDAT elevation was also observed in TCMR (albeit less on average), but because TCMR is treatable such changes did not indicate a poor outcome. Thus high ENDAT without antibody and antibody without endothelial changes had little impact on graft survival.

The biology underlying increased ENDAT expression could include a number of processes, including endothelial injury/dysfunction, repair, activation and/or angiogenesis. For example, seven ENDATs elevated in C4d+ ABMR (VWF, MCAM, CDH5, SELE, PECAM1, CD34, CAV1) are associated with endothelial activation and endothelial cell-cell interactions (16,26). SOX7, another endothelial transcript that increased in C4d+ ABMR, plays redundant roles in endothelial differentiation and some members of SOX family are involved in postnatal angiogenesis in mice (38). We conclude that ENDAT changes in renal transplants occur in rejection and in other forms of renal injury, and their impact on transplant glomerulopathy and graft loss is principally in patients with circulating HLA antibodies.

If ABMR is defined as alloantibody-mediated damage to the microcirculation, then the present results confirm previous suspicions that C4d positivity, albeit very useful, is not synonymous with ABMR. Absence of diffuse C4d staining does not exclude antibody-mediated injury, recalling previous studies showing transplant glomerulopathy with alloantibody but no C4d staining (22,26,27). Moreover, as shown in other studies, focal C4d staining was not a reliable indicator of ABMR (39), and in this study was not associated with elevated ENDAT expression. C4d positivity is not a guarantee of ABMR: diffuse C4d staining can occur with no morphologic injury or impaired outcome in ABO-incompatible allografts (40).

These results do not directly imply complement independent mechanisms in C4d negative ABMR. It is possible that complement still mediates damage in such cases despite the lack of diffuse C4d staining (41). It is shown that treatment by anti-HLA antibody alone can also trigger release of VWF, P-selectin, and increased leukocyte adherence to human endothelial cells (42). Thus new mechanistic studies are needed to explorate whether different mechanisms operate in C4d positive versus C4d negative ABMR.

The low frequency of anti-HLA in patients requiring an early biopsy for clinical indication, and of early ABMR reflects the effectiveness of current antibody screening technologies, and our center practice of avoiding desensitization and positive crossmatches at the time of transplant. The question of whether C4d negative ABMR exists in early post-transplant, should be explored in a further study, which would require a population of positive crossmatch transplants.

ENDAT expression in biopsy provides a new tool for understanding the pathogenesis of late kidney graft loss and ABMR, and for predicting graft outcomes and defining ABMR even in C4d negative biopsies in patients with antibodies. This could potentially change practice in kidney transplant follow-up clinics, encouraging more biopsies. Many transplants with evidence of deterioration do not get biopsied because clinicians believe that biopsies offer too little clinical utility. Our results show that combining alloantibody with ENDATs identifies cases of ABMR missed by C4d staining. The combination of C4d negative ABMR plus typical C4d+ ABMR accounts for the majority of kidney transplant losses after a biopsy for cause, offering improved prediction of prognosis. Longer study may show that some patients who appear as ‘false positives’ (i.e. Ab+ ENDAT+ biopsies without ABMR histopathology features or graft loss) may be at an earlier stage in progression of these time dependent features. Such patients may be the most important to detect, before potentially irreversible changes appear, and could be the focus of trials to establish effective treatments by suppressing antibodies and reducing adverse endothelial changes before the full phenotype emerges (2,22,23,25).

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 wish to thank Ms. Anna Hutton, Ms. Stacey Lacoste, and Mr. Vido Ramassar for technical assistance; Dr. Zija Jacaj for assistance with retrieval of clinical records; Drs. Gunilla Einecke, Konrad Famulski, Luis Hidalgo, Deborah James, Michael Mengel, Thomas Mueller, and Jeff Reeve for discussions and 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, Capital Health, Roche Molecular Systems, Hoffmann-La Roche Canada Ltd., Alberta Innovation & Science, the Roche Organ Transplant Research Foundation, the Kidney Foundation of Canada, and Astellas Canada. Dr. Halloran also holds the Muttart Chair in Clinical Immunology and was supported by a Canada Research Chair in Transplant Immunology. Dr. Sis' research has been supported by funding from Roche Organ Transplantation Research Foundation, University of Alberta Hospital Foundation, and Kidney & Urology Foundation of America–Renal Pathology Society.

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

Table S1. Patient demographics, clinical data, pathology diagnoses and Banff lesions in the original study population.

Table S2. The individual endothelial genes that are differentially expressed between C4d+ antibody-mediated rejection and C4d negative T-cell-mediated rejection.

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

FilenameFormatSizeDescription
AJT_2761_sm_SuppMat.doc39KSupporting info item
AJT_2761_sm_TableS1.doc57KSupporting info item
AJT_2761_sm_TableS2.doc70KSupporting 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.