SEARCH

SEARCH BY CITATION

Keywords:

  • Fibrosis;
  • gene expression;
  • graft failure;
  • inflammation;
  • renal injury

Abstract

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

We previously reported that kidney transplants with early acute injury express transcripts indicating injury repair—the acute kidney injury signal. This study investigated the significance of this signal in transplants with other conditions, including rejection and recurrent disease. The injury signal was elevated in biopsies in many different conditions, including T cell-mediated rejection and potentially progressive diseases such as antibody-mediated rejection and glomerulonephritis. A high injury signal correlated with poor function and with inflammation in areas of fibrosis, but not with fibrosis without inflammation. In multivariate survival analysis, the injury signal in late kidney transplant biopsies strongly predicted future graft loss, similar to a published molecular risk score derived in late kidneys. Indeed, the injury signal shared many individual transcripts with the risk score, e.g. ITGB6, VCAN, NNMT. The injury signal was a better predictor of future graft loss than fibrosis, inflammation or expression of collagen genes. Thus the acute injury signal, first defined in early reversible injury, is present in many diseases as a reflection of parenchymal distress, where its significance is dictated by the inducing insult, i.e. treatable/self-limited versus untreatable and sustained. Progression in troubled transplants is primarily a function of ongoing parenchymal injury by disease, not fibrogenesis.


Abbreviations
ABMR

antibody-mediated rejection

AKI

acute kidney injury

CKD

chronic kidney disease

eGFR

estimated GFR

IRRATs

injury–repair response-associated transcripts

TCMR

T cell-mediated rejection

Introduction

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

We recently reported that microarray analysis of biopsies from injured human kidney transplants reveal a complex coordinate injury–repair response [1] analogous to wound healing, which provides an acute kidney injury (AKI) signal that is not detectable by histology. The AKI signal comprises injury-repair-associated transcripts (IRRATs) that represent the response of kidney tissue to an insult [2-4]. In early kidney transplants with AKI, the AKI signal correlated with depression of function, future recovery of function and interstitial inflammation [1]. Biological functions of the AKI transcripts are shared with carcinoma [5], e.g. synthesis of extracellular matrix components, reexpression of developmental programs, increased cell migration and TGFB1 effects [6-11], recalling the concept of “cancer as wounds that do not heal” [6]. The AKI transcripts include known AKI biomarkers previously reported in urine and body fluids, indicating that biomarkers are often products of genes induced by the parenchymal response to wounding, e.g. HAVCR1 (KIM-1) [12]. All kidneys diagnosed as early pure AKI and expressing the AKI signal recovered, and almost all kidneys [27, 28] survived long term, consistent with successful resolution of the acute injuries of donation and implantation with no late progressive deterioration during follow-up.

The injury-repair mechanisms in kidney parenchyma must respond to injury caused by disease mechanisms as well as external agents, raising the issue of whether the AKI signal is expressed in other types of renal pathology, particularly progressive diseases [2]. This question has also been raised by the association of AKI biomarkers with chronic changes such as aging [13] or chronic kidney disease (CKD), and with experimental models of progressive disease [14] and rejection [15]. This study was undertaken to determine the extent to which the AKI signal is triggered by diseases such as rejection and recurrent glomerulonephritis. We studied prospectively collected kidney transplant indication biopsies taken months or years after transplantation, particularly the relationship of the AKI signal to histopathology, function and prognosis. Our hypothesis was that renal parenchyma has one fundamental the injury-repair response to wounding, which is inherently self-limited but persists when the insult in ongoing. We further postulated that this response would provide a signal that the parenchyma is being injured, the significance of which would be determined by the nature of the insult, i.e. self-limited, treatable, or potentially progressive and untreatable.

Methods

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

Human biopsies for clinical indications, histopathology and diagnoses

The collection of 315 human kidney grafts, represented by 403 indication biopsies was described in [16]. Biopsies were obtained under ultrasound guidance by spring-loaded needles (ASAP Automatic Biopsy, Microvasive, Watertown, MA, USA). The study was approved by the University of Alberta Health Research Ethics Board (Issue # 5299), by the University of Illinois Chicago Office for the Protection of Research Subjects (protocol # 2006–0544) and the University of Minnesota, protocol HSR#0606 M87646). The 6-week pristine protocol biopsies were described in [1].

Biopsies were assessed by a pathologist, blinded to the results of molecular studies. Inflammation was scored in fibrotic parenchyma alone (i-IFTA). The i-IFTA score and the interstitial fibrosis were recorded as continuous variables [17]. The i-IFTA score > 0 corresponded to ≥10% of parenchyma affected. Interstitial fibrosis score, ci > 0, corresponded to ≥6% of parenchyma affected. Banff histopathologic diagnoses were described previously by us [18].

We analyzed 50 biopsies with early AKI, 35 biopsies with Banff TCMR, 65 with ABMR (including [17] C4d positive and 48 suspicious for ABMR, aka C4d negative ABMR), 22 with mixed (ABMR/TCMR) rejection, 42 with borderline rejection and 189 remaining nonrejecting biopsies, including 41 with glomerulonephritis and other glomerular diseases (GN), 34 with fibrosis alone and no features of a disease process (IFTA, ci score > 1), 12 with polyoma virus, 72 with no major abnormalities and no features of a disease process (ci score < 2), 10 with at least one microcirculation lesion being PRA positive but DSA negative (possible ABMR) and 20 “others” with uncommon entities such as C4d deposition with no pathology (n = 5), thrombotic microangiopathy (n = 3), suspicious viral nephropathy (n = 4), transplant glomerulopathy (n = 2), posttransplant lymphoproliferative disorder (n = 1), oxalosis (n = 1), obstruction (n = 1), acute pyelonephritis (n = 1), tubulointerstitial nephritis (n = 1) and biopsies inadequate for histology assessment (n = 1). All diagnoses reflect updated donor-specific antibody status.

Renal function, the molecular risk classifier of graft failure and graft survival

Renal function was defined by estimated creatinine clearance, using a four-variable MDRD equation at the time of biopsy and 6 months later. We also calculated the change in eGFR at 6 months after the biopsy. We have previously built a gene-based classifier to predict graft loss and used this molecular classifier to assign a prognostic molecular risk score to each biopsy [19]. Here we used the same approach to calculate the risk score based on all 315 patients.

RNA extraction and microarrays

As previously reported [20], in addition to cores for diagnostic histopathology, one 18-gauge biopsy core was collected for gene expression analysis, placed immediately in RNALater (Life Technologies, Carlsbad, CA, USA), kept at 4°C for 4–24 h, then stored at −20°C. RNA extraction, quality control and HG_U133_Plus_2.0 GeneChip (Affymetrix, Santa Clara, CA, USA) processing were described previously [20]. Detailed protocols are available at (www.affymetrix.com).

Microarray data analysis

Microarray data files (GSE36059) for 403 indication biopsies and eight nephrectomies were preprocessed using Robust Multiarray Analysis in Bioconductor [19]. The AKI signal was defined as the summarized expression (geometric mean) of the top 30 injury-repair response-associated transcripts (IRRATs) [1]. To derive the collagen signal (COL signal), we compared 28 biopsies with pure AKI to nephrectomy controls [1] and found 19 probe sets (corrected p-value, FDR = 0.05) showing increased expression of COL1A1, COL1A2, COL3A1, COL5A1, COL5A2, COL6A1, COL6A3, which overrepresented the GO category: fibrillar collagen (GSE30718). All probe sets correlated with the AKI signal (r = 0.6–0.4), with exception of two probe sets representing COL6A1. We removed these COL6A1 probe sets and calculated the summarized collagen signal (COL signal) using the remaining probe sets.

Statistical analyses

Significance of differences between the groups of biopsies was tested by the Welch t-test with Bonferroni correction. A chi-square test was used for contingency tables. p-values of <0.05 were assumed significant.

For the analysis of survival curves, we selected the last biopsy per patient (n = 315) and last, late biopsies (≥1 year posttransplant, n = 186). Using these two populations, we calculated the median AKI signal to plot the Kaplan–Meier survival curves. The graft survival was assessed as the time between biopsy and graft failure/censoring. The median time of follow-up of patients with graft failures was 1218 (8–2196) days and there were 72 failures.

Univariate associations between the AKI signal and histological and clinical features with graft survival were assessed by Cox regression analysis. In multivariate regression (the full model) all six input variables were used—a stepwise method was not used.

To test the ability of single transcripts to predict the graft failure, we used the Cox regression analysis.

Results

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

Relationship of the AKI signal to histological diagnoses

The AKI signal, calculated here as the mean expression of the injury-repair-associated transcripts (IRRATs), was originally defined in early transplants with pure AKI [1]. In this study, we examined the AKI signal in a prospective study of 403 unselected indication biopsies from 315 patients, representing the spectrum of diseases that affect kidney transplants, including 50 early kidney biopsies with AKI. The demographics of the 403 indication biopsies have been reported previously [16].

We compared the range of the AKI signal in disparate histological diagnoses to biopsies with no major histologic abnormalities (NOMOA) (Figure 1). Many indication biopsies had high AKI signals including biopsies diagnosed with antibody-mediated rejection (ABMR), T cell-mediated rejection (TCMR), mixed TCMR and ABMR, and borderline rejection. Many nonrejecting biopsies diagnosed with recurrent glomerulonephritis (GN) or atrophy-fibrosis also displayed high AKI signals. On average, the AKI signal was higher in all diagnostic groups than in NOMOA controls (see means at the top of Figure 1). Pristine protocol biopsies (PP) and control kidneys (“Con”) had similar AKI signal to NOMOA biopsies confirming them as the “normal” range for a kidney transplant and for normal controls (normal kidney tissue from tumor nephrectomy specimens.

image

Figure 1. Relationship between the AKI signal and the histological diagnoses. The diagnostic categories were ordered by descending median values of the AKI signal. The boxes represent the interquartile range, the solid horizontal lines within the boxes indicate the median of the AKI signal and the whiskers represent 1.5 times the interquartile range. Shaded horizontal box indicates the interquartile range of the AKI signal in the NOMOA group across all diagnoses. Upper panel: all diagnoses were compared to the NOMOA group. Means and Bonferroni-corrected p-values for 12 comparisons (Student's two-tailed t-test) are shown.

Download figure to PowerPoint

Correlates of the AKI signal in indication biopsies

We analyzed the significance of the AKI signal in all biopsies and in the selected groups. The selection criteria are presented in Table 1. We excluded the 50 AKI biopsies because AKI had been analyzed previously. Then we divided the remaining 353 biopsies into biopsies into four groups; group 1: TCMR; group 2: biopsies without potentially progressive or untreatable diseases; group 3: biopsies with potentially progressive or untreatable diseases; and group 4: biopsies with unclassified findings. We also defined all biopsies that had no histologically defined rejection: group 5. In each case we could study either the biopsy population, or the patients who had these biopsies. In addition we defined a group of 186 patients having late biopsies, since we had previously shown that this population had most of the graft losses [21]: group 6.

Table 1. Selection criteria for biopsy and patient groups
 AllGroup 1: TCMRGroup 2: no diagnosis of a potentially progressive disease: borderline, NOMOA, IFTAGroup 3: diagnosis of a potentially progressive disease: ABMR, mixed rejection, GNGroup 4: unclassified diagnoses: possible ABMR, BK, otherGroup 5: nonrejecting biopsies, i.e. no histologic diagnosis of rejection: (i.e. excluding TCMR, ABMR, mixed, borderline)
  1. Biopsy and patient groups were based on the histological diagnoses shown in Figure 1.

  2. 1One biopsy (not included) was insufficient for histology.

  3. Bolded numbers indicate biopsies or patients used in a particular analysis.

Biopsy groups based on histology diagnosis (excluding AKI) (see the x-axis, Figure 1)
Number of biopsies40313514712842188
Shown/analyzed inFigure 1Table 2Tables 2, 3Table 2Table 2Table 4
Patient groups based on findings in latest biopsy
All patients3152112610229158
(number of failures)7241443931
Shown/analyzed inFigure 2, Table 8Table 2Tables 2, 3Table 2Table 2Tables 4, 5
Patient groups presenting for biopsy after 1 year, analyzing only the latest biopsy
Group 6 patient presenting for biopsy > 1 year1864759116109
(number of failures)6421342728
Shown/analyzed inFigure 2, table 6

We analyzed the relationships of the AKI signal to renal function (Table 2). The mean AKI signal was highest in group 1. The eGFR at biopsy was similar in all groups. However, the 6-month eGFR was highest in groups 1 (TCMR) and 2 (with no progressive diseases), due to recovery of eGFR. The most progression to graft failure was in group 3 (potentially progressive diseases). The four grafts with TCMR progressed to failure due to nonadherence and ABMR, but not due to pure TCMR [22].

Table 2. Kidney function in the four diagnosis groups: group 1 with TCMR; group 2 with no progressive or untreatable diseases; group 3 with potentially progressive and untreatable diseases; and group 4 of unclassified findings
 Group 1: TCMRGroup 2: No diagnosis of a progressive disease: borderline, NOMOA, IFTAGroup 3: Diagnosis of a potentially progressive or nontreatable diseases ABMR, mixed, GNGroup 4: Unclassified findings possible ABMR, BK, other
  1. The groups 1, 3, 4 were compared to the group 2 (to the group of biopsies without progressive or untreatable diseases).

  2. Recovery of eGFR in the TCMR group—p-value by the paired t-test <0.0001, recovery of eGFR in the group without progressive diseases—p-value <0.0001, recovery of eGFR in the group of potentially progressive diseases.

  3. *p < 0.001, **p < 0.01, ***p > 0.05 by the two-tailed Welch t-test with Bonferroni correction for three comparisons in each of rows 5–8. Means ± SEM are shown.

  4. 1Fold change versus nephrectomies (linear values).

  5. 2Recovery of eGFR is measured as the difference between the eGFR at the time of biopsy and 6 months after the biopsy within a group.

  6. p value = 0.41, recovery of eGFR in the unclassified group p-value = 0.52.

  7. 3Significant difference in the distribution of failures among the groups by the chi-square test, p-value < 0.001.

Kidney function in relationship to biopsy diagnosis
Number of biopsies3514712842
Median time to biopsy (months)4.919.061.111.0
AKI signal11.93 ± 0.23*1.29 ± 0.421.46 ± 0.19**1.65 ± 0.31**
eGFR at biopsy33.3 ± 1.1***34.3 ± 1.134.1 ± 1.3***31.3 ± 1.9***
eGFR at 6 months242.4 ± 2.4***42.8 ± 1.434.5 ± 1.5*33.8 ± 2.4**
delta eGFR from biopsy to 6 months10.5 ± 1.9***8.2 ± 1.4−0.9 ± 1.1*3.6 ± 2.4***
Survival analysis (death censored) classified by diagnosis in latest biopsy only
Number of patients2112610229
Failures (death censored)34 (19%)14 (11%)43 (42%)9 (31%)

We analyzed the biopsies with no diagnosis of progressive diseases (group 2) in more detail (Table 3). Biopsies diagnosed as a borderline had the highest AKI signal, probably because some borderline biopsies actually have TCMR [23]. Biopsies with NOMOA had the lowest signal. Both had good recovery of eGFR, and a low frequency of progression to subsequent failure. Biopsies with atrophy-fibrosis with no diagnosed diseases had moderately elevated AKI signals but no recovery; 6/29 progressed to failure (probably due to diseases that could not be diagnosed by histology at the time of this biopsy).

Table 3. Kidney function in group 2: biopsies in which no diagnosis of a potentially progressive or untreatable disease
 BorderlineNOMOAAtrophy-fibrosis (IFTA)
  1. The borderline and IFTA groups were compared to the NOMOA group.

  2. *p < 0.001, **p < 0.01, ***p > 0.05, ****p = 0.06, by the two-tailed Welch t-test with Bonferroni correction for two comparisons in each of rows 5–8. Means ± SEM are shown.

  3. 1Fold change versus nephrectomies (linear values).

  4. 2Recovery of eGFR is measured as the difference between the eGFR at the time of biopsy and 6 months after the biopsy within a group.

  5. Recovery of eGFR in the borderline and the NOMOA group by the paired t-test—p-value <0.0001, recovery of eGFR in the IFFTA group p-value = 0.33.

Kidney function in relationship to biopsy diagnosis
Number of biopsies427233
Median time to biopsy (months)6.916.579.2
AKI signal11.61 ± 0.37*0.93 ± 0.391.34 ± 0.37*
eGFR at biopsy32.5 ± 2.3***37.2 ± 1.630.3 ± 1.7**
eGFR at 6 months248.1 ± 2.9***44.5 ± 1.832.9 ± 2.4*
ΔeGFR from biopsy to 6 months15.0 ± 3.3****7.0 ± 1.52.6 ± 2.6***
Survival analysis (death censored) classified by diagnosis in the latest biopsy only
Number of patients326529
Failures (death censored)2 (6%)6 (9%)6 (21%)

The AKI signal is high in scarred and inflamed biopsies with no histologic diagnosis of rejection

The strong histological correlates of the AKI signal in entire cohort of patients were inflammation in scarred areas (i-IFTA, r = 0.31, p < 0.0001) and the amount of atrophy/fibrosis (ci, r = 0.28, p < 0.0001). AKI signal was not related to the histologic features of acute tubular injury (not shown).

Because atrophy-fibrosis in indication biopsies is usually inflamed [17], we studied whether the AKI signal was primarily associated with atrophy-fibrosis or with inflammation. We compared biopsies with no scarring or inflammation (ci = 0, i-IFTA = 0), biopsies with scarring alone (ci > 0, i-IFTA = 0) and biopsies with scarring with inflammation (ci > 0, i-IFTA > 0) (Table 4). We studied group 5, which excludes all histologic rejection (ABMR, TCMR, mixed and borderline), to avoid confusion between rejection-induced inflammation and the reactive inflammation in atrophy scarring. The mean AKI signal was only increased when scarring was accompanied by inflammation. The AKI signal correlated with inflammation in scarred areas (r = 0.4), whereas biopsies with scarring that lacked inflammation showed low AKI signals.

Table 4. Distribution of 188 nonrejecting biopsies (i.e. no histologic diagnosis of rejection: group 5), according to their inflammation/fibrosis status and AKI signal
  Number of biopsies
Assessment of scarring (ci) and inflammation with scarred areas (i-IFTA) in biopsies with no rejectionAKI signal linearBottom tertile of AKI signalMiddle tertile of AKI signalTop tertile of AKI signal
  1. The AKI signal was discretized into tertiles. Of 189 biopsies, one biopsy was left out as inadequate for histology assessment.

  2. 1Significant difference in the distribution among the AKI tertiles by a chi-square test (p-value < 0.001).

  3. 2Significantly different from remaining groups of biopsies by the two-tailed Welch t-test with Bonferroni correction for two comparisons (p < 0.0001). Means ± SEM are shown.

No scarring ci = 0 i-IFTA = 0 (n = 57) 10.95 ± 0.07301512
Scarring with no inflammation ci > 0 i-IFTA = 0 (n = 48) 11.04 ± 0.0923178
Scarring with inflammation ci > 0 i-IFTA > 0 (n = 83) 11.62 ± 0.102103043
Total biopsies (n = 188) 636263

Dividing the biopsies into tertiles by their AKI signal confirms that biopsies with uninflamed scarring were distributed like biopsies with no scarring, mainly with low-AKI signals, whereas biopsies with inflamed scarring mainly had high AKI signals. Thus atrophy-fibrosis with inflammation represents a response to recent or ongoing injury, manifested by the AKI signal, whereas uninflamed atrophy-fibrosis reflects remote injury and is not accompanied by the AKI signal.

Because fibrosis accompanied by inflammation is associated with an increased risk of failure [17], we examined the AKI signal and i-IFTA in the 31 of 158 kidneys in group 5 that progressed to failure. Most (22/31) had both high AKI signal and inflammation in areas of atrophy-fibrosis (Table 5). The AKI signal in the latest biopsy was highly predictive of subsequent progression to failure, with Cox regression p value 8.97 × 10−9.

Table 5. Findings in the latest biopsy of kidneys with no histologic diagnosis of rejection (group 5) that subsequently failed: impact of scarring, inflammation and AKI signal
 Distribution of failures by inflammation, scarring and AKI signal
Scoring only inflammation in fibrotic areasBottom tertile of the AKI signal (four failures)Middle tertile of the AKI signal (five failures)Top tertile of the AKI signal (22 failures)
  1. 1Significant difference in the distribution among the AKI tertiles by a chi-square test (p-value < 0.05). Only one (last) biopsy per patient was analyzed.

ci = 0 i-IFTA = 0211
ci > 0 i-IFTA = 0001
ci > 0 i-IFTA > 012420

Relationship of the AKI signal to risk of graft loss in patients presenting with late biopsies

Because of the association of the AKI signal with graft loss, we extended our analysis to all patients biopsied after 1-year posttransplant (group 6), the patient population at risk for progression to failure [19]. The AKI signal was higher in the 64 transplants that progressed to failure after the biopsy, compared to the 122 transplants that have not progressed to failure, 1.8 versus 1.1, respectively (p < 0.001).

Therefore we studied the ability of the AKI signal in group 6 biopsies to predict future graft loss, compared to variables such as scarring (ci), inflammation (i-IFTA), eGFR at the biopsy, and proteinuria (Table 6). There were 57 kidney failures in 171 kidneys for which we had complete histology and proteinuria data. We also included those diagnosed with potentially progressive diseases (ABMR, mixed rejection, or GN) based on our earlier results [16]. (We did not include polyoma virus nephropathy because there was only one failure.)

Table 6. Univariate and multivariate analysis of associations of findings in last biopsy, eGFR and proteinuria with death censored graft survival in group 6 of patients presenting late
 p-value (likelihood ratio)Hazard ratio (95% confidence limits)
  1. Only latest biopsy per patient was included in the analysis. There were 57 events and 171 grafts analyzed from the 186 patients in group 6, because 15 patients (including 7 failures) were excluded due to missing histology or proteinuria data. Bolded numbers indicate significant p-values.

Univariate analysis
Potentially progressive disease0.00202.2 (1.4–3.8)
Scarring (ci)8.99×10−51.8 (1.3–2.4)
eGFR at biopsy2.00×10−90.92 (0.90–0.95)
Proteinuria1.17×10−75.2 (2.8–9.4)
AKI signal6.66×10−164.5 (3.1–6.4)
i-IFTA0.00261.02 (1.01–1.02)
Multivariate analysis: full model
Potentially progressive disease0.111.6 (0.9–2.8)
Scarring (ci)0.500.86 (0.56–1.3)
eGFR at biopsy0.00200.95 (0.93–0.98)
Proteinuria5.66×10−54.5 (2.4–8.7)
AKI signal7.33×10−74.6 (2.5–8.5)
i-IFTA0.330.99 (0.97–1.0)

In univariate analysis, all variables were associated with future graft loss; the best p-value was for the AKI signal (p = 6.66×10−16).

In multivariate analysis, the AKI signal was the strongest predictor (p = 7.33×10−7), followed by proteinuria and eGFR. Fibrosis (ci), inflammation in scarred areas (i-IFTA) and diagnosis of a potentially progressive disease were no longer significant. The interaction of disease and the AKI signal was not significant in any model.

Thus the correlation of potentially progressive diseases such as recurrent GN or ABMR with future failure is primarily because these diseases can induce parenchymal injury, manifested by the AKI signal. Areas of scarring with inflammation also reflect recent or ongoing parenchymal injury, but this is better reflected in the AKI signal.

Similarity of the AKI signal to the Risk Score

We compared the AKI signal with the previously defined molecular risk score for graft loss [19], either for all 315 patients (Figure 2A) or for 186 patients presenting for late biopsies (group 6, Figure 2B). The kidneys were divided into high or low AKI signal by the median; and into high and low risk scores by above or below zero. Kidneys with high AKI signals and those with high risk scores had similarly poor graft survival. The Cox regression p-values were significant for the AKI signal and risk score when all patients were analyzed (see Table 8). Thus the AKI signal discovered in early transplants with reversible injury predicted future deterioration when it was detected in late indication biopsies (Cox regression p-value 1.8 × 10−8).

image

Figure 2. Survival curves of kidney grafts based on the AKI signal and the molecular risk score using last biopsies. The analysis used the entire cohort of patients (A) and the subset of patients who presented for biopsy after 1 year post transplant (group 6 in table 1). RS–risk score. Survival was not significantly different between grafts with either AKI signal > median and RS >0 or AKI signal < median and RS < 0. Survival was significantly different between grafts with AKI signal > median versus AKI signal < median or RS > 0 versus RS < 0. F–failed, N–functioning.

Download figure to PowerPoint

Many individual IRRATs predict graft failure

Not only did the AKI signal predict progression, but many individual transcripts used in the AKI signal (IRRATs) also predicted progression. Thus 19/30 IRRATs had p-values < 0.05 in a Cox regression model for graft loss within 3 years after biopsy (Table 7). Fourteen transcripts in the AKI signal (IRRAT) had also been independently selected for use by the risk score classifier.

Table 7. Many individual transcripts constituting the AKI signal predict future graft failure and are shared with the molecular Risk Score classifier
Gene symbolGene nameCox regressionHazard ratio p-value95% C.I.Used by the risk score classifier
  1. Transcripts are ordered by their ascending Cox regression p values. Bolded symbols indicate significant Cox regression p-values.

  2. The molecular risk score was derived for all 403 biopsies from 315 patients. 95% C.I. = range of values in which the estimate prevalence will fall 95% of the time.

NNMTNicotinamide N-methyltransferase0.0001.81(1.44, 2.26)Yes
MEGF11MEGF11 protein0.0001.81(1.38, 2.37)Yes
VCANChondroitin sulfate proteoglycan 2 (versican)0.0001.49(1.21, 1.84)Yes
NFKBIZNuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, zeta0.0001.99(1.41, 2.82)Yes
LTFLactotransferrin0.0001.36(1.15, 1.60)Yes
ITGB6Integrin, beta 60.0021.37(1.13, 1.66)Yes
SLPISecretory leukocyte protease inhibitor (antileukoproteinase)0.0031.39(1.15, 1.68)Yes
LCN2Lipocalin 2 (oncogene 24p3)0.0031.39(1.14, 1.69)Yes
VMP1Vacuole membrane protein 10.0031.55(1.16, 2.07)Yes
RARRES1retinoic acid receptor responder (tazarotene induced) 10.0041.44(1.13, 1.85)Yes
CTSSCathepsin S0.0051.41(1.11, 1.80)Yes
SOD2superoxide dismutase 2, mitochondrial0.0061.57(1.16, 2.12) 
OSMROncostatin M receptor0.0061.61(1.15, 2.25) 
FCGR3AFc fragment of IgG, low affinity IIIa, receptor for (CD16)0.0071.30(1.08, 1.57)Yes
CDH6Cadherin 6, type 2, K-cadherin (fetal kidney)0.0081.59(1.13, 2.25) 
ITGB3Integrin, beta 3 (platelet glycoprotein IIIa, antigen CD61)0.0081.49(1.12, 1.99) 
AKAP12A kinase (PRKA) anchor protein (gravin) 120.0121.48(1.09, 2.01) 
PTPRCProtein tyrosine phosphatase, receptor type, C0.0191.30(1.04, 1.63)Yes
EVI2AEcotropic viral integration site 2A0.0321.30(1.02, 1.64)Yes
SERPINA3Serine (or cysteine) proteinase inhibitor, clade A, member 30.0561.23(1.00, 1.51) 
OLFM4Olfactomedin 40.0600.80(0.63, 1.02) 
ADAMTS1A disintegrin-like and metalloprotease with Thrombospondin type 1 motif, 10.0671.34(0.99, 1.83) 
EGR1Early growth response 10.1001.24(0.97, 1.59) 
ADAM9A disintegrin and metalloproteinase domain 9 (meltrin gamma)0.1311.31(0.91, 1.88) 
C9orf71Chromosome 9 open reading frame 710.1481.23(0.93, 1.62) 
FOSThe v-fos FBJ murine osteosarcoma viral oncogene homolog0.2341.17(0.91, 1.50) 
S100A8S100 calcium binding protein A8 (calgranulin A)0.4310.91(0.73, 1.15) 
PTX3Pentaxin-related gene, rapidly induced by IL-1 beta0.4561.11(0.86, 1.43) 
PI15Peptidase inhibitor 150.6861.07(0.78, 1.46) 
MTND6NADH dehydrogenase, subunit 6 (complex I)0.8371.04(0.78, 1.46) 

Thus many individual AKI transcripts identified in early AKI that recovered were strongly associated with the risk of graft failure when expressed in late kidney transplants, probably because they reflect a parenchymal injury-repair response induced by many types of injury and diseases.

Comparison of AKI signal to fibroblast collagens

Although progression of chronic kidney disease is associated with fibrosis, fibrillar collagen transcripts were not selected by the algorithm that created the AKI signal or by the algorithm that created the risk score. We studied the relationship of fibroblast collagen (COL signal, described in the Methods section) gene expression to the AKI signal, the risk score, inflammation in atrophy-fibrosis, scarring (ci) and future graft failure in the latest biopsies of all 315 patients (Table 8).

Table 8. Association between molecular features, scarring and risk of failure in all 315 patients
 Spearman correlation withRisk of failure (Hazard ratio [95% confidence limits]) and Cox regression p-value
Molecular variableInflammation in scarred areas (i-IFTA)Scarring (ci) 
  1. Only latest biopsy per patient was included in the analysis. Bolded numbers indicate significant Cox regression p-values.

Fibrillary collagen0.070.01.11 (0.92–1.33)
(COL) gene set signalp = 0.89p = 0.93p = 0.76
HAVCR1 (“KIM-1”) expression0.230.221.59 (1.32–1.90)
 p < 0.0001p < 0.0001p = 6.00×10−6
Risk score0.600.552.92 (2.17–3.92)
 p < 0.0001p < 0.0001p = 1.33×10−11
AKI signal0.310.282.18 (1.66–2.85)
 p < 0.0001p < 0.0001p = 7.02×10−8

The COL signal did not correlate with inflammation in atrophy-fibrosis (i-IFTA), scarring (ci), or graft loss (Table 8). The risk score and the AKI signal correlated with i-IFTA and ci as expected and predicted future graft loss. Biomarker HAVCR1 correlated with i-IFTA, ci and future graft loss but less strongly than the AKI signal and the risk score.

Discussion

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

We used microarrays to explore the significance of the AKI signal in indication biopsies from kidney transplants. The AKI signal consists of injury-repair transcripts originally defined in early kidney transplants with reversible acute injury without rejection or disease. This study found that many diseases triggering biopsies of kidney transplants induce the AKI signal: TCMR, ABMR, BK nephropathy, glomerulonephritis and others. The AKI signal was associated with impaired function, but its association with future progression was dependent on the triggering insult. In TCMR, which is reversible when treated, the signal was associated with future recovery of GFR, similar to its associations in early AKI. In biopsies with potentially progressive diseases (group 3) the AKI signal had more ominous association because the ineffectiveness of treatments for many diseases that trigger indication biopsies, e.g. recurrent GN and ABMR. Thus transplants presenting late with elevated AKI signals often progressed to failure. The main histological feature corresponding with the AKI signal was inflammation in areas of atrophy-fibrosis, but not scarring without inflammation. The AKI signal shared many transcripts with the risk score classifier, e.g. ITGB6, NNMT, VCAN, LCN2, but fibroblast collagen transcripts were not associated with progression. Thus the AKI signal indicates recent parenchymal wounding and attempted repair, but its prognostic significance reflects the nature of the underlying wounding process. If the injury is self-limited (early AKI) or is arrested by effective treatment (TCMR), the signal correlates with depressed function at the time of biopsy but the kidney recovers and does not progress to failure. If the injury reflects an untreatable and sustained insult, the AKI signal correlates with depressed function at the time of biopsy but also indicates probable progression to graft failure.

Two limitations to this study to be addressed in future studies are that it did not identify the underlying diseases in some late biopsies with high AKI signals, and that is was restricted to indication biopsies. The failure to identify the origin of the AKI signal in many troubled transplants in group 5 (i.e. no histologic rejection) probably reflects missed diagnoses due to the limitations of the histologic diagnostic system. For example, scarring in late biopsies can obscure diagnoses such as glomerulonephritis, ABMR, or TCMR. The problem with indications biopsies is that they cannot establish the frequency of AKI signals in stable transplants without biopsy indications. This is offset by the NOMOA biopsies in the population, which act as an internal control, but it would be still be interesting to know the frequency and significance of AKI signals in protocol biopsies many years after transplantation. However, the frequency of serious ongoing diseases in stable kidneys with no indications for biopsy is low, and the risk and expense of protocol biopsies is probably not justifiable.

The results require a change in our concept of progression in kidney transplants with diseases, namely that progression is due to injury caused by diseases, with secondary fibrosis, rather than dysregulated fibrosis. In multivariate analysis, inflammation in atrophy-fibrosis dropped out when we included the AKI signal, because both reflect the response to recent wounding but the AKI is a better reflection. In contrast, fibroblast collagens failed to predict progression, indicating that progression is not due to fibrosis but to ongoing parenchymal injury. The study of chronic kidney disease has long faced two interpretations of fibrogenesis: a reaction to parenchymal wounding by a disease versus a dysregulated process triggered by a disease but becoming self-sustaining [24]. Our data strongly favor the former idea, wherein persistent disease damages nephrons and induces wound healing (the AKI signal), with fibrosis and inflammation. Wound healing [25] is readily mistaken for an independent autonomous process when the diagnosis of primary diseases is missed, due to scarring, sampling error, or poor diagnostic criteria (e.g. ABMR) [21]. The autonomous dysregulated-repair model is suggested by rodent models [26], but such models manifest features not found in human kidney disease, e.g. a high propensity for progressive focal segmental glomerulosclerosis after nonspecific injuries.

The association of inflammation with injury-repair explains why inflammation in areas of scarring is associated with progression in kidney transplants [27, 28], because inflammation in scarring indicates recent injury. Inflammatory cells, particularly macrophages, are recruited to wounded tissue [28-32]. Inflammation is confusing because it is sometimes aggressive (as in TCMR) and sometimes reactive, as in the healing of a recent wound. In protocol biopsies of stable kidney transplants at 6 weeks, inflammation correlates with previous injury [33], and does not predict poor outcome.

Inflammation has a disproportionate impact on biopsy interpretation because it is readily detected by histology, whereas the parenchymal injury is often silent by histology, creating the misleading impression that the inflammation is causing progression. Acute epithelial injury is poorly assessed by conventional microscopy: the epithelial features that pathologists term acute tubular injury correlated poorly with the AKI signal or with function, both in this study and in kidneys with early acute injury. New assays for acute epithelial injury, such as confocal microscopy to assess loss of brush border solute carriers [34] or the AKI signal itself will change this, adding a clinically useful dimension to biopsy assessment by identifying the degree of recent or ongoing epithelial injury.

The fact that many insults—acute injury, TCMR, potentially progressive diseases—trigger a common injury-repair program makes the structure of this response of considerable interest. This program involves dedifferentiation and suspension of function, reexpression of developmental programs, stromal and microcirculation remodeling, and recruitment of macrophages and other inflammatory cells. Injury to one segment of the nephron can trigger a response of the entire unit through feedback mechanisms, e.g. the juxtaglomerular sensor of distal fluid contents [35]. Parenchymal injury from diseases that exceed the capacity to repair the nephrons results in permanent nephron shutdown and atrophy, requiring remodeling of the stroma and microcirculation [36]. The injury-repair program stabilizes injured tissue and attempts to restore its integrity, but also remodels the tissue when nephrons cannot be restored and must be permanently taken out of service.

The prominence of the AKI signal in kidneys with progressive diseases such as ABMR has encouraging implications for potential treatments. If progression is due to injury from the ongoing effects of the primary disease, progression will stop if the disease causing injury is arrested. If progression were an autonomous second process such as dysfunctional fibrosis (or FSGS, as in rodents), it would be necessary to treat both the primary disease and the secondary autonomous process. Thus the ability to identify aggressive processes [37, 38] becomes important for clinical progress [39, 40].

The association of protein biomarkers such as HAVCR1 (KIM-1) and LCN2 (NGAL) in body fluids with kidney injury and progression is probably because they reflect products of the parenchymal response to injury. The best empirically defined biomarkers are among the molecules most induced by parenchymal injury, again not related to fibroblast collagen synthesis. It will be important to identify biomarkers that quantitatively reflect the parenchymal injury-repair response and the AKI signal in the tissue, and to avoid biomarkers that are too readily induced by relatively trivial injury.

Acknowledgments

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

The ATAGC has been supported by the Canada Foundation for Innovation, Genome Canada, the University of Alberta, the University of Alberta Hospital Foundation, Alberta Advanced Education and Technology, Roche Molecular Systems, Hoffmann-La Roche Canada Ltd., the Alberta Ministry of Advanced Education and Technology, the Kidney Foundation of Canada, Stromedix Inc. and Astellas Canada. Dr. Halloran also holds a Canada Research Chair in Transplant Immunology and the Muttart Chair in Clinical Immunology. The authors thank Anna Hutton and Vido Ramassar for technical support and Zija Jacaj for collecting the clinical data. The authors thank Dr. Bruce Kaplan and Dr. Arthur Matas for biopsy material, and Michael Mengel and Banu Sis for their assessment of the histopathology.

Disclosure

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

P. F. Halloran holds shares in Transcriptome Sciences Inc., a company with an interest in molecular diagnostics. The other authors have no competing financial interests to disclose as described by the American Journal of Transplantation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  • 1
    Famulski KS, de Freitas DG, Kreepala C, et al. Molecular phenotypes of acute kidney injury in human kidney transplants. JASN 2012; 23: 948958.
  • 2
    Venkatachalam MA, Griffin KA, Lan RP, Geng H, Saikumar P, Bidani AK. Acute kidney injury: A springboard for progression in chronic kidney disease. Am J Physiol Renal Physiol 2010; 298: F1078F1094.
  • 3
    Sharfuddin AA, Molitoris BA. Pathophysiology of ischemic acute kidney injury. Nat Rev Nephrol 2011; 7: 189200.
  • 4
    Devarajan P. Update on mechanisms of ischemic acute kidney injury. JASN 2006; 17: 15031520.
  • 5
    Chang HY, Sneddon JB, Alizadeh AA, Sood R, West RB, Montgomery K, et al. Gene expression signature of fibroblast serum response predicts human cancer progression: Similarities between tumors and wounds. Plos Biology 2004; 2: 206214.
  • 6
    Riss J, Khanna C, Koo S, et al. Cancers as wounds that do not heal: Differences and similarities between renal regeneration/repair and renal cell carcinoma. Cancer Research 2006; 66: 72167224.
  • 7
    Bissell MJ, Radisky D. Putting tumours in context. Nat Rev Cancer 2001; 1: 4654.
  • 8
    Igarashi P, Shao X, McNally BT, Hiesberger T. Roles of HNF-1beta in kidney development and congenital cystic diseases. Kidney Int 2005; 68: 19441947.
  • 9
    Wight TN, Merrilees MJ. Proteoglycans in atherosclerosis and restenosis: Key roles for versican. Circ Res 2004; 94: 11581167.
  • 10
    Kinsella MG, Bressler SL, Wight TN. The regulated synthesis of versican, decorin, and biglycan: Extracellular matrix proteoglycans that influence cellular phenotype. Crit Rev Eukaryot Gene Expr 2004; 14: 203234.
  • 11
    Wu Y, Siadaty MS, Berens ME, Hampton GM, Theodorescu D. Overlapping gene expression profiles of cell migration and tumor invasion in human bladder cancer identify metallothionein 1E and nicotinamide N-methyltransferase as novel regulators of cell migration. Oncogene 2008; 27: 66796689.
  • 12
    Siew ED, Ware LB, Ikizler TA. Biological markers of acute kidney injury. J Am Soc Nephrol 2011; 22: 810820.
  • 13
    Anderson S, Eldadah B, Halter JB, et al. Acute kidney injury in older adults. J Am Soc Nephrol 2011; 22: 2838.
  • 14
    Ko GJ, Grigoryev DN, Linfert D, Jang HR, Watkins T, Cheadle C, et al. Transcriptional analysis of kidneys during repair from AKI reveals possible roles for NGAL and KIM-1 as biomarkers of AKI-to-CKD transition. Am J Physiol Renal Physiol 2010; 298: F1472F1483.
  • 15
    Famulski KS, Broderick G, Hay K, et al. Transcriptome analysis reveals heterogeneity in the injury response of kidney transplants. Am J Transplant 2007; 7: 24832495.
  • 16
    Sellares J, de Freitas DG, Mengel M, et al. Inflammation lesions in kidney transplant biopsies: association with survival is due to the underlying diseases. Am J Transplant 2011; 11: 489499.
  • 17
    Mengel M, Reeve J, Bunnag S, et al. Scoring total inflammation is superior to the current Banff inflammation score in predicting outcome and the degree of molecular disturbance in renal allografts. Am J Transplant 2009; 9: 18591867.
  • 18
    Famulski KS, Einecke G, Sis B, et al. Defining the canonical form of T-cell-mediated rejection in human kidney transplants. Am J Transplant 2010; 10: 810820.
  • 19
    Einecke G, Reeve J, Sis B, et al. A molecular classifier for predicting future graft loss in late kidney transplant biopsies. J Clin Invest 2010; 120: 18621872.
  • 20
    Mueller TF, Einecke G, Reeve J, et al. Microarray analysis of rejection in human kidney transplants using pathogenesis-based transcript sets. Am J Transplant 2007; 7: 27122722.
  • 21
    Einecke G, Sis B, Reeve J, et al. Antibody-mediated microcirculation injury is the major cause of late kidney transplant failure. Am J Transplant 2009; 9: 25202531.
  • 22
    Sellares J, De Freitas D, Mengel M, et al. Understanding the causes of kidney transplant failure: the dominant role of antibody-mediated rejection and non-adherence. Am J Transplant 2012; 12: 388399.
  • 23
    de Freitas DG, Sellares J, Mengel M, et al. The nature of biopsies with “Borderline Rejection” and prospects for eliminating this category. Am J Transplant 2012; 12: 191201.
  • 24
    Zeisberg M, Neilson EG. Mechanisms of tubulointerstitial fibrosis. J Am Soc Nephrol 2010; 21: 18191834.
  • 25
    Wight TN, Potter-Perigo S. The extracellular matrix: An active or passive player in fibrosis? Am J Physiol Gastrointest Liver Physiol 2011; 301: G950G955.
  • 26
    Eddy AA, Lopez-Guisa JM, Okamura DM, Yamaguchi I. Investigating mechanisms of chronic kidney disease in mouse models. Pediatr Nephrol 2012; 27: 12331247.
  • 27
    Cosio FG, Grande JP, Wadei H, Larson TS, Griffin MD, Stegall MD. Predicting subsequent decline in kidney allograft function from early surveillance biopsies. Am J Transplant 2005; 5: 24642472.
  • 28
    Park WD, Griffin MD, Cornell LD, Cosio FG, Stegall MD. Fibrosis with inflammation at one year predicts transplant functional decline. J Am Soc Nephrol 2010; 21: 19871997.
  • 29
    Wen X, Murugan R, Peng Z, Kellum JA. Pathophysiology of acute kidney injury: A new perspective. Contrib Nephrol 2010; 165: 3945.
  • 30
    Akcay A, Nguyen Q, Edelstein CL. Mediators of inflammation in acute kidney injury. Mediators Inflamm 2009; 2009: 137072.
  • 31
    Mannon RB, Matas AJ, Grande J, et al. Inflammation in areas of tubular atrophy in kidney allograft biopsies: A potent predictor of allograft failure. Am J Transplant 2010; 10: 20662073.
  • 32
    Anders HJ, Ryu M. Renal microenvironments and macrophage phenotypes determine progression or resolution of renal inflammation and fibrosis. Kidney Int 2011; 80: 915925.
  • 33
    Mengel M, Chang J, Kayser D, et al. The molecular phenotype of six-week protocol biopsies from human renal allografts: Reflections of prior injury but not future course. Am J Transplant 2011; 11: 708718.
  • 34
    Einecke G, Kayser D, Vanslambrouck JM, et al. Loss of solute carriers in T cell mediated rejection in mouse and human kidneys: An active epithelial injury-repair response. Am J Transplant 2010; 10: 22412251.
  • 35
    Schnermann J, Levine DZ. Paracrine factors in tubuloglomerular feedback: Adenosine, ATP, and nitric oxide. Annu Rev Physiol 2003; 65: 501529.
  • 36
    Bonventre JV, Yang L. Cellular pathophysiology of ischemic acute kidney injury. J Clin Invest 2011; 121: 42104221.
  • 37
    Racusen LC, Regele H. The pathology of chronic allograft dysfunction. Kidney Int Suppl 2010; 19: S27S32.
  • 38
    Lee SB, Kalluri R. Mechanistic connection between inflammation and fibrosis. Kidney Int Suppl 2010; 19: S22S26.
  • 39
    Akalin E, O'Connell PJ. Genomics of chronic allograft injury. Kidney Int Suppl 2010; 19: S33S37.
  • 40
    Mannon RB. Immune monitoring and biomarkers to predict chronic allograft dysfunction. Kidney Int Suppl 2010; 19: S59S65.