Because of increased interest in renal transplant protocol biopsies (1), a symposium was organized to assess their utility for patient management, clinical trial outcome analysis and elucidation of pathogenetic mechanisms. The participants were chosen for their scientific productivity in the field of renal transplantation related to protocol biopsies. No attempt was made to present a formal debate on the pros and cons of protocol biopsies. The program consisted of 11 presentations in three sections, as summarized below.
A 1-day symposium on the application of protocol biopsies in renal transplantation was held in Boston, 21 July 2006. Representatives from centers with extensive experience in the use of protocol biopsies for routine patient care and research reported results on the pathological findings and their value in patient management. The consensus was that protocol biopsies, in experienced hands, are a safe and valuable means of detecting subclinical disease that can benefit from modification of therapy. Furthermore, molecular studies reveal evidence of activity or progression not readily appreciated by histological techniques. Wider application is expected in multicenter clinical trials to predict and validate outcomes. The principal barrier to wider use of protocol biopsies is knowledge of the benefits of intervention.
Clinical and Pathological Data from Protocol Biopsy Programs
Jeremy Chapman presented data on behalf of Brian Nankivell and colleagues at Westmead Hospital, where protocol biopsies at 0, 3, and 12 months and at yearly intervals subsequently are the standard of care (2). Their hypothesis is that treatable, subclinical immune and nonimmune processes that lead to late graft loss begin early and can be detected by protocol biopsies (1–3 months). Chronic tubulo-interstitial and vascular changes, previously known as chronic allograft nephropathy (CAN), can be seen in one-third of transplants after 1 year and at later times become nearly universal. Biopsies at 3 months scored as Banff ci0 and cv0 have a significantly better graft survival at 5 years.
The prevalence of cellular subclinical rejection episodes depends on HLA-match, time posttransplant, and the immunosuppressive protocol, for example, tacrolimus and MMF therapy decreased the prevalence. Subclinical rejection in 3-month biopsies predicted interstitial fibrosis at 1 year; GFR 2 years after transplantation was lower in patients with repeated biopsies demonstrating subclinical rejection. Furthermore, treatment of subclinical rejection in early biopsies with corticosteroids alone often resulted in improved structure, function and graft survival as shown initially by Rush (3). However, this conclusion must be confirmed in larger studies, and in biopsies at late time points.
Chronic calcineurin inhibitor (CNI) toxicity, whose hallmark is arteriolar hyalinosis, is a universal finding after several years. Hyalinosis in any form was related to CNI dose, not glucose intolerance and was evident before hypertension. Therefore de novo arteriolar hyalinosis in allografts, even if not in the nodular pattern, can largely be attributed to CNI. Glomerulosclerosis increases from 5% at 1 year to 30% at 5 years after transplantation, associated with prior interstitial fibrosis and arteriolar hyalinosis. Substitution of MMF for azathioprine reduces fibrosis, glomerular scarring and mesangial expansion, arguing that CNI toxicity can be modified (Wavamunno et al., World Transplant Congress (WTC) Abstract 1296).
Microarray analysis of protocol biopsies showed that expression of fibrosis-related genes was already increased at 1 month, even in histologically normal biopsies; early analysis suggests that 1-month gene expression may predict overt fibrosis at 3 months; at 3 months the gene expression signature of epithelial-to-mesenchymal transition represented a minority of fibrosis associated genes in those with TA and IF (Vitalone et al., WTC Abstracts 844 and 1057).
Daniel Serón, in collaboration with Jean-Louis Bosmans (University of Antwerp, Belgium), presented their substantial experience with protocol biopsies as a surrogate marker of graft survival (4). A surrogate marker is a substitute variable that has a validated, quantitative ability to predict incremental changes in true endpoints (graft or patient survival are the gold standards).
The presence of chronic transplant arteriopathy, subclinical rejection or transplant glomerulopathy, superimposed on interstitial fibrosis and tubular atrophy (IF/TA) predicts a significantly poorer survival (1). Chronic lesions (Banff cg, ci, ct, cv) progress from 4 to 14 months after transplantation even with stable creatinine levels (5). Withdrawal of CNI at 3 months in patients receiving cyclosporine and sirolimus was not only associated with less severe IF/TA at 3 years, but also with better graft survival at 4 years, showing for the first time that a change in the severity of IF/TA is associated with a change in graft survival (6,7). However, these results may not apply to patients with high immunologic risk. In any case, these data suggest that protocol biopsies can meet certain criteria of a surrogate marker of graft survival. Prospective interventional studies have not been done to demonstrate the consequences of changing the biopsy marker, necessary for full validation as a surrogate marker.
Power calculations suggested that the minimum sample size to detect reduction of the prevalence of IF/TA at 6 months by 50% is approximately 300 patients per group (8). However, in trials comparing a CNI-based with a CNI-free regimen, ≤50 cases per group were sufficient to demonstrate a significant difference in IF/TA (9,10). Thus, the utility of protocol biopsies to differentiate between treatments may be better than expected. As the prevalence of IF/TA decreases, however, sample size will need to increase, unless more sensitive markers can be identified, such as molecular or morphometric assays.
Fernando Cosio presented an update of the Mayo Clinic results, where the standard of care includes protocol biopsies 0, 4, 12 and 24 months after transplantation, and more frequently in sensitized patients (11,12). In their large series of protocol biopsies, most commonly in recipients of living donor kidneys, approximately half had pathological findings at 1 year and IF/TA was the most common lesion. However, 1–2 year biopsies had other pathologies, including recurrent disease, subclinical rejection, polyomavirus nephritis and transplant glomerulopathy. Increasing IF/TA was related to poorer allograft function and inferior graft survival (11). Allografts with fibrosis and simultaneous inflammation (≥i1) had a worse survival than those with fibrosis without inflammation (11).
Transplant glomerulopathy increased in frequency from 1 to 5 years post-transplant (5–14% of protocol biopsies), frequently had a subclinical onset, and affected graft survival more adversely than IF/TA and inflammation (11). There was an association between transplant glomerulopathy, history of alloantibody responses and episodes of acute rejection, particularly antibody-mediated rejection episodes (12). At 1 year posttransplant 8% of conventional transplants had transplant glomerulopathy compared to 17% of ABO incompatible and 22% of positive crossmatch grafts. Transplant glomerulopathy was also associated with Hepatitis C virus, perhaps related increased frequency of anti-HLA antibodies.
Michael Mengel focused on the pathological phenotype of infiltrating cells in protocol biopsies. In the Banff schema, the diagnosis of acute cellular rejection is based in part on a quantitative increase of interstitial inflammation and tubulitis. Infiltrates below empirically defined thresholds for rejection are regarded as ‘borderline’ or not clinically relevant. Furthermore, ‘non-specific’ infiltrates, such as those in areas of interstitial fibrosis, around large blood vessels, in nodules or in subcapsular areas, do not qualify as criteria for active rejection (13).
Mengel analyzed more than 1100 protocol and indication biopsies (14,15) for the presence of the ‘nonspecific’ inflammatory infiltrates and characterized the phenotype of the mononuclear cells. The prevalence of infiltrates was equal in protocol and indication biopsies (87%) (15). Only 20% of the biopsies for graft dysfunction met the Banff criteria for rejection. A diffuse infiltrate was significantly more frequent in grafts with dysfunction and was associated with tubulitis, supporting the Banff criteria. However, ‘nonspecific’ infiltrates were also more common in biopsies that met the Banff criteria for acute cellular rejection. Nodular infiltrates were more common in protocol biopsies and were rich in B cells. Their data confirmed that serum creatinine is not a reliable marker for intragraft pathology. The best predictor of allograft function 1 year after transplantation was persistent inflammation, as measured by the number of infiltrates of any type. Their data suggest that no infiltrate in an allograft can be considered per se as harmless, and that a more comprehensive notation of infiltrates is desirable. Thus an emerging theme is the importance of inflammation, rather than simply fibrosis, as a predictor of outcome even when it failed to meet the criteria for rejection. Other studies indicate that the combination of inflammation and fibrosis has the worst outcome than either alone in biopsies taken 1 year or more after transplantation (1,11). In Mengel's studies, inflammation in areas of atrophic tubules was present in about half of the protocol biopsies. Analyzing patients with chronic tubulo-interstitial changes (i.e. at least Banff CAN grade I) in their 6-month protocol biopsies revealed that these patients had significantly more often inflammation in areas of tubular atrophy already in their prior protocol biopsy at 3 months after transplantation. Furthermore, patients with infiltrates in areas of tubular atrophy in their 6-month protocol biopsy had a significantly lower creatinine clearance at 2 years compared to those who lacked this type of inflammation.
Design and Analysis of Protocol Biopsy Studies
Michael Mihatsch presented the diagnostic criteria for CNI toxicity, gained from being a histological eyewitness from the dawn of cyclosporine therapy. He classifies CNI-arteriolopathy, the hallmark of chronic CNI toxicity, as a variant of thrombotic microangiopathy, with a slow, subclinical course. Differentiation from arteriolar hyalinosis of other causes (i.e. hypertension, diabetes and aging) is challenging. The major discriminating feature is that in CNI toxicity, smooth muscle cells are replaced by nodular hyaline deposits in the outer media (the typical lesion), while in arteriolar hyalinosis of other causes, smooth muscle cells are preserved and hyaline accumulates underneath the endothelium (16).
Mihatsch has introduced a new, more reproducible scoring system of CNI-arteriolopathy (17): grade 1 = CNI-arteriolopathy present in one arteriole, no circular involvement; grade 2 = CNI-arteriolopathy present in more than one arteriole, no circular involvement; and grade 3 = CNI-arteriolopathy with circular involvement independent of the number of arterioles involved. These scores significantly correlate with the time after transplantation, CNI-glomerulopathy, obsolescent glomeruli, focal segmental glomerulosclerosis, striped fibrosis with tubular atrophy and interstitial infiltrates without tubulitis. CNI-arteriolopathy did not correlate with features of acute rejection (C4d deposition, glomerulitis or Banff type II) and were negatively correlated with type I rejection. Randomized studies with long-term follow-up by protocol biopsies are needed to establish therapeutic implications of the various signs of CNI-toxicity.
Ian Roberts presented a critical review and proposal for reliable methods for scoring fibrosis in protocol biopsies. Estimation of fibrosis with any method is hampered by sampling error within small biopsies. Traditional approaches for estimating chronic damage involve semiquantitative assessment (eyeballing) of interstitial fibrosis, tubular atrophy, vascular intimal fibrosis, glomerulosclerosis and transplant glomerulopathy following the Banff or CADI scores. These semiquantitative scores are easy to apply but lack high interobserver reproducibility (18), contributing to their variable predictive and diagnostic value. Computer-based image analysis for collagens (e.g. by Sirius Red or immunohistochemistry) is an alternative, but the reproducibility can still be confounded by differences in tissue fixation, section thickness, microscope light levels, and arbitrarily set thresholds. Moreover, any method must eventually pass the test of delivering information of clinical value for diagnosis, staging and prognosis.
It is important to discriminate patients at risk of progressive fibrosis from those with established stable chronic damage, common in donor kidneys. Fibrotic activity may be assessed by expression of genes involved in pro- and antifibrotic pathways or the accumulation of cells typical of early fibrosis, such as myofibroblasts and mast cells (19). Evidence presented at this meeting supports the evaluation and grading of inflammatory activity in areas of fibrosis for assessing the activity of chronic allograft damage. This urgently needs further validation in prospective studies with protocol biopsies.
The safety of protocol biopsies was addressed by Anke Schwarz, who updated her published experience of 1171 protocol biopsies (20). At Hannover, 1705 protocol renal transplant biopsies over the last 6 years have been done as an outpatient procedure with 4-h surveillance. The complications were gross hematuria (3.1%), perirenal hematoma (3.3%), vasovagal reaction (0.8%), and A-V fistula with mostly spontaneous resolution (9.0%). Overall, 2.0% required hospitalization, 0.6% urinary catheter, 0.1% operative intervention and 0.3% blood transfusion. There were no deaths or graft losses, comparable to a multicenter study in which graft loss due to biopsy was 0.04% (1/2127) (21). In the Hannover series, all of the complications became evident in the first 4 h after the biopsy. The initially used 18-gauge needle was substituted for a 16-gauge needle, which resulted in an increase of specimen adequacy from 40% to 72% (7 glomeruli, 1 artery), with no adverse consequences other than a slight increase in hematomas (2.0–4.4%).
In the discussion that ensued, the suggestion was made that the centers with a large experience collate and publish their safety and complication rates, to provide the best guide for clinicians and investigators contemplating a protocol biopsy program. Chapman pointed out that the rate of graft loss from protocol biopsies of about 0.03% has to be weighed against the annual graft loss rate of about 5% due to rejection and the potential benefit of early intervention guided by the biopsy findings.
Marc Cavaillé-Coll provided a perspective on the role of protocol biopsies in multicenter trials. Experience with protocol biopsies is still limited at the FDA, but evaluation of the use of biopsies in clinical trials should be encouraged. Their value depends on the rigor of study design, conduct and compliance. Bias is minimized by randomization, masking of treatment groups, compliance with protocol, complete follow-up, standardized criteria for grading of allograft damage and centralized biopsy review. To be of value, protocol biopsies should provide results predictive of long-term graft function and survival.
Accelerated approval of new drugs that treat serious or life-threatening illnesses can be based on a surrogate endpoint, other than survival or irreversible injury, defined as one that is ‘reasonably likely, based on epidemiologic, therapeutic, pathophysiologic or other evidence, to predict an effect on a clinical end-point other than survival or irreversible morbidity’ (21CFR§ 314.510). These must be validated, and if used, require that the ‘applicant study the drug further, to verify and describe its clinical benefit, where there is uncertainty as to the relationship of the surrogate endpoint to clinical benefit, or the observed clinical benefit to ultimate outcome’.
In the discussion, Halloran emphasized the problem created in the interpretation of data from trials in which not all patients consent to a protocol biopsy, since these patients may not be a random sample of the larger population in the trial. Statistically robust ways of dealing with this problem are sorely needed. Cavaillé-Coll suggested that methods to impute values for missing data be defined prospectively. Several suggested a central registry of standardized data from multicenter clinical trials would be extremely valuable to provide validation of protocol biopsy results (as well as clinical data) with long-term outcome.
Molecular Markers in Protocol Biopsies: Insights into Pathogenesis and Applications
The promise and challenge of microarray analysis of gene expression in biopsies was the focus of the presentation by Philip Halloran. The comprehensive approach his group has taken is to identify the transcriptome events and functional families of genes in readily manipulated animal models (renal transplantation in the mouse) and apply these to the human.
Among the pathogenesis transcript sets' (PBT) defined in the mouse are cytotoxic T cell associated transcripts (CATs), interferon-γ dependent rejection induced transcripts (GRITs), macrophage activation transcripts (MATs), and solute carrier epithelial transcripts (Slcs). These provide a biologically relevant and potentially incisive approach to increase the signal-to-noise ratio in broadband microarray data. In the mouse CATs appear before tubulitis is obvious and are independent of CD103, perforin or granzyme A or B (22). Furthermore, epithelial injury, as judged by loss of epithelial specific transcripts, appears long before morphologic changes. Their group has proposed that T-cell-mediated rejection and the consequent epithelial injury is not due to tubulitis, but rather tubulitis is a consequence of injury.
Initial application in clinical biopsies is underway. Among 88 renal biopsies taken for graft dysfunction there was a high degree of correlation between the CATs, GRITS and a negative correlation with Slcs (Einecke et al., WTC Abstract). PBTs correlated with Banff scores for t, g, ci, ct and i but not with endarteritis (Banff v) or C4d. Tubulitis, glomerulitis and interstitial mononuclear inflammation were predictors of CATs and GRITs. Tubulitis was the only independent predictor of decreased epithelial transcripts (Slcs). PBTs correlated better with the retrospective diagnosis of rejection by the clinical course than by the Banff histopathology classification, among 54 biopsies with a t score of ≥1 (Mueller et al., WTC Abstract). Tubulitis sometimes occurred without elevation of CATs, GRITs or loss of Slcs and therefore may not be pathogenic. PBTs offer an objective new approach that can extend histopathology and provide new insights into the mechanisms of graft rejection. Their application to protocol biopsies will be of considerable interest.
Allan Kirk presented an update of the extensive gene expression analysis done by his group on protocol and indication biopsies addressing the role of molecular analysis in detecting rejection and tolerance (23). Protocols at the NIH typically involve biopsies taken at the time of transplant (pre- and or post-reperfusion), one and 6 months later and at 1, 3 and 5 years post-transplant (23). Gene expression analyses have evolved to a custom 98 gene cDNA array in 384 well microfluidic format that takes 4 h to perform (24).
The question of whether molecular analysis can reliably tell us things that histology cannot can only be assessed using protocol biopsies, since common findings can factitiously associate with adverse conditions in biopsies taken for graft dysfunction, and prognostic findings can only be assessed before the development of the condition. An example of increased sensitivity of molecular assays is provided by postreperfusion biopsies, which have high levels of adhesion molecule, chemokine and cytokine transcripts, despite minimal histologic abnormalities (23).
Protocol biopsies taken after 1 month exhibiting either normal histology or subclinical rejection (SCR) were compared with biopsies for cause that had acute clinical rejection (ACR). Many genes were upregulated to a similar degree in subclinical and clinical rejection, indicating that these transcripts were manifestations of the infiltrate, though not indicators of functionally significant rejection. T-bet, FasL and CD152 (CTLA-4) transcripts, however, distinguished between clinical and subclinical rejection (23). The subclinical group was heterogeneous for expression levels of genes (e.g. CD152, CD154, CD28, TGF-beta), suggesting a divergent mechanisms warranting further investigation. Tubulitis (t3 vs. t0) correlated well with IFN-γ inducible chemokines (CXCL9, 10 and 11), some of which are produced by tubular epithelial cells in culture in response to IFN-γ (Kirk et al., WTC Abstract). Polyomavirus nephritis has a similar molecular signature as acute rejection, indicating the need for specific viral markers (Mannon et al., WTC Abstract 2883).
Molecular signals were predictive of later pathology. High levels of many genes were expressed in histologically normal biopsies prior to the detection of tubulitis (Hale et al., WTC Abstract). Six-month protocol biopsies taken before CAN developed showed increased levels of epithelial-mesenchymal transition markers alpha-smooth muscle actin, S100A4 and vimentin, as well as growth factors FGF2, TGFβ and PDGF (Cheng et al., WTC Abstract). These data support the hypothesis that early events in fibrosis are better detected by molecular signals than routine histology.
Hermann Haller focused on endothelial injury in allografts. Graft microvascular loss is associated with progressive IF/TA and graft dysfunction (25). Circulating endothelial progenitors correlate inversely with graft function as they do in patients with chronic renal failure (26). We need improved methods to evaluate the status of the endothelium to support and interpret the potential benefit of therapies directed at endothelial repopulation.
The debate is not whether protocol biopsies reveal subclinical important diseases and mechanisms, but whether their safety justifies the gain. Clinical studies from many centers involving thousands of patients have confirmed that the protocol biopsy in the first 6 months contains information that correlates with later graft outcome and these and later biopsies can reveal unsuspected pathological processes (subclinical rejection, polyomavirus infection, drug toxicity). New molecular tests have the power to enhance the prognostic and diagnostic yield. A critical need is for interventional studies to validate the biopsy as a surrogate marker.
Among the ‘verbs’ from this conference are recommendations to:
- 1establish a registry of clinical trials involving protocol biopsies, to help validate surrogate markers;
- 2refine, by consensus, pathologic scoring for protocol biopsies, such the different patterns of infiltrate and fibrosis, to expand the range of lesions scored by the Banff system;
- 3pool and publish safety data on protocol biopsies;
- 4develop robust statistical methods for the confounding effects of variable protocol biopsy compliance.
A satellite symposium held on 21 July 2006 under the auspices of the World Transplant Congress, supported in part by unrestricted educational grants from Genzyme Corporation, Novartis Pharmaceuticals and Roche Pharmaceuticals and organized by R. B. Colvin, M. Mengel, H. Haller and M. Stegall.