A Need for Biomarkers of Operational Tolerance in Liver and Kidney Transplantation

Authors

  • M.-C. Londoño,

    1. Liver Transplant Unit, Hospital Clinic, IDIBAPS, CIBEREHD, Barcelona, Spain
    Search for more papers by this author
  • R. Danger,

    1. Institut National de la Sante Et de la Recherche Medicale INSERM U643, and Institut de Transplantation Urologie Néphrologie du Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    Search for more papers by this author
  • M. Giral,

    1. Institut National de la Sante Et de la Recherche Medicale INSERM U643, and Institut de Transplantation Urologie Néphrologie du Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    2. Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    Search for more papers by this author
  • J.-P. Soulillou,

    1. Institut National de la Sante Et de la Recherche Medicale INSERM U643, and Institut de Transplantation Urologie Néphrologie du Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    2. Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    Search for more papers by this author
  • A. Sánchez-Fueyo,

    Corresponding author
    1. Liver Transplant Unit, Hospital Clinic, IDIBAPS, CIBEREHD, Barcelona, Spain
    2. Institute of Liver Studies, Cheyne Wing, King's College London, Denmark Hill, London
    Search for more papers by this author
  • S. Brouard

    Corresponding author
    1. Institut National de la Sante Et de la Recherche Medicale INSERM U643, and Institut de Transplantation Urologie Néphrologie du Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    2. Centre Hospitalier Universitaire Hôtel Dieu, Nantes, France
    Search for more papers by this author

Sophie Brouard, Sophie.Brouard@univ-nantes.fr and Alberto Sánchez-Fueyo, afueyo@clinic.ub.es

Senior authors equally contributed.

Abstract

Both kidney and particularly liver recipients can occasionally discontinue all immunosuppressive drugs without undergoing rejection. These patients, who maintain stable graft function off immunosuppressive drugs without clinically significant detrimental immune responses and/or immune deficits, are conventionally termed operationally tolerant and offer a unique paradigm of tolerance in humans. The immune characterization of operationally tolerant transplant recipients has recently received substantial attention. Operationally tolerant patients might exhibit a signature of tolerance that could potentially be useful to select recipients amenable to drug minimization or withdrawal. Furthermore, elucidation of the molecular pathways associated with the operational tolerance phenotype could provide novel targets for therapy. Particular emphasis has been placed on the use of blood samples and high-throughput transcriptional profiling techniques. In liver transplantation, natural killer related transcripts seem to be the most robust markers of operational tolerance, whereas enrichment in B cell-related gene expression markers has been consistently found in blood samples from operationally tolerant kidney recipients, suggesting that different mechanisms operate in the two situations. In this minireview, we summarize the main achievements of recently published reports focused on the identification of transcriptional markers of operational tolerance, we highlight their mechanistic and clinical implications and describe their methodological limitations.

Abbreviations: 
AKT

v-akt murine thymoma viral oncogene homolog 1

BANK1

B-cell scaffold protein with ankyrin repeats 1

FcR

Receptor to Fc fragment of IgG

HC

healthy controls

IL10

interleukin 10

ITN

Immune Tolerance Network

mRNA

messenger RNA

NK

natural killer

PBMC

peripheral blood mononuclear cell

qPCR

quantitative real-time PCR

TGFβ

transforming growth factor, beta 1

TOL

operationally tolerant patients

Introduction

Transplant recipients still exhibit much higher morbidity and mortality than the general population. Although this is in part due to the effects of chronic allograft injury, the main causes are comorbidities influenced by chronic immunosuppressive drug usage (1–5). To redress this situation, research priorities in organ transplantation are moving away from the search of novel powerful immunosuppressive drugs toward the identification of strategies to minimize immunosuppression. The most extreme approach at long-term immunosuppression minimization is the complete withdrawal of immunosuppressive drugs. This has been occasionally described in kidney recipients but seems to be particularly frequent in liver transplantation, where approximately 20% of selected recipients enrolled in drug withdrawal trials are able to discontinue all drugs (6,7). These patients, who maintain stable graft function off immunosuppressive drugs without clinically significant detrimental immune responses and/or immune deficits, are conventionally termed operationally tolerant. The exact mechanisms contributing to this phenomenon (ignorance, anergy, deletion and regulation) are still unknown. Furthermore, critical aspects of the clinical phenotype, such as its stability, prevalence and histopathology correlates, still need to be unambiguously established. Despite these limitations, operationally tolerant recipients constitute the best proof-of-principle of the viability of drug-free graft survival in clinical organ transplantation and have consequently received substantial attention by the transplant community. Recently, there have been several internationally orchestrated collaborative efforts identify biomarkers of operational tolerance in liver and kidney transplantation. Although access to the tolerated organs would present several advantages to explore the underlying mechanisms of operational tolerance, until now this has only been performed in liver recipients. In kidney transplantation most recipients’ doctors and ethical committees are reluctant to and do not recommend biopsy on these functioning grafts. Most studies have concentrated on the use of peripheral blood samples and transcriptional profiling techniques in the ultimate hope to offer clinicians a less invasive predictive and monitoring tool of tolerance or low immunological risk in human (Table 1). In this review, we concentrate on recent advances made in the development of transcriptional biomarkers of human operational tolerance. We summarize the main achievements of reported studies, highlight their mechanistic and clinical implications and point out their conceptual and methodological caveats.

Table 1.  Studies using gene expression profiling to identify transcriptional biomarkers of tolerance
StudyRef.Patients (n)OrganSampleMicroarray platformValidation in additional transcriptional platform
  1. TOL = operationally tolerant patients; Non-TOL = non-operationally tolerant patients; CR = kidney recipients with chronic rejection; MS = kidney recipients on monotherapy with steroids; STA = stable kidney recipients on standard immunosuppressive treatment; AR = kidney recipients with biopsy-proven acute rejection; HC = healthy controls; PBMC = peripheral blood mononuclear cell.

Martínez-Llordella (2007)(8)16 TOL 16 Non-TOL 16 HCLiverPBMCAffimetrix whole-genomeqPCR
Kawasaki (2007)(10)11 TOL 11 HCLiverPBMCAgilent cDNAqPCR
Brouard (2007)(13)17 TOL 22 CR 10 MS 12 STA 14 AR 16 HCKidneyPBMCCustom cDNA LymphochipqPCR
Braud (2008) Sivozhelezov (2008)(14, 29)8 TOL 18 CR 8 HCKidneyPBMCGlass-slide spotted custom array 
Pons (2008)(11)5 TOL 7 Non-TOLLiverPBMC Only qPCR
Martínez-Llordella (2008)(9)28 TOL 33 Non-TOL 16 HCLiverPBMCAffimetrix whole-genomeqPCR
Sagoo (2010)(18)11 TOL 11 MS 40 STA 10 CR 19 HCKidneyWhole bloodRISET 2.0 Agilent custom microarrayqPCR
Newell (2010)(17)25 TOL 33 STA 42 HCKidneyWhole blood urineAffimetrix whole genomeMassARRAY qPCR
Lozano (2011)(30)12 Liver TOL 12 Liver Non-TOL 12 Kidney TOL 12 Kidney CR 12 HCLiver KidneyPBMCAffimetrix whole-genomeqPCR

Identification of Transcriptional Biomarkers of Operational Tolerance

A “biomarker” is defined by the Biomarkers Definitions Working Group as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes or pharmacological responses to a therapeutic intervention” (8). A biomarker that can be employed as a substitute for a clinical endpoint is known as a “surrogate endpoint”. Characterization of a biomarker as a surrogate endpoint requires demonstration within clinical trials and/or epidemiological studies that the biomarker is capable of predicting clinical benefit with accuracy and reproducibility (8). Among the many immunomonitoring tools currently available to develop diagnostic and/or prognostic biomarkers in human transplantation, most studies focused on the study of operationally tolerant patients have favored the use of unbiased and comprehensive analytical platforms such as microarray-based transcriptional profiling. The specific goals of these studies have been: (1) to identify the genes differentially expressed between operationally tolerant recipients and one or more comparison groups (“class comparison” studies); and (2) to design classification rules capable of discriminating between operationally tolerant and nontolerant recipients on the basis of their expression profiles (“class prediction” studies). Gene expression microarrays are the most accessible, affordable and well standardized of all high-throughput “-omics” molecular profiling technologies. However, the use of this technology in clinical studies involving small sample sizes is statistically challenging, particularly in class prediction studies. Indeed, when there is a large disproportion between the high number of genes analyzed and the limited number of samples available, there are always one or more sets of genes capable of accurately classifying the samples from which the gene models were derived even when data are completely random. To address this problem a gene model has to be validated on a set of data completely independent from the data employed for its development. To do so, the study cases need to be partitioned into training and validation groups, and this can be accomplished using a variety of different approaches (e.g. the split-sample method or the various cross-validation resampling methods (9). However, these strategies are often applied in invalid ways leading to grossly biased results (10). Furthermore, even when appropriately applied, these techniques only partially tackle the problem if there are no large cohorts of patients available for study. Indeed, the experience accumulated in cancer research, where much larger microarray studies have been performed, indicate that at least 200 samples are required to generate accurate and stable predictive signatures (11). In human transplantation tolerance studies such sample sizes are very difficult to achieve. Microarray platforms are also very sensitive to experimental “noise” (i.e. inter- or intraplatform variability that arises when hybridizations are made in different laboratories or in the same laboratory but on different batches; Ref. 12). For this reason, technical validation of microarray expression results on a different transcriptional platform such as quantitative real-time PCR (qPCR), with standardization protocols, efficiency and RNA quality control, is still highly recommended (13,14).

Operational Tolerance in Kidney Transplantation

Brouard et al. (15) performed the first molecular profiling analysis of operationally tolerant recipients. The study included 17 operationally tolerant recipients (i.e. stable graft function without immunosuppression for at least 2 years), 22 chronic rejectors, 10 patients on corticosteroid monotherapy, 12 stable patients on standard immunosuppressive treatment and 14 recipients with acute rejection. A custom immunology-focused cDNA microarray (“Lymphochip”) was employed to analyze peripheral blood mononuclear cell (PMBC) samples from a training group of 5 operationally tolerant recipients, 11 chronic rejectors and 8 age-matched healthy subjects. A set of 49 genes able to differentiate operationally tolerant patients from healthy subjects and chronic rejectors was identified. Thirty-three out of the 49 genes were validated by qPCR and could segregate the operationally tolerant from the chronic rejector patients with 99% specificity and 86% sensitivity. This tolerance footprint classified 40% of long-term minimally immunosuppressed patients receiving corticosteroid monotherapy as potentially tolerant, whereas this occurred in only 8% of stable recipients on conventional maintenance bitherapy immunosuppression. The tolerance-related expression pattern was enriched in T cell-related genes and in genes involved in cell-cycle regulation. Furthermore, 27% of the differentially expressed genes were identified as potentially modulated by TGFβ (15). In parallel, in an independent nonstatistical analysis on the same samples, B-cell scaffold protein with ankyrin repeats 1 (BANK1), a negative modulator of CD40-mediated v-akt murine thymoma viral oncogene homolog 1 (AKT) activation involved in B-cell responses, seems as ‘“key leader” within the tolerance-associated gene expression network (16,17). The same set of samples was re-analyzed employing a different microarray platform (16). Seven B-cell-related pathways involving genes of cell cycle, proliferation, maturation and differentiation of B cells were identified in this analysis, with confirmation of an upregulation of the transcripts of BANK1 in blood from operationally tolerant recipients (18). A detailed description of the characteristics of circulating B cells found in this cohort of operationally tolerant kidney recipients was subsequently reported (19). Operationally tolerant kidney recipients exhibited an increased frequency and absolute number of circulating B cells with a memory phenotype and B cells expressing higher level of molecules of costimulation, activation and homing such as CD80, CD86, CD40, CD62L (19). In parallel, B cells from operationally tolerant recipients display a potentially inhibitory profile characterized by a decreased mRNA FcγRIIa/FcγRIIb (CD32a/CD32b) ratio mainly due to an increased level of transcripts of the inhibitory FcγRIIb (CD32b) receptor on B cells from these patients, as well as an increased number of B cells expressing the CD5 and CD1d, two molecules that have been associated to the potentially regulatory B10 phenotype (19).

A significant contribution of circulating B cells to the blood gene expression profiles noted in operationally tolerant kidney recipients was also described by the investigators of two large multicenter studies (20,21). Newell et al. reported the results of the Immune Tolerance Network (ITN) US multicenter study, in which 25 operationally tolerant kidney patients (i.e. stable graft function without immunosuppression for at least 1 year) and 33 stable kidney recipients on triple immunosuppression were analyzed. Whole blood Affymetrix microarray experiments were performed on samples collected from a training set of 19 operationally tolerant recipients, 27 immunosuppressed recipients and 12 healthy controls. Twenty-two out of the 30 genes differentially expressed less than twofold when comparing tolerant and immunosuppressed recipients were B-cell specific. Confirmatory transcriptional experiments on 228 genes were conducted employing the “Sequenom MassARRAY” platform on whole blood samples from all participants. Thirty-one genes were differentially expressed between tolerant and immunosuppressed recipients. No differences between operationally tolerant patients and healthy controls were noted. A model comprising three genes discriminated tolerant from immunosuppressed recipients with a sensitivity/specificity of 79%/86% in the training set and 100%/83% in an independent test set (six operationally tolerant patients and six patients under immunosuppression). As compared with immunosuppressed recipients, operationally tolerant recipients exhibited a higher number of naïve, memory, transitional and total circulating B cells. In contrast, healthy individuals and tolerant patients only differed in the number of naïve and IL10-producing transitional circulating B cells but interestingly, the authors report that ex vivo stimulated transitional B cells from tolerant patients produced increased amounts of IL-10, a hallmark of regulatory B cells (20). A more clear evidence of the association between B cells and operational tolerance was provided by the demonstration that CD20 transcript levels were increased in blood (15) and urine (20) from operationally tolerant recipients as compared with both immunosuppressed patients and healthy individuals. The Indices of Tolerance/RISET Consortium European multicenter study (21) evaluated whole-blood samples from 11 operationally tolerant patients, 11 stable recipients under corticosteroid monotherapy, 10 patients on immunosuppressive treatment without calcineurin inhibitors, 30 patients under calcineurin inhibitor-based treatment, 9 patients with chronic rejection, and 19 healthy controls. In addition to these study groups, samples from 89 kidney recipients from the ITN study (including 24 tolerant recipients) were also analyzed. Operationally tolerant patients exhibited a higher number of circulating B and natural killer (NK) cells compared to healthy controls and all other study groups. The expansion in the number of B cells was mainly due to an increase in the number of transitional and naïve B-cell subsets. Gene expression profiling was performed using an Agilent custom microarray platform (RISET 2.0 comprising 5069 transplantation-related probes). The two cohorts of operationally tolerant patients expressed a unique set of genes not shared by the remaining comparison groups of kidney grafted recipients and enriched in B-cell-related pathways. A signature comprising the 10 top ranked genes could discriminate tolerant recipients from the remaining study groups with 100% sensitivity and 100% specificity (European cohort), and with 90% sensitivity and 92% specificity (US cohort). Of note, a truly independent validation cohort was not included in the microarray analyses and healthy individuals were excluded from the classification analyses.

Finally, in an attempt to estimate the potential prevalence of operational tolerance among immunosuppressed kidney recipients, Brouard et al. employed the 20 most informative genes among their previously described 49 qPCR-based gene set (15) to classify a selected cohort of 111 immunosuppressed kidney recipients with stable renal function 5 years after transplantation (22). Only 4.6% of them were identified as potentially tolerant. This goes along with the rarity of cases of tolerant kidney recipients reported in the literature (7).

Operational Tolerance in Liver Transplantation

In the first study employing microarray gene expression profiling to characterize operationally tolerant liver transplant patients, Martínez-Llordella et al. (23) used whole-genome Affymetrix microarrays to analyze peripheral blood mononuclear cell (PBMC) samples from 16 operationally tolerant liver transplant recipients and 16 nontolerant recipients (patients in whom the failure of a previous attempt at drug weaning had formally demonstrated the absence of tolerance). Tolerant recipients exhibited 462 positively and 166 negatively regulated genes. The 462 upregulated gene set was significantly enriched in genes preferentially expressed by gamma-delta T cells (γδ T cell) and NK cells, and in genes encoding for proteins involved in cycle cell arrest. In contrast, upregulation of pro-inflammatory genes (e.g. TNFα, IL-1, IL-6, ICAM-1) was noted in nontolerant recipients, although this was restricted to nontolerant recipients with active hepatitis C virus infection. Operationally tolerant and nontolerant recipients also differed in the distribution of circulating γδ T cell subsets and in the frequency of potentially regulatory CD4+CD25+FoxP3+ T cells (Tregs). This cell subset was increased in tolerant patients, albeit no differences in FoxP3+ transcript levels between tolerant and nontolerant recipients were noted (23). Martínez-Llordella et al. subsequently expanded their initial observations on a larger cohort of patients (28 operationally tolerant and 33 nontolerant recipients). Whole-genome Affymetrix microarray experiments conducted on PBMC samples from 17 operationally tolerant and 21 nontolerant recipients revealed 1932 differentially expressed genes. Quantitative real-time PCR validation experiments were conducted on 68 genes and a very good correlation between the two transcriptional platforms was observed. Three predictive qPCR-derived gene expression signatures (containing 2–7 genes) were identified. These three signatures showed very high classification accuracy not only in the training set of 38 recipients from whom they were derived, but also in an independent validation cohort of 11 operationally tolerant and 12 nontolerant recipients. The tolerance-related expression data set was enriched in NK and γδ T cell-related transcripts, and significantly correlated with the frequency of circulating NK and γδ (but not CD4+CD25+FoxP3+) cells (24). Significant transcriptional differences were also observed in magnetically sorted PBMC subsets from tolerant and nontolerant recipients. The authors concluded that tolerance-related differential expression profiles were attributed both to differences in the distribution of PBMC subsets and to differences in the specific transcriptional programs of individual PBMC subsets.

In an independent microarray study from Japan, Kawasaki et al. employed an Agilent cDNA microarray platform comprising 12814 genes to analyze PBMC samples collected from 11 operationally tolerant live donor liver transplant recipients and 11 healthy individuals. This resulted in the identification of 717 differentially expressed genes (627 upregulated and 90 downregulated in operationally tolerant recipients; ten of them validated by qPCR). Among differentially expressed genes “immune response” was the most highly overrepresented functional category (25).

In addition to microarray studies, other groups have reported on the expression levels of individual genes. Thus, Pons et al. quantified FoxP3 expression in PBMC samples sequentially collected from 12 liver transplant patients on maintenance immunosuppression with cyclosporine A undergoing complete drug withdrawal. Before the initiation of weaning no differences in CD4+CD25high T cells or FoxP3 transcript levels were noted between recipients who subsequently rejected and those who successfully discontinued treatment. As weaning progressed, however, both CD4+CD25high T cells and FoxP3 gene expression gradually increased in the group of operationally tolerant recipients, but not in those who developed rejection (26). Li et al. evaluated whether the presence of intragraft Tregs correlated with drug-free graft acceptance. For that purpose, they gathered liver tissue samples from 28 operationally tolerant recipients, 29 patients under maintenance immunosuppression, 7 patients with chronic rejection and 12 patients with normal liver, and quantified FoxP3 expression levels by qPCR and CD4+FoxP3+ T cells by immunostaining. Operationally tolerant patients exhibited a higher number of CD4+FoxP3+ T cells compared to patients with chronic rejection and normal livers. In contrast, FoxP3 mRNA levels were similar in operationally tolerant and chronic rejectors (27).

Böhne et al. recently reported the results of the first prospective immunosuppression withdrawal trial in liver transplant recipients including blood and liver tissue transcriptional biomarker studies (28). In this study, 75 recipients out of the 98 liver recipients completed the trial, 57 of whom rejected whereas 41 were successfully weaned. Sequential blood and/or liver tissue samples from 75 recipients were analyzed employing whole-genome microarrays and qPCR. The enrichment in PBMC NK-related transcripts described by Martínez-Llordella et al. (23,24) was already present at enrolment (i.e. before the discontinuation of immunosuppressive drugs was initiated), thus confirming the results of previous retrospective studies. In contrast, no differences in Foxp3 mRNA levels or CD4+FoxP3+ T cells were noted between tolerant and nontolerant individuals at enrolment. Transcriptional differences between tolerant and nontolerant recipients were also identified at the graft itself. The intragraft gene expression profile, however, was mainly enriched in genes involved in iron homeostasis, and showed no overlap with the PBMC-derived expression markers. These markers were found to be independent from the type of immunosuppression and all other clinical parameters associated with the success of drug withdrawal. In agreement with the PBMC results, at enrolment no differences between tolerant and nontolerant patients in intragraft CD4+Foxp3+ T cells or Foxp3 mRNA levels were detected. Importantly, in a side-by-side comparison liver tissue-derived transcriptional signatures were found to be more accurate and reproducible at predicting the success of drug withdrawal than PBMC-derived signatures (28).

Altogether, these studies suggest that operationally tolerant liver recipients have unique blood and liver tissue gene expression patterns that could be useful to discriminate between tolerant and nontolerant recipients. NK cells and related transcripts seem to be the most robust blood markers of operational tolerance, as they are already present in the blood of tolerant liver recipients before immunosuppressive drugs are discontinued. The role of potentially regulatory CD4+CD25+Foxp3+ T cells is less clear due to the confounding effects of pharmacological immunosuppression. Importantly, blood cellular and transcriptional markers do not seem to reflect the nature of intragraft immune responses. Furthermore, liver tissue-derived biomarkers are more accurate than blood-related markers at predicting the success of drug withdrawal strategies.

Blood Transcriptional Similarities Between Operationally Tolerant Liver and Kidney Recipients

To find a common immunologic ground shared by recipients operationally tolerant to different organs, Lozano et al. directly compared on the same microarray platform blood samples collected from operationally tolerant and nontolerant liver and kidney recipients (30). As described by previous studies, liver and kidney operationally tolerant recipients exhibited distinct blood transcriptional and cell phenotypic patterns, which were related to NK and B cells, respectively. However, no or minimal overlaps in blood cell phenotype and whole-genome expression patterns were observed between the two groups of tolerant recipients. In the same report, Lozano et al. performed the first cross-platform comparison of tolerance-related expression data sets by re-analyzing their own expression data on tolerant kidney recipients side-by-side with the microarray data contained in Newell et al. and Sagoo et al. reports (20,21,30). Despite the heterogeneity of the study cohorts, 35 genes, most of them B-cell-related, were differentially expressed in the three operationally tolerant cohorts when compared with nontolerant transplant recipients. Except for a very limited number of genes, however, the overrepresentation of B-cell-related transcripts was not seen when kidney operationally tolerant patients were compared with healthy individuals. This report has important implications. First, it suggests that the mechanisms responsible for development of operational tolerance in kidney and liver recipients might be different. Second, it confirms that the previously described over-representation of B-cell-related transcripts in the blood of tolerant kidney recipients is highly reproducible across all cohorts analyzed. Third, the differences noted between liver and kidney tolerant recipients suggest that tolerance-related transcriptional signatures cannot be solely attributed to the presence or absence of pharmacological immunosuppression (30). It is enticing to speculate that the expansion of B cells could be directly related to the tolerance state (e.g. by delivering immunoregulatory effects; Ref. 31). Verification of this hypothesis will require additional investigation including the performance of prospective clinical trials.

Methodological Limitations

In kidney transplantation, the specificity for tolerance of the gene signatures identified in the retrospective cross-sectional studies described earlier is difficult to establish and need further validations. The main problem derives from the lack of appropriate comparators and the difficulty to perform prospective studies, which has been possible in liver transplantation. Indeed, there is no such thing as an adequate control group for allograft kidney recipients who exhibit stable graft function despite receiving no immunosuppressive drugs. The absence of biopsies from tolerated organs is another potential problem. First, a “minimal” form of subclinical rejection cannot be excluded. Second, whether analyses on peripheral blood samples constitute a good surrogate of the immune responses taking place within the graft is unclear. The stability and prognostic value of the kidney tolerance-related signatures is also unknown and will need to be determined in long-term longitudinal studies. In liver transplantation the availability of a prospective immunosuppression withdrawal trial (28) incorporating protocol liver biopsies partially addresses some of the methodological limitations of kidney studies. The cohort of liver recipients from whom the predictive gene signatures were derived is still relatively small. Therefore, the robustness and stability of these signatures in large independent validation cohorts cannot be taken for granted. Furthermore, the long-term outcome and overall clinical benefit of drug withdrawal in liver transplantation is still unknown.

The use of blood as the source for mRNA, while highly desirable due to its accessibility, might also be associated with technical problems. When PBMCs are employed, Ficoll isolation may cause significant interlaboratory variability. Although whole blood RNA stabilization associated with globin mRNA reduction and strict sample handling protocols can address this problem (32), the small differences in transcript levels observed when comparing blood samples from operationally tolerant and nontolerant recipients (less than twofold for most differentially expressed genes) could compromise the reproducibility of predictive models. Another limitation of these studies has been, so far, the fact that they are starting from well-defined but size-limited clinical situations instead of using large cohorts of patients that would encompass many clinical situations in addition to operational tolerance. The influence of blood cell subset heterogeneity (33), still not well studied in the context of transplantation tolerance, can also be a source of variability. Clearly, the use of “deconvoluting” bioinformatic algorithms needs to be further pursued (34). Alternatively, transcriptional analyses of isolated PBMC subsets, a strategy that has proven useful in autoimmune diseases (35), should also be explored as a way to generate robust predictive signatures.

The Long Road to Clinical Application

One of the ultimate goals of the studies described earlier is to find biomarkers to identify recipients who would benefit from partial or complete immunosuppressive drug discontinuation. Although recently published reports represent significant advances in this direction, only one study in liver transplantation has formally demonstrated that tolerance can be predicted. Even in this case additional validation studies will need to be conducted before the accuracy and clinical applicability of the biomarkers can be established. The fact that the three most documented studies on transcriptional signatures of operational tolerance reveal only few overlapping genes reinforces this point (30). The road to clinical application should necessarily be different for liver and for kidney transplantation. In liver transplantation, replication of the reported signatures in a large immunosuppression withdrawal trial would constitute the most robust validation strategy. Before this, adequate standardization of the transcriptional platform to be employed would need to be conducted. In kidney transplantation, where drug withdrawal trials are not acceptable, validation will need to be performed much more cautiously. One approach would be to test the biomarkers predictive capacity in drug minimization trials conducted in subgroups of kidney recipients with low immunological risk. This is currently under process in a clinical study of CNI weaning (http://clinicaltrials.gov/ct2/show/NCT01292525) conducted with the center of the CENTAURE network (http://www.fondation-centaure.org/). As an alternative, biomarker-guided minimization trials could be proposed. It might be preferable however, to gain deeper understanding of the pathogenic role of circulating B-cell subsets in the development of kidney allograft tolerance before biomarker-guided minimization studies are conducted. Until these studies are completed, investigators should be discouraged from conducting immunosuppression weaning attempts on the unique basis of transcriptional biomarkers whose validity has not been confirmed signatures.

Conclusions

Although transplant recipients maintaining stable graft function in the absence of immunosuppressive drugs were first described more than two decades ago, advances in their immune characterization have only been achieved in recent years. Demonstration that operationally tolerant liver and kidney recipients exhibit distinct molecular patterns in peripheral blood and at the graft itself has raised for the first time the prospect of deciphering the mechanisms responsible for drug-free graft survival and of identifying recipients who might safely discontinue all immunosuppressive therapy. Although the studies performed so far exhibit a number of methodological limitations, their results have fueled a renewed interest in the characterization of operationally tolerant patients and have provided the rationale for undertaking well-designed large-scale prospective drug withdrawal or drug minimization clinical trials in liver and kidney transplantation. These studies will be essential to validate the already described biomarkers of tolerance that could be part of composite scores, including clinical characteristics and cell phenotyping. In addition, they will provide answers to critical questions such as the risk/benefit ratio of drug weaning protocols, the stability of the operationally tolerant phenotype, the influence of clinical and immunologic parameters on tolerance development and the long-term histological outcomes associated with drug withdrawal.

Acknowledgments

R.D. was funded by a grant from the French Transplantation Research Network (RTRS) supported by the “Fondation pour la Recherche Médicale”, the “Fondation de Coopération Scientifique” CENTAURE and ROTRF grant. CIBEREHD is funded by the Instituto de Salud Carlos III Spain. INSERM U643 is supported by the Agence de la biomédecine.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

Ancillary