SEARCH

SEARCH BY CITATION

Martínez-Llordella M, Lozano JJ, Puig-Pey I, Orlando G, Tisone G, Lerut J, et al. Using transcriptional profiling to develop a diagnostic test of operational tolerance in liver transplant recipients. J Clin Invest 2008;118:2845–2857. (Reprinted with permission.)

Abstract

  1. Top of page
  2. Abstract
  3. Comment
  4. References

A fraction of liver transplant recipients are able to discontinue all immunosuppressive therapies without rejecting their grafts and are said to be operationally tolerant to the transplant. However, accurate identification of these recipients remains a challenge. To design a clinically applicable molecular test of operational tolerance in liver transplantation, we studied transcriptional patterns in the peripheral blood of 80 liver transplant recipients and 16 nontransplanted healthy individuals by employing oligonucleotide microarrays and quantitative real-time PCR. This resulted in the discovery and validation of several gene signatures comprising a modest number of genes capable of identifying tolerant and nontolerant recipients with high accuracy. Multiple peripheral blood lymphocyte subsets contributed to the tolerance-associated transcriptional patterns, although NK and γδTCR+ T cells exerted the predominant influence. These data suggest that transcriptional profiling of peripheral blood can be employed to identify liver transplant recipients who can discontinue immunosuppressive therapy and that innate immune cells are likely to play a major role in the maintenance of operational tolerance in liver transplantation.

Comment

  1. Top of page
  2. Abstract
  3. Comment
  4. References

It is well established that liver transplants are less likely to be rejected than other organs. This was first observed in pigs, where liver transplants between unrelated animals were often accepted without requiring immunosuppression.1 Subsequent studies in rodents have confirmed the liver's resistance to rejection and its ability to be accepted without treatment. This is reflected in the outcome of clinical liver transplantation, where a significant proportion of liver transplant recipients can be successfully weaned off all immunosuppressive drugs. The first such study reported that more than 20% of stable liver transplant patients could be weaned,2 although the low proportion of stable patients in the total liver transplant recipient population means that the actual percentage weaned is considerably less.3 The problem is to identify those stable patients that can be successfully weaned from those who are likely to reject. The article by Martinez-Llordella et al.4 makes a considerable stride toward achieving this aim and does so using peripheral blood, a relatively noninvasive approach.

They used microarray analysis of stable liver transplant patients who could be completely weaned from immunosuppression (TOL) compared to those in whom weaning led to rejection and the necessity for recommencement of immunosuppression (non-TOL). Using the well-established “Significance Analysis of Microarrays” (SAM) methodology5 they identified 1932 genes out of a total of 47,000 that were significantly differently expressed between 17 TOL and 21 non-TOL patients (with an estimated false positive rate of 5% among the 1932 genes called as differentially expressed). The authors then sought to identify an appropriately small group of assayed genes that could be used to predict TOL. To do so, they used statistical classification methods based on nearest shrunken centroids analysis6 termed “Predictive Analysis of Microarrays” (PAM). Such approaches, based on multivariate statistical analysis and often referred to as “supervised learning”, attempt to find the optimal set of “features” (gene expression measurements, in this case) that can distinguish two groups of samples (patient groups in the current setting). This analysis yielded 26 microarray probes representing 24 unique genes (Table 1). These 24 genes were tested on a cohort of 19 stable patients who had not yet been weaned and identified five patients (26%) as TOL. This was satisfyingly close to the predicted percentage of potentially TOL patients in this group. It also identified the patients with the highest levels of Vδ1TCR+ T cells and the highest ratios of Vδ1:Vδ2 T cells, both of which had been previously demonstrated to be associated with TOL.7, 8

Table 1. Genes Identified by Microarray Analysis (PAM) and PCR (MiPP) as Differentially Expressed Between TOL and Non-TOL Patients
PAM GenesPCR ValidatedPCR Fold ChangeP (TOL vs. non-TOL)MiPP Genes
  • y, yes; n, no

  • *

    Bolded values are significantly different by PCR analysis

  • **

    Italicized genes are represented twice in the 26 probe PAM signature

ARID5Bn-  
CD160y2.078*0.00002 
CLK4n-  
**EIF5Ay−1.0640.43ALG8
FEM1Cy−1.380.0008CLIC3
FEZ1y2.2190.00002CX3CR1
IL-8y−4.5790.04GANCG
MRPS31y1.2610.6GNPTAB
NOTCH2y1.110.14IL2RB
OSBPL5y1.6990.0001KLRB1
PDE4By−1.5210.007KLRF1
POU2F1n- NKG7
PRIM1n- PSMD14
PTGDRy1.5640.0007RGS3
SLAMF7y1.4140.00001SLAMF7
TGFBR3y1.0910.22 
TRA @y−1.1730.6 
UQCRC2n-  
WDR67y1.2480.003 
ZBTB10n-  
ZNF267y−1.5210.008 
ZNF295y−1.8790.001 
ZNF331n-  

Although supervised learning approaches have been extensively used for microarray data to identify “predictive” gene sets, that can, for example, distinguish between cancer subtypes9 or survival times,10 these analyses contain often unappreciated complexities. Attempting to distinguish groups of patients using data sets in which there are many more genes than patient measurements is a somewhat arbitrary process, in which it is possible to generate a large number of equally sensitive and specific predictive sets.11, 12 The authors do not address this explicitly, because their test analysis was undertaken within a single patient selection; however, their subsequent analysis largely addresses this concern. Armed with their list of 24 genes from PAM analysis, they validated the microarray results by quantitative polymerase chain reaction (PCR) of the TOL and non-TOL patients as well as 16 normal controls. For this validation, they included another 44 genes that were most significant by SAM analysis and six genes that were previously reported to be associated with TOL.

This yielded a total of 68 genes that were tested by quantitative PCR. Thirty-four of the 68 genes could be validated by PCR and 30 genes were differentially expressed by arrays but not by PCR. Of these, there was no overlap between PCR primers and microarray probes in 11 cases, making 34 of 53 genes (64%) that were differentially expressed by both microarrays and PCR. Sixteen of the 24 genes identified by PAM were confirmed by PCR; however, only 11 of these were significantly different between TOL and non-TOL by PCR (Table 1). Using the same group of TOL and non-TOL patients and a method for measuring the performance of predictive “gene signatures” (called the misclassified penalized posterior [MiPP] algorithm13) they found three signatures consisting of only 12 genes that correctly classified TOL and non-TOL patients. Expression of these gene signatures was tested on separate groups of 11 TOL and 12 non-TOL patients for whom there was no previous data and gave acceptable levels of prediction, with overall error rates for each signature ranging from 0.13 to 0.17. Only one of the 12 genes identified by MiPP analysis of PCR results, SLAMF7, corresponded to the genes identified on microarrays by PAM (Table 1).

The authors thus had two largely independent sources of data to identify TOL patients: the microarray results analyzed by both SAM and PAM, and the quantitative PCR results analyzed by MiPP. The question now is which method or combination of methods will best identify those patients who can be safely weaned off immunosuppression? The relative accuracy of the two methods was not tested and future studies will need to focus on this. An additional question is whether the gene signatures observed by PAM and MiPP will be applicable to other transplant units where the patient demographics, type of immunosuppression, and prevalence of infections such as hepatitis C might differ. The authors attempted to answer this by examining a number of clinical variables including age, sex, type of immunosuppression, time from transplantation, peripheral blood leukocyte subsets, and hepatitis C status. Only time from transplantation, which had a weak effect on the PAM signature, and hepatitis C, which had a strong effect, showed any correlation. Despite its strong effect, infection with hepatitis C virus did not prevent the correct identification of TOL status. The effect of clinical variables on the MiPP gene signatures was not examined.

As well as the microarray and PCR data, the authors also performed flow cytometry analysis of the peripheral blood leukocyte subsets and correlated the gene expression data with leukocyte phenotype on the same sample. CD4, CD8, Foxp3, CD19, and natural killer T cells showed no correlation with the 24 gene PAM set, whereas natural killer (NK) cells were significantly associated with 10 of the gene probes, Vδ1TCR+ T cells were significantly associated with nine, and γδTCR+ T cells were significantly associated with 13 of the probes. These results show that NK cells and γδTCR+ T cells are associated with the gene signature of liver transplant tolerance and led the investigators to examine the expression of 22 of the most significantly different genes identified by PCR in subsets of peripheral blood leukocytes. This used five TOL and five non-TOL patients and sorted cells into CD4, CD8, γδTCR+ T cell, and non-T cell subsets. There were significant differences in gene expression in one or more of these subsets between TOL and non-TOL patients in 14 of the 22 genes. Finally, in six TOL, six non-TOL, and five control patients, expression of protein levels of SLAMF7, interleukin-2Rβ, KLRB1, KLRF1, CD9, CD160, and CD244 were assessed by flow cytometry on CD4, CD8, γδTCR+ T cells, T cells, NK cells, and CD19 (B cells). These proteins were mainly expressed on NK and γδTCR+ T cells, reflecting the increased levels of γδTCR+ T cells in TOL.7, 8

These studies suggest an approach that yields a robust method to identify potentially tolerant liver transplant recipients. They show that a range of techniques, including microarray analysis, quantitative PCR, flow cytometry, and gene expression studies of sorted leukocyte subsets, can discriminate between TOL and non-TOL patients. In addition, the role of γδTCR+ T cells and NK cells in liver transplant acceptance is of interest because they have been identified by a number of studies of tolerant liver transplant patients.

References

  1. Top of page
  2. Abstract
  3. Comment
  4. References