Analysis of the peripheral T-cell repertoire in kidney transplant patients


  • Patrick Miqueu,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    2. TcLand Expression S.A., Nantes, France
    3. EA 4275 Biostatistique, Recherche Clinique & Mesures Subjectives en Santé, Faculté de Pharmacie, Nantes, France
    Search for more papers by this author
    • These authors have contributed equally as first and senior authors, respectively.

  • Nicolas Degauque,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
    • These authors have contributed equally as first and senior authors, respectively.

  • Marina Guillet,

    1. TcLand Expression S.A., Nantes, France
    Search for more papers by this author
  • Magali Giral,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
  • Catherine Ruiz,

    1. TcLand Expression S.A., Nantes, France
    Search for more papers by this author
  • Annaïck Pallier,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
  • Cécile Braudeau,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
  • Gwenaëlle Roussey-Kesler,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
  • Joanna Ashton-Chess,

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    2. TcLand Expression S.A., Nantes, France
    Search for more papers by this author
  • Jean-Christophe Doré,

    1. UMR 5154, CNRS/MNHN, Paris, France
    Search for more papers by this author
  • Eric Thervet,

    1. Université Paris Descartes and Hôpital Necker, AP-HP, Paris, France
    Search for more papers by this author
  • Christophe Legendre,

    1. Université Paris Descartes and Hôpital Necker, AP-HP, Paris, France
    Search for more papers by this author
  • Maria P. Hernandez-Fuentes,

    1. Medical Research Council Center for Transplantation, King's College London, School of Medicine, Guy's Hospital, London, UK
    Search for more papers by this author
  • Anthony N. Warrens,

    1. Immunology and Renal Medicine, Imperial College, London, UK
    Search for more papers by this author
  • Michel Goldman,

    1. Institute for Medical Immunology, Université Libre de Bruxelles, Charleroi, Belgium
    Search for more papers by this author
  • Hans-Dieter Volk,

    1. Institute for Medical Immunology and Berlin-Brandenburg Center for Regenerative Medicine, Charité University Medicine, Berlin, Germany
    Search for more papers by this author
  • Uwe Janssen,

    1. MACSmolecular Business Unit, Miltenyi Biotec GmbH, Cologne, Germany
    Search for more papers by this author
  • Kathryn J. Wood,

    1. Transplantation Research Immunology Group, Nuffield Department of Surgery, University of Oxford, John Radcliffe Hospital, Oxford, UK
    Search for more papers by this author
  • Robert I. Lechler,

    1. Medical Research Council Center for Transplantation, King's College London, School of Medicine, Guy's Hospital, London, UK
    Search for more papers by this author
  • Dominique Bertrand,

    1. INRA-BIA, Nantes, France
    Search for more papers by this author
  • Véronique Sébille,

    1. EA 4275 Biostatistique, Recherche Clinique & Mesures Subjectives en Santé, Faculté de Pharmacie, Nantes, France
    Search for more papers by this author
  • Jean-Paul Soulillou,

    Corresponding author
    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    • INSERM U643 30, Bd Jean Monnet, 44093 Nantes Cedex 1, France Fax: +33-2-40-08-74-11
    Search for more papers by this author
    • These authors have contributed equally as first and senior authors, respectively.

  • Sophie Brouard

    1. INSERM Unité 643, CHU de Nantes, Institut de Transplantation et de Recherche en Transplantation, Université de Nantes, Faculté de Médecine de Nantes. RTRS “Centaure”, Nantes, France
    Search for more papers by this author
    • These authors have contributed equally as first and senior authors, respectively.


The long-term stability of renal grafts depends on the absence of chronic rejection. As T cells play a key role in rejection processes, analyzing the T-cell repertoire may be useful for understanding graft function outcomes. We have therefore investigated the power of a new statistical tool, used to analyze the peripheral blood TCR repertoire, for determining immunological differences in a group of 229 stable renal transplant patients undergoing immunosuppression. Despite selecting the patients according to stringent criteria, the patients displayed heterogeneous T-cell repertoire usage, ranging from unbiased to highly selected TCR repertoires; a skewed TCR repertoire correlating with an increase in the CD8+/CD4+ T-cell ratio. T-cell repertoire patterns were compared in patients with clinically opposing outcomes i.e. stable drug-free operationally tolerant recipients and patients with the “suspicious” form of humoral chronic rejection and were found significantly different, from polyclonal to highly selected TCR repertoires, respectively. Moreover, a selected TCR repertoire was found to positively correlate with the Banff score grade. Collectively, these data suggest that TCR repertoire categorization might be included in the calculation of a composite score for the follow-up of patients after kidney transplantation.


To prevent graft rejection following kidney transplantation, recipients take lifelong immunosuppression. Despite continuous improvements in such treatments, the half-life of a kidney graft has not increased significantly in the past two decades 1. Manifest by a decrease in renal function that is associated with specific histological lesions 2, chronic rejection remains the major problem of late allograft loss 3. The identification of biomarkers predictive of chronic rejection in patients with a stable graft function would therefore be a valuable tool in patient management 4–6. In contrast to the patients who develop chronic rejection, rare cases exist of kidney recipients who tolerate their graft despite discontinuation of immunosuppression 7.

Operational tolerance and suspicious chronic Ab-mediated rejection are clinical and immunological situations, representing the two opposing endpoints for patients with stable kidney graft function. Indeed, because T cells have been shown to be involved in both chronic rejection and tolerance 8, we have explored the T-cell repertoire in a cohort of patients with stable kidney graft function. We have previously shown, in a small cohort of patients, that both drug-free operationally tolerant patients (TOL patients) and patients with the “suspicious” form of chronic rejection (CHR patients) display a TCR repertoire that differs from healthy, non-transplanted individuals 9–11. In this article, we revisit the relationship between peripheral T-cell repertoire and clinical status, using a new statistical method to analyze the TCR repertoires presented as TcLandscape (TcL) data. We show that, despite being selected according to the stringent clinical and biological criteria, patients with stable graft function display heterogeneous usage of their T-cell repertoire, ranging from unbiased to highly selected profiles. We confirm that the TcL pattern reveals immunological differences between TOL and CHR patients. Furthermore, a positive correlation between peripheral T-cell repertoire profiles and Banff grade is demonstrated. Altogether, these data suggest that the shape of the T-cell repertoire could constitute a valuable parameter which could be used to assess graft outcome, guide the medical management of patient with chronic rejection and indicate the necessity of the long-term follow-up of those stable patients who have an altered T-cell repertoire.


Patients with stable graft function and under immunosuppression display highly diverse TCR repertoire

Evaluating the complexity of the TCR repertoire from spectratyping data, as produced by the TcL technology, needs appropriate statistical method 12. An unsupervised analysis was conducted to explore both the qualitative (Kurtosis of CDR3-length distribution (CDR3-LD) and the quantitative (amount of Vβ transcripts) diversity of the TCR repertoire of the 209 patients with stable graft function (stable for an average of 9 years; range 1.9–22.9 years) on immunosuppressants (mycophenolate mofetil or azathioprine) plus calcineurin inhibitors (STA). Principal component analysis (PCA), a statistical method used to reduce the complexity of data sets, was adapted to TcL data. A factorial map, where a patient's TcL location reflects its overall TCR repertoire diversity was produced (Fig. 1). Eigenvalue decomposition of the covariance matrix shows that PCA C1 and PCA C2 account for a significant amount of the variability (Supporting Information Fig. 1). The widespread location of the patient's TcL in the factorial map highlights the heterogeneity of their T-cell repertoire. As shown in Fig. 1, TcL patterns stemmed from Gaussian repertoire to highly selected TCR usage (low and high PCA C1 values, respectively). To analyze this heterogeneity, a K-means clustering algorithm was applied to the distribution of the C1 coordinate values of the 209 STA patients and four classes of TcL shapes were defined by C1 boundary values of −0.032, 0.008 and 0.071 (dotted lines Fig. 2A). A representative TcL for each of the four classes is shown in Fig. 2B. TcL pattern 1 is composed of “Gaussian-like” Vβ CDR3-LD (Kurtosis KGr1 median=0.10, inter-quartile range (IQR)=0.60). TcL patterns 2 and 3 exhibit an increased level of Vβ CDR3-LD alterations (increased Kurtosis) compared with pattern 1 (Kurtosis KGr2 median=0.89, IQR=0.57; Kurtosis KGr3 median=2.24, IQR=0.73). Pattern 4 characterizes altered TcL with distinct oligoclonal Vβ CDR3-LD (Kurtosis KGr4 median=3.22, IQR=1.05). Multiple group comparisons on Kurtosis show that the four TcL classes are significantly different. Most of the patients (36%) display a TcL belonging to pattern 1, whereas 27%, 18% and 18% of the patients display TcL pattern of classes 2, 3 and 4, respectively (Fig. 2A).

Figure 1.

PCA of 286 TcL. TCR repertoires of TOL (n=14, square), CHR (n=21, triangle) and STA (n=209, circle) patients are plotted against PCA C1 (x axis) and C2 (y axis) (see Supporting Information Fig. 3 for all individuals). PCA allows multidimensional dataset of all TcL to be visualized in a factorial map. See Supporting Information Fig. 1 for variance decomposition.

Figure 2.

Clinically homogenous transplanted patients display diverse usage of their TCR repertoire. (A) Distribution of the PCA C1 coordinates of the STA patients. According to the observed segmentation of the distribution of the coordinates of the 209 STA on the C1 axis, three boundaries at C1 values of −0.032, 0.008 and 0.071 define four classes of TcL patterns (K-means clustering). (B) Examples of TcL of a representative STA patient for each of the four TcL classes (x: CDR3 length, y: TCR Vβ gene, z: qPCR Vβ gene/HPRT, color: perturbation from Gaussian distribution). (C) Percentages and numbers of STA, STN and MIS patients belonging to the four different TcL classes based on their respective PCA C1 coordinates.

The stability of the TcL pattern from STA patients was also investigated by analyzing blood samples harvested at two different time points (between 2.5 and 9.4 months; Supporting Information Fig. 2). The TcL pattern remained stable, displaying similar patterns for the two time-points. Indeed, for each individual with a TcL pattern class 3/4, similar Vβ families with a high Vβ/HPRT ratio and a skewed CDR3 LD were identified. The “Gaussian-like” TCR Vβ repertoire which characterized TcL pattern class 1 was also conserved.

Influence of clinical confounding factors on the peripheral TCR repertoire classification

To investigate the effect of the treatment, and particularly of calcineurin inhibitors on the TCR repertoire classification, we compared the repertoire of the STA patients (n=209) with patients with stable graft function on immunosuppressants (mycophenolate mofetil or azathioprine) but without calcineurin inhibitors (STN patients, n=8) and with patients with stable function under minimal immunosuppression (corticosteroid,<10 mg/day) (MIS patients, n=12). STN and MIS patients (i.e. groups without calcineurin inhibitor) showed no significant difference in term of distribution among the four TcL classes (Fig. 2C and Supporting Information Fig. 3). Thus, immunosuppressive drugs, and especially calcineurin inhibitors, do not have an effect on the TCR repertoire shape.

The influence of clinical and biological parameters on the TcL shape for the STA GenHomme cohort (defined in Materials and methods section) was investigated. Among the different variables investigated, a strong positive correlation was observed between the PCA C1 coordinate and the CD8+/CD4+ T-cell ratio (Spearman test, ρ=0.58, p<0.01). Low correlations were also observed between the shape of the TcL and the recipient age (Spearman test, ρ=0.26, p<0.01), the donor age (Spearman test, ρ=0.24, p<0.01) and the CMV serology (Kendall test, τ=0.298, p<0.01). It is worth noting that the quality of the graft function (proteinuria and creatinemia), numbers of HLA mismatch and the presence of anti-HLA Ab did not influence the shape of the TcL. No strong correlation was found between PCA C2 and the biological and the demographics variables. The relationship between occurrence of bacterial, fungal or viral infections and the TcL shape was explored. Ongoing infections could not account for the skewing of the repertoire, as they were one of the exclusion criteria. The occurrence of these infection episodes did not differ between patients within different TcL classes, except for past CMV disease (Kruskal–Wallis test, p=0.002; Supporting Information Table 1). As expected, all the CMV episodes occurred shortly after the transplantation (median time between transplantation and CMV reactivation episodes: 41, 42.5 and 48 days for patients within TcL classes 1, 2 and 3/4, respectively) and at distance of the TcL analysis (median time between transplantation and TcL analysis >2600 days) (Supporting Information Table 2).

Influence of CMV seropositivity and CMV-specific T-cell response on the TCR repertoire

Due to statistically different CD8+/CD4+ T-cell ratios between TcL classes (Fig. 3) and a weak correlation between CMV seropositivity and TcL class belonging (τ=0.298, p<0.01), we analyzed viral serological status and the incidence of infections in the STA GenHomme cohort. As herpes viruses such as CMV have been shown to produce long-term alterations in the CD8+ TCR repertoire 13, we noticed that CMV seropositivity was weakly correlated to the absolute number of peripheral CD8+ T cells (Kendall test, τ=0.166, p<0.01). The categorization into four TcL classes highlighted that patients within TcL classes 3 and 4, when compared with patients within TcL class 1, exhibit a greater number of CD8+ T cells (TcL classes 3 and 4=669±360 cells/μL, TcL class 1=370±216 cells/μL, p<0.01) and a greater prevalence of CMV seropositivity (CMV+ patients: TcL classes 3 and 4=57%, TcL 1=18%, χ2, p<0.01). Moreover, although 36% of the STA patients display CMV seropositivity, 70% of the TOL recipients with lowest level of TCR repertoire alteration and 64% of CHR with the highest level of TCR repertoire alterations were positive for anti-CMV IgG, showing that no correlation with CMV exists within these two groups of patients.

Figure 3.

Higher CD8+/CD4+ T-cell ratios are associated with an enhancement of TCR alteration. Distributions of the CD8+/CD4+ T-cell ratios of the STA individuals, according to their TcL class. Box-and-whisker plot show the median (center line), the 25th and 75th percentiles (box) and whiskers indicating the lowest/highest data still within 1.5 IQR of the lower/upper quartile. Outliers data, which are beyond the whiskers, are plotted with dots. *p<0.01, Kruskal–Wallis.

To confirm this result, we investigated the contribution of CMV-specific clones to the highly selected T-cell clones in the PBMC and CD8+ T-cell subset. Among the STA cohort, we selected CMV seropositive HLA-A2 patients (n=2) that belong to TcL class 3. CD8+ T cells, pp65-HLA-A2 tetramer-positive and -negative fractions were FACS-sorted. pp65 specificity was chosen because 70–90% of all CTL recognizing CMV-infected cells are pp65 specific 14. By comparing the CDR3-LD (Supporting Information Fig. 4A and B), we confirmed that CD8+ T cells account for the majority of the alterations found in the PBMC repertoire 10, 15. Interestingly, these alterations are found in the pp65-HLA-A2 tetramer-negative CD8+ T cells. A quantitative analysis of the differences calculated Vβ by Vβ between all fractions confirm for both individuals that pp65-HLA-A2 tetramer-negative CD8+ T cells are highly similar to total CD8+ T-cell fraction, whereas important differences are noted with the pp65-HLA-A2 tetramer-positive fraction (Supporting Information Fig. 5). A detailed review of the CDR3-LD in Supporting Information Fig. 4A shows three situations: (i) the pp65-HLA-A2 tetramer-positive CD8+ T cells exhibit the same Vβ CDR3-LD as the negative fraction (Vβ1, Vβ2, Vβ11, Vβ12.1, Vβ15, Vβ17 and Vβ24); (ii) particular expansions are revealed in the positive fraction, but without modifying the CD8+ T-cell Vβ spectratype (Vβ3, Vβ4, Vβ5.2, Vβ6.4, Vβ7, Vβ8, Vβ9, Vβ12.2, Vβ13.5, Vβ14, Vβ16, Vβ18, Vβ21, Vβ22 and Vβ23) and (iii), a few CD8+ T-cell Vβ CDR3-LD are modified by expansions in the pp65-HLA-A2 tetramer-positive fraction (Vβ5.1, Vβ6.1, Vβ6.5 and Vβ13) but have no impact on the PBMC profile. As a conclusion, the pp65-HLA-A2 tetramer+ fraction does not alter the TcL typology of these two patients. Altogether, these data suggest that, even if CMV is positively correlated with TCR repertoire shape, the TCR classification of these patients is not driven by the specific anti-pp65 CMV-specific T-cell response.

Higher expression of cytotoxic/inflammatory gene in patients with an altered TCR Vβ repertoire

TCR Vβ repertoire alteration could be associated with a bias of regulatory/cytopathic immune gene balance. To test this hypothesis, we measured the gene expression of FOXP3 (prototypic regulatory-associated gene), GZMB (prototypic cytotoxicity-associated gene) and T-bet (prototypic inflammation-associated gene) in the PBMC of patients within the STA GenHomme cohort. Patients belonging to the TcL classes 3 and 4 exhibit a decrease in FOXP3 (p=0.0001) expression, and an increase in GZMB (p=0.001) and T-bet (p<0.0001) expression as compared with patients belonging to TcL class 1 (Fig. 4A). Correlations between PCA C1 and gene expression of FOXP3, GZMB and T-bet at the individual level (Fig. 4B) show that FOXP3 gene expression decreased when the PCA C1 value increased (slope=−3.01±0.61; p<0.001). On the other hand, GZMB and T-bet gene expression is increased when the PCA C1 value increased (slope=2.14±0.71, p=0.003 and slope=3.34±0.52, p<0.001 respectively).

Figure 4.

Increase of inflammation-associated genes with the alteration of TCRVβ repertoire. (A) Expression levels of the FOXP3 (regulatory), GZMB and TBX21 (pro-inflammatory) genes were measured by quantitative real-time PCR in the PBMC of patients belonging to TcL class 1 (n=71) and TcL classes 3 and 4 (n=46). Expression levels were calculated using the 2−Δct method after normalization to the housekeeping gene HPRT (mean represented by horizontal bars). p-values are calculated by the Mann–Whitney test. (B) Linear regression between PCA C1 and FOXP3 (slope=−3.01±0.61; p<0.001), GZMB (slope=2.14±0.71, p=0.003) and TBX21 (slope=3.34±0.52, p<0.001) gene expression in PBMC of the STA patients of the GenHomme cohort. Slope and p-value were evaluated using the least squares method. In total, 95% confidence intervals (dotted grey lines) are shown.

Operational tolerance and chronic humoral rejection differently affect TCR repertoires

Finally, we investigated whether the TcL pattern allowed the discrimination of patients with distinct clinical status (operational tolerance versus chronic rejection). PCA C1 values from TOL or CHR patients differ significantly (Mann–Whitney Test, p<0.01; TOL PCA C1 median=−0.04 versus CHR PCA C1 median=0.02; Fig. 1) and sign the immunological differences between the two conditions (Supporting Information Fig. 3). The repertoire of CHR patients displays a higher level of clonal CDR3-LD associated with a higher quantity of Vβ transcripts as compared with the repertoire of TOL patients. Using the four TcL patterns previously defined, we confirmed this observation. More than 90% of TOL patients have the TcL pattern classes 1 and 2 (>60% with a TcL class 1; Fig. 5A). CHR patients exhibit predominately the TcL pattern classes 3 and 4.

Figure 5.

TcL patterns efficiently discriminate CHR and TOL patients. (A) Frequency of CHR (n=21) and TOL (n=14) patients within the four classes of TcL pattern. (B) Correlation between Banff grade and PCA C1 of CHR patients (Banff 1-2 PCA C1 median=−0.002, IQR=0.127, Banff 3 PCA C1 median=0.077, IQR=0.099; *p<0.05, Mann–Whitney).

Interestingly, we noticed that CHR PCA C1 values are directly correlated to the Banff score of patients. Patients with high Banff score show a significantly more altered repertoire than patients with low Banff score (PCA C1 median=0.077, IQR=0.099 versus PCA C1 median=−0.002, IQR=0.127 for patients with grade 3 versus patients with Banff grade 1 Mann–Whitney Test, p=0.0317; Fig. 5B).


We have used a new statistical approach to compare the TCR repertoire typology of a large cohort of 286 patients including TOL, CHR, STA and STN patients. Special emphasis has been put on unsupervised analysis to identify TCR Vβ transcriptional patterns without statistical a priori16. This approach led us to use the Kurtosis of the CDR3-LD, an unbiased metric, which is pertinent for revealing the alteration of CDR3-LD and to estimate its “clonality” 17. Moreover, our statistical approach weights the importance of a CDR3-LD according to the quantity of transcripts rather than performing an independent analysis of these data. We first show that kidney recipients selected for clinical stability (good graft function at least 5 years post-transplantation) displayed heterogeneous TCR patterns from Gaussian to highly selected profiles. Given the large size of the analyzed cohort, we looked for correlation of the TcL topology with the biological and clinical variables collected in the GenHomme database. The factor with the strongest correlation (ρ=0.58, p<0.01) was the CD8+/CD4+ T-cell ratio. Stable recipients displaying class 1 TcL patterns have low to moderate CD8+/CD4+ T-cell ratios, whereas those with classes 3 and 4 patterns have a higher CD8+/CD4+ T-cell ratios. This observation and the fact that altered TCR patterns were positively correlated with the CD8+/CD4+ T-cell ratio are not surprising since CD8+ T cells have been shown to be the main contributor of the alterations of T-cell repertoire in different situations including healthy individuals 18, 19, HIV-infected patients 20, EBV-infected patients 21, 22 and kidney graft recipients 10. We thus identified a sub-group of highly clinically stable patients that accumulated antigen-experienced CD8+ T cells. This observation was strengthen by the fact that inflammation related genes (i.e. GZMB and T-bet) were increased and regulatory associate gene (i.e. FOXP3) was decreased in patients with a skewed Vβ repertoire.

We also found that TCR repertoire usage was significantly different between operationally tolerant recipients and patients with chronic rejection. Patients with chronic rejection displayed peaked Vβ transcript CDR3-LD associated with higher quantity of transcripts, indicating accumulation of oligo or monoclonal Vβ expansions. This skewed TCR usage was not found in patients with chronic renal failure (RFA), suggesting that T-cell alterations reflected rejection process and not kidney dysfunction (Supporting Information Fig. 3). Such results are in agreement with those of Matsutani et al., who reported that the level of alterations of TCR usage was significantly greater in recipients with graft failure 23.

Both persistent and non-persistent viruses have been shown to induce a highly biased T-cell repertoire 21, 24, 25. Among the virus-specific T cells, the T-cell response to CMV has been shown to be large, comprising up to 10% of all CD8 T cells 26–29. In this study, only a low correlation was found between CMV seropositivity status and peripheral TCR repertoire usage of the patients with stable graft function. Briefly, 18% of the patients within TcL class 1 have anti-CMV IgG, whereas 36% of the patients with a stable graft function, whose TcL belong to classes 3 and 4, have anti-CMV IgG. Based on this observation, CMV reactivation was also found to be more frequent in patients with the TcL classes 3 and 4 than in patients with a TcL class 1. These observations raise the question of whether the alterations are related to the CMV seropositivity of stable patients. Based on the aforementioned literature, finding a higher prevalence in patients with altered TCR Vβ repertoire could be expected. However, several lines of evidence suggest that viral infection and CMV infection in particular were not the main reason for the profound perturbation of the TCR Vβ repertoire observed. First, active inflammatory processes (including viral and bacterial infection) at the inclusion time and episodes of acute rejection were exclusion criteria for the recruitment of patients in the GenHomme cohort. The influence of CMV infectious episodes observed shortly after the transplantation in patients from the GenHomme cohort and thus at distance from the TcL analysis was studied. Similar prevalence of anti-CMV IgG was found in operationally tolerant recipients and patients with chronic humoral rejection despite exhibiting dramatically different repertoire usages. Furthermore, in these two groups, no correlation was found between TCR Vβ repertoire usage and CMV serology. Moreover, the analysis of the impact of the CMV pp65-specific T cells on the overall shape of the CD8+ repertoire showed that the TcL typology is not perturbed by CMV pp65-specific clones. Taken together, these data suggest that the TCR classification of the patients cannot be solely related to the CMV response.

We then can hypothesize that such peripheral expansions, and particularly in patients with chronic rejection, could be related to dominant indirect 3 or direct 30 alloimmune responses against the graft. The role of T cells and especially CD8+ T cells had been likely undermined in the process of chronic rejection, whereas several studies confirmed the presence of CD8+ T cells infiltrate in the graft 31–33. Moreover, we have shown that blood of animals (as reported here in patients) with chronic rejection exhibited strong alteration of the CD8+ T-cell repertoire 34. The correlation between the Banff score and the shape of the TcL in this study reinforces the hypothesis that CD8+ T cells may be an instrumental player in chronic rejection. As the magnitude of the clonal selection in recipients with chronic rejection correlates with the severity of the rejection, TcL usage could be a useful tool for graft monitoring in these patients. Further studies on sorted Vβ families with strong alteration, on reactivity against donor cells and a long-term follow-up of the stable patient cohort are awaited for improving the interpretation of TCR alteration in long-term graft recipient. Combined with other biomarker data 9–11 and associated with the expression of inflammation or regulatory-related genes (GZMB, T-bet versus FOXP3) as shown, TCR repertoire categorization might be included in the calculation of a “composite score” for the follow-up of patients to prevent rejection or helping to decide upon immunosuppressant withdrawal.

Materials and methods


Two hundred and eighty-six individuals were included in the study. Two hundred and twenty-five patients have been recruited within a collaborative project (GenHomme, Research French ministry) involving the Nantes Institute of Transplantation, the Center for Adult Transplantation of the Necker Hospital (Paris, France) and the Biotechnology Company, TcLand Expression (Nantes, France). Sixty-one additional patients were recruited in the framework of the European “Indices of Tolerance” Network. The protocol of the study was approved by the Ethical Committees of Nantes and Paris Universities and of the European Commission. All patients signed a written informed consent before inclusion.

Clinical groups

Several different clinical groups were studied (Table 1). Operationally tolerant patients (TOL, n=14) are defined by a stable kidney graft function (Creatininemia<150 μmol/L, Proteinuria<1 g 24 h−1) off immunosuppressive drugs for more than 1 year (mean drug-free duration=8.3±5.7 years) at the time of testing. This definition fulfills EU criteria for operational tolerance (for review, see 4). Immunosuppressive treatment, including corticosteroids, was stopped on account of non-compliance (n=11), calcineurin inhibitor toxicity (n=1), post-transplant lymphoproliferative disorder (n=1) or cancer (n=1).

Table 1. Clinical and biological characteristics of the cohort of patientsa)
 TOL (n=14)CHR (n=21)STA (n=209)MIS (n=12)RFA (n=8)
  • a)

    a) CNI, calcineurin inhibitor; MIS, patients with a stable graft function under minimal immunosupression. This group comprises patients from the GenHomme framework and from the “Indice of Tolerance” framework. RFA Non-transplanted patients with “non-immune” renal failure. Values are mean±SD for continuous variables or number of patients (%) for categorical variables. *One missing data and more than one missing data.

Age (years; mean±SD)49±1649±1852±1450±1238±15
Gender male/female (% male)10/4 (71%)9/12 (43%)126/83 (60%)8/4 (67%)6/2 (75%)
CMV status (%pos)7 (70%)10 (50%)127 (63%)♯2 (67%)♯1 (12%)
Time between sampling and Tx (years)17±108±59±415±7 
Creatinemia (μmol/L)109±25264±14111830117±33654±193
Proteinuria >1 g 24 h−1 (pos/total)2/1413/215/189♯2/111/8
Banff grade2.2±0.8
C4d staining (%)14(67%)
Anti-HLA class I and/or class II Ab (%)3 (23%)♯11(52%)37 (18%)♯3 (25%) 
Donor age (years; mean±SD)29±12*46±1437±14♯25±13♯
Gender male/female (% male)9/1 (90%)11/9 (55%)171/82 (68%)♯5/1 (83%)♯
Number of total HLA mismatches (A, B, DR)2.5±1.23.1±1.72.6±1.7♯2.9±1.4
Medical treatment     
CNI – Tacrolimus, Cyclosporine A (%pos)0 (0%)14 (70%)158 (77%)♯1 (8%)0 (0%)
Corticosteroid – Prednisone (%pos)0 (0%)9 (45%)53 (26%)♯12 (100%)0 (0%)
MMF – CellCept (%pos)0 (0%)11 (55%)95 (47%)♯0 (0%)♯8(100%)
Azathioprine (%pos)0 (0%)0 (0%)62 (30%)♯1 (8%)0 (0%)

Patients with the “suspicious” form of chronic humoral rejection (CHR, n=21) all had a progressive degradation of their renal function (Creatininemia >150 μmol/L and Proteinuria >1 g 24 h−1). In all cases, transplant renal biopsies documented histological signs of chronic humoral rejection at the time of the blood test (Banff 05 grade II or IIIb) with either C4d deposition (in 14 patients out of 21) or circulating anti-donor class II Ab in 11 out of 21 patients. Because the patients had not necessarily both circulating anti-donor class II Ab and C4d deposits, they were referred to as “suspicious” of chronic humoral rejection, as suggested by Banff '07 classification 2.

Long-term stable patients (n=229) comprised patients who had stable kidney graft function on immunosuppresants (either mycophenolate mofetil or azathioprine), supplemented with calcineurin inhibitors treatment in some (n=209 referred as STA) but not in others cases (n=8, referred as STN). Patients also received corticosteroids. The cohort of 209 STA patients is composed of 182 patients recruited from the GenHomme study (patients who have been transplanted at least 5 years previously) and 27 patients from the “Indices of Tolerance” network. Patients were included based on the function of their kidney graft assessed at least 5 years after transplantation (Creatininemia <150 μmol/L, Proteinuria <1 g 24 h−1). Ongoing infection and episodes of rejection defined the exclusion criteria. Among the 229 patients, some were minimally immunosuppressed (n=12, referred as MIS): those were patients with stable graft function on steroid monotherapy (<10 mg/day) in whom any calcineurin inhibitors, mycophenolate mofetil or azathioprine has been discontinued on account of cancer, uncontrolled infectious diseases or calcineurin inhibitor toxicity. Multiple clinical parameters were obtained for the long-term stable patients within the GenHomme project, including donor and recipient demographic characteristics, clinical history of renal graft failure, transplantation monitoring, full blood counts and medications biochemical screening.

Non-transplanted patients with “non-immune” RFA (n=8) had a creatinemia 654±193 μmol/L and proteinuria >1 g 24 h−1. The causes of RFA were polycystic kidney (4/8 patients), renal dysplasia (2/8 patients), interstitial nephropathy (1/8 patients) and malformative uropathy (1/8 patients). Finally, healthy individuals (HEI, non-transplanted individuals, n=14) with normal renal function and no known infectious pathology for at least 6 months prior to the study were enrolled.

Purification of pp65-HLA-A2 CD8+ T cells

PBMC from HLA-A2 CMV+ patients were stained with PE-labeled anti-human CD8 mAb, Alexa700-labeled anti-human CD3 mAb, Alexa 647-labeled anti-human CD4 mAb and pp65-HLA-A2 APC-labeled multimer. DAPI was used to exclude dead cells. pp65-HLA-A2 APC-labeled multimer was prepared by incubating for 1h APC-streptavidin with biotinylated pp65-HLA-A2 monomer. All mAb were purchased from BD Biosciences and biotinylated pp65-HLA-A2 monomer was produced by INSERM core facility (Nantes, France). DAPICD3+CD4CD8+, DAPICD3+CD4CD8+pp65-HLA-A2 multimer and DAPICD3+CD4CD8+pp65-HLA-A2 multimer+ were separated from PBMC using a high-speed cell sorter (FACSAria, BD Biosciences). Purity was greater than 98%.

RNA extraction, preparation of cDNA and T-cell repertoire analysis

Blood, collected in EDTA tubes, was obtained from a peripheral vein or arteriovenous fistula. PBMC were separated on an MSL layer (Eurobio) and frozen in TRIzol® reagent (Invitrogen) for RNA extraction. Total RNA was reverse-transcribed using a classical MMLV cDNA synthesis (Invitrogen). Complementary DNA was amplified by PCR using pairs of primers specific of each Vβ gene 10, elongated and electrophorezed using a gel sequencer (ABI Prism 377 DNA sequencer – Applied Biosystems) 35. The CDR3 profiles obtained were transformed into mathematical distributions and normalized so that the total area was equal to one. In parallel, the level of Vβ family transcripts was measured by real-time quantitative PCR and normalized by a housekeeping gene (HPRT). The CDR3-LD was then combined with each normalized Vβ transcript amounts to obtain the TcL data as described previously 15, 36, 37.


Several parameters or metrics can be used to describe, and summarize with one value, the shape of the Vβ CDR3-LD. Indeed, the distribution of 13 lengths of Vβ CDR3 reflects different immunological situations which can be analyzed 12. Kurtosis, a mathematical index, has been chosen to quantify the CDR3-LD diversity 17. The Kurtosis reflects the degree of “peakedness” of a distribution 38 and is perfectly suitable for describing CDR3-LD with expansions.

Statistical analysis of the TcL

To better explore the level of repertoire diversity 12, we developed a totally unsupervised statistical method. This three-step procedure allows combining the shape of the 26 Vβ CDR3-LD and their respective quantity of transcripts. First, the Kurtosis of each given CDR3-LD of each patient Vβ family is calculated. Second, the Kurtosis value is weighted by the quantity of the Vβ transcripts. Third, PCA, an exploratory statistical technique is used to reduce and extract the major trends of the dataset 38, 39. Indeed, PCA provides “projections” of complex datasets onto a reduced, easily visualized space defined by axes, named component (C). In our context, PCA displays the patients TcL data in a factorial space where the distance between two patients illustrates their TcL similarity. The data processing has been carried out in the Matlab environment (The Mathworks) using the SAISIR package 40 (

Real-time quantitative PCR

Quantitative real-time PCR was performed using an Applied Biosystems GenAmp 7900 sequence detection system. The expression of the genes of interest was analyzed using TaqMan primer-probe sets purchased as “Assay-on-Demand” from Applied Biosystems (Foster City, CA), and normalized to the expression of HPRT. Transcript levels were calculated according to the 2−ΔCt method as described by Applied Biosystems.

Statistical analysis

When data are not normally distributed, median and IQR are calculated. Statistical tests have been performed using SPSS 12.0 and data representation using PAST software (Palstat: Statistics for Palaeontologists and Palaeobiologists. Whalley, J. S., Ryan, P. D., 1995) and Excel 2007 (Microsoft). All correlations are based on non-parametric Spearman ρ and Kendall τ statistics for continuous and ordinal variables, respectively. Kruskal–Wallis and Mann–Whitney tests were considered statistically significant at p<0.05. Least squares method was used to evaluate the linear regression. Bonferroni adjustment has been used for multiple group comparisons. χ2 tests were performed to assess independence between variables, with the Yates' correction for continuity. K-means clustering algorithm has been used to partition a dataset into a predefined number of clusters (PAST software).


This work was supported by the GenHomme funding (French Ministry of Research), the Indices of Tolerance Network ( and the European Consortium RISET (Reprogramming the Immune System for the Establishment of Tolerance, P. Miqueu was supported by TcLand Expression, and N. Degauque is a recipient of a Transplant Society Research Fellowship.

Conflict of interest: Patrick Miqueu, Marina Guillet, Catherine Ruiz and Joanna Ashton-Chess are employees of TcLand Expression S.A., whose statistical tools are used in this study. Uwe Janssen is an employee of Miltenyi Biotec.