Expression Patterns of Regulatory T-Cell Markers in Accepted and Rejected Nonhuman Primate Kidney Allografts


  • The authors declare that no conflict of interest exists.

* Corresponding author: K. Haanstra,


The identification of FOXP3 expressing cells in recipients of an allograft, in particular inside the graft itself, may help to define criteria for immunosuppressive drug withdrawal. We therefore examined expression patterns of several regulatory T-cell (Treg) markers in kidney biopsies and kidney tissues taken at the time of graft rejection from monkeys treated with αCD40, αCD86, CsA, a combination of these or after drug withdrawal.

In advanced stages of rejection, organized multifocal nodular infiltrates, with mature dendritic cells, T cells and B cells could be found. In contrast, interstitial infiltrates contain more macrophages, less T cells and few B cells.

Cells expressing FOXP3, CD25 and CTLA-4 were mainly found in nodular infiltrates of rejected tissue samples. A significant correlation was found between the percentage FOXP3+ cells and markers for rejection, i.e. creatinine levels and Banff interstitial and tubular infiltrate scores. The type of immunosuppression did not influence the percentage of cells expressing Treg markers. Three animals with prolonged drug-free survival showed low numbers of FOXP3+ cells.

We conclude that the presence of intragraft FOXP3+ cells is not confined to tolerated grafts, but should be considered as part of the normal immune response during rejection.


Research on the prevention of graft rejection has a strong focus on the induction of Tregs. They are thought to be important for the induction and maintenance of tolerance towards an allograft, obviating the need for life-long immunosuppression. Although several phenotypically distinct Tregs have been described, the best characterized are naturally occurring CD4+CD25+ T cells, which express CTLA-4, GITR and in particular FOXP3 (1–6). The mechanism by which Tregs suppress allograft rejection is unknown. One of the reported mechanisms is the secretion of immunomodulating cytokines such as TGF-β and IL-10 (5–8). We have observed in monkeys treated with costimulation blockade that as early as 3 weeks after transplantation many animals show significant graft infiltration without loss of graft function (8–11). Tregs with the functional capacity to regulate effector T cells were found to be present inside tolerated murine skin allografts (12,13). We hypothesize that induction of Tregs to the allograft is not restricted to the draining lymph nodes, but may also take place within the graft. A positive correlation between Tregs present in the graft and the absence of graft rejection could then be a useful marker for graft tolerance. As FOXP3 has been described to be an exclusive marker for Tregs, we have investigated the intragraft FOXP3 expression of accepted versus rejected kidney grafts by immunohistochemistry. The expression of other markers associated with Tregs, such as CD25, CTLA-4 (CD152), GITR and CD103, was investigated as well and compared this with a marker for effector T cells: granzyme B.

Materials and Methods


Rhesus monkeys (Macaca mulatta) were BPRC-bred or purchased from a commercial breeding station and housed at the Biomedical Primate Research Centre. All procedures were performed in accordance with the guidelines of the Animal Care and Use Committee installed by Dutch law. Heterotopic kidney allotransplantation with bilateral nephrectomy was performed as described previously (14,15). Kidney biopsies were taken at regular intervals: approximately monthly during the first 3 months posttransplantation and three times per year thereafter. The biopsies and the necropsy materials were scored for rejection according to the Banff'97 criteria (16). Results of these studies have been reported previously (9–11,15,17–20).

Immunosuppressive treatment

The following immunosuppressive treatments were given at times of tissue sampling: CsA (Novartis; intramuscular (i.m.), 2 to 10 mg/kg) once daily either up to day 35 posttransplantation or for 4–6 months (15, 18–20), or in one case, for 12 months (19). Some animals received prior to tissue sampling Rapamycine (Wyeth), orally once daily up to day 20 posttransplantation (18) and/or αCD3 toxin, intravenous (i.v.) as a bolus injection twice in the first week posttransplantation (18). αCD40 and αCD86 were given twice weekly i.v. for 8 weeks (10). When CsA treatment followed αCD40 +αCD86, this was given over a 3-month period starting in week 6 posttransplantation with tapering dosages (9). Details of immunosuppression and tissue sampling time points can be found in supplementary material Table (S1) (available online).


Frozen 5 μm sections were air-dried, fixed and preincubated with avidine and biotine to block aspecific reactions. When a peroxidase technique was used, endogenous peroxidase activity was blocked using Peroxidase blocking reagent (Dako, Belgium). The following antibodies were used: CD3 (FN18, BPRC, The Netherlands), CD8, CD20, CD68 (clones DK25, L26 and KP1, Dako), CD25 (B-B10, Diaclone, France), CD83 (HB15A; Beckman Coulter, The Netherlands), CD103 (2G5; Beckman Coulter), CD152 (CTLA-4; clone BNI3, BD Pharmingen, Belgium), GITR (AF689, polyclonal goat antibodies, R&D Systems, UK), FOXP3 (hFOXY, eBioscience, San Diego, CA, or ab2484, polyclonal goat antibodies, Abcam, UK); and for CD4, a mix of the following antibodies was used: OKT4, OKT4a, (Johnson & Johnson, NJ) RIV6, RIV7 (RIVM, The Netherlands) and MT-310 (gift of prof. Rieber, Germany). The slides were developed using a biotinylated rabbit anti-mouse Ig (Dako) or biotinlylated donkey anti-goat Ig (Jackson Immunoresearch, UK) for the polyclonal goat antibodies. Next, slides were incubated with HRP- or AP-labelled StreptABcomplex (Dako). Staining was visualized using DAB Fuchsine (Dako) and slides were counterstained with haematoxylin. Slides were scored blind. The whole tissue sample was examined and then representative areas of renal cortex and medulla containing mononuclear infiltrates were selected. Although the medulla in general has a lower intensity of infiltrates, no difference was found in the percentages of cells stained for the markers analyzed, both for the nodular as well as for the diffuse infiltrates. All infiltrating cells including the trivial interstitial inflammatory cells were included in the counts. The results were expressed as a percentage of all mononuclear infiltrating cells in at least three fields of view per infiltration type (nodular, diffuse). For biopsies with nodular infiltrates all present fields were evaluated. The diffuse infiltrates were evaluated through the entire sample. The results for the tubuli are expressed as the percentage of tubuli with positive cells.

Immunofluorescence stainings

A selection of tissue samples was used to double stain FOXP3 with CD4, CD8, CD25 or CTLA-4. Frozen sections were treated as described above. Slides were incubated with FOXP3 (hFOXY, eBioscience), followed by a biotin labelled donkey anti-mouse antibody (Jackson Immunoresearch) and Cy3 labelled streptavidin (Jackson Immunoresearch) CD4 (1F6, Monosan, The Netherlands), CD8, CD25 or CTLA-4 are labelled with Zenon Alexa Fluor 488 mouse IgG1 labeling kit (Invitrogen, The Netherlands) and slides were incubated. Slides were mounted in DAPI containing Vectashield hardset mounting medium.

Quantitative real-time polymerase chain reaction Q-PCR

mRNA extraction, cDNA transcription and DNA amplification was performed as described in detail before (21). Briefly, the relative level of FOXP3 mRNA was determined on the ABI Prism 7700 sequence detector (Applied Biosystems, Foster City, CA) using the Assay-on-Demand product for detection and quantification of FOXP3 (Hs00203958_m1) and 18S (Hs99999901_s1) mRNA (Applied Biosystems). cDNA was added to the PCR mixture containing Absolute Q-PCR ROX dUTP Mix (Abgene, UK) and specific Primer & Probe-on-Demand mix. The number of copies of cDNA was calculated as follows: 2(total Ct–sample Ct). The obtained values were normalized to the housekeeping gene 18S present in each sample and multiplied by 106.

Statistical analysis

The data were analyzed according to the groups as listed in Table 1. As using multiple samples from one animal might lead to pseudoreplication, we have corrected for this. Where more than one tissue sample per animal was present per group, the mean values of these samples were used for the calculation of mean values. Data are given as mean ± SEM, unless otherwise indicated. GraphPad Prism for Mac OS X (version 4.0b) software was used for statistical analysis. Spearman r was used to calculate correlations. Significances of differences between tissues with and without rejection were calculated using the Mann–Whitney U-test. Significances of differences between immunosuppressive groups were calculated using the Kruskal–Wallis test. A p-value of ≤0.05 was considered significant.

Table 1.  Identification of kidney tissues and groups
GroupImmunosuppression at the time of biopsyGroup IDCreatinineN1Acronym
  1. Biopsies were divided into groups according to the presence or absence of rejection and subsequently according to the type of rejection. Tissues from animals without rejection and without immunosuppression were subdivided into tissues taken within 6 months after cessation of immunsuppression (group 1c) and tissues taken after more than 6 months after cessation of treatment (long-term survival (LTS), group 1d). Creatinine levels at the time of tissue sampling have been given as median (μmol/L) (range).

  2. 1Number of tissue samples/number of animals

  3. 2Five biopsies (3 in COS-no rej and 2 in CsA-no rej) taken with a 20% creatinine rise were found to be due to another cause (CMV, occluded urether) and were identified as no rejection.

1aCOS-blockade (anti-CD40 + CD86)COS-no rej108 (61–926)229/18No rejection (no rej) 86/281
1bCsA (through level > 300 ng/mL)CsA-no rej80 (61–1120)216/10 
1cNo immunosuppression, subsequent rejectionnoIS-no rej103 (56–198)14/10 
1dNo immunosuppression, long-term survivalLTS110 (70–141)27/4 
2aCOS-blockade (anti-CD40 + CD86)COS-rej828 (505–1207)4/4Rejection (rej) 46/39
2bCsA (through level > 300 ng/mL)CsA-rej947 (234–1589)11/11 
2cNonenoIS-rej588 (218–1705)31/24 


Histological aspect of kidney grafts

The tissue samples were divided into groups according to the presence or absence of rejection. Tissue samples were classified as ‘rejection’ when histological signs of rejection according to Banff (acute rejection score IA, (16)) were present, accompanied by a significant serum creatinine rise of ≥20% compared to the preceding time point 3 to 4 days earlier. Tissue samples were classified as ‘nonrejection’ when no significant creatinine rise due to rejection of the graft was observed. This group includes samples free of rejection and samples with sub-clinical rejection according to the Banff classification and may have an acute rejection score of maximally IA. Included in the nonrejection group are five samples with creatinine rises due to cytomegalovirus (CMV) or an occluded urether. A significant correlation between the Banff score and serum creatinine was evident in the complete dataset. Both infiltration (i) and tubulitis (t) scores were correlated with serum creatinine (Spearman r = 0.58, p < 0.0001; Figure 1A and Spearman r = 0.49, p < 0.0001, respectively). The tissues were grouped as indicated in Table 1. A total of 86 tissue samples from 28 animals without rejection and 46 samples from 39 animals with rejection were analysed. The tissues were subdivided according to the type of immunosuppression at the time of tissue sampling. If more than one biopsy per animal was available, the average of all available samples per animal was taken, so that each animal is represented once per type of immunosuppression group. Biopsies were classified as long-term survival (LTS) when monkeys had no signs of rejection and all immunosuppression had been stopped at least 6 months prior to the biopsy collection.

Figure 1.

Interstitial and nodular infiltrates in kidney biopsies. (A) Increasing Banff infiltration scores (i) are moderately correlated with increasing serum creatinine levels (Spearman r = 0.55, p < 0.0001). (B,C) H&E staining clearly identifies infiltrating cells, which sometimes present as interstitial infiltrates (B,D,F) or as nodular infiltrates (C,E,G). CD3 staining (D,E) and CD20 staining (F,G) demonstrates the lymphoid like organization of nodular infiltrates, with clear T- and B-cell areas.

First differences in graft infiltrates between tissue samples with and without rejection were analyzed. Infiltrates consist of nodular infiltrates and diffuse interstitial infiltrates (Figures 1B and C). The nodular infiltrates seem to develop over time and include all nodular types as described by Mengel et al. (22). Biopsies without rejection have relatively small nodular infiltrates (Figures 1D and C); while biopsies with rejection contain larger nodular infiltrates that resemble lymphoid follicles with clear T-cell (Figure 1E) and B-cell areas (Figure 1G). The composition of the infiltrates differs between rejected and nonrejected tissues (Table 2). The nodular infiltrates in nonrejected tissues are composed of 26% CD8+ and 67% CD4+ cells, 11% are CD68+ macrophages and 12% are CD20+ B cells. As CD4 also stains macrophages the number of CD4+ T cells is probably less than 67%. Diffuse infiltrates differ substantially from nodular infiltrates as they contain less CD4+ T cells, B cells and CD83+ DCs, but more CD68+ macrophages (Table 2, all p-values <0.01). The percentage of CD8+ T cells is remarkably similar in nodular and diffuse infiltrates, in both rejected and nonrejected tissues. CD4 in rejected tissues and CD103 in nonrejected tissues are also not significant between nodular and diffuse infiltrates. These findings indicate that the infrastructure for antigen presentation is present inside the graft and, in particular, in the nodular infiltrates that resemble lymphoid follicles. A clear difference between rejected and nonrejected tissue samples was noted with regard to tubular infiltrating cells. While CD4+ T cells were equally present in both groups, more tubules contained CD8+ and/or CD103+ T cells in rejection tissue samples (Table 2).

Table 2.  Infiltrate analyses of tissue samples, divided by the presence or absence of rejection
 Nodular infiltratesDiffuse infiltratesTubular infiltrates
No rejectionRejectionp-Value1No rejectionRejectionp-Value1No rejectionRejectionp-Value1
  1. Numbers of cells positive for indicated markers in nodular or diffuse infiltrates are expressed as percentage of the total number of infiltrating cells. Tubular infiltrates are expressed as the number of tubules positive for indicated markers. Given is the mean ± SEM (number of animals analyzed).

  2. 1p-values t-test comparing rejected versus nonrejected tissues; 2 NA not applicable.

  3. No cells positive for these markers are found within tubuli or nodular infiltrates.

CD371 ± 3 (27)75 ± 2 (21) 45 ± 3 (28) 55 ± 3 (22)0.012912 ± 2 (28)31 ± 4 (23)<0.0001
CD467 ± 2 (28)71 ± 3 (27) 36 ± 3 (34) 62 ± 4 (28)<0.00015.5 ± 1 (34)6.6 ± 1 (28) 
CD826 ± 2 (29)26 ± 3 (26) 25 ± 2 (34) 27 ± 2 (27) 10 ± 2 (28)20 ± 3 (27)0.0147
Granzyme B   1.7 ± 0.4 (28)   1.7 ± 0.5 (21) 5.5 ± 0.9 (33)4.0 ± 0.5 (21) 1.1 ± 0.3 (33)1.5 ± 0.5 (21) 
CD2012 ± 2 (27)21 ± 3 (22)0.00132.1 ± 0.4 (30)3.7 ± 0.6 (23)0.0084 NA2 
CD6811 ± 1 (25)12 ± 1 (21) 27 ± 2 (29) 36 ± 3 (22)0.0137 NA 
CD8315 ± 2 (32)12 ± 2 (20) 2.5 ± 0.5 (32)5.5 ± 1 (21)0.0246 NA 
CD25   4.1 ± 0.7 (39)   9.0 ± 0.9 (38)<0.00011.6 ± 0.4 (38)4.8 ± 0.8 (37)<0.0001 NA 
FOXP3   2.8 ± 0.4 (37)   6.4 ± 0.6 (32)<0.00012.2 ± 0.6 (38)3.7 ± 0.8 (29)0.0091 NA 
CTLA-413 ± 1 (35)20 ± 3 (32) 6.0 ± 0.9 (31)11 ± 1 (29)0.0007 NA 
GITR   4.5 ± 0.6 (38)   3.6 ± 0.6 (27) 1.3 ± 0.3 (36)1.3 ± 0.3 (25) NA 
CD103   2.1 ± 0.4 (32)   2.9 ± 0.8 (21) 2.9 ± 0.4 (34)6.8 ± 1.2 (21)0.000411 ± 2 (34)26 ± 4 (22)0.0008

Subsequently, we analyzed if the type of immunosuppression influenced the nature of the infiltrates of rejected tissues. Similar patterns of graft infiltrates were seen in tissue samples taken during costimulation blockade (COS-rej) or without immunosuppression (noIS-rej), while reduced percentages of CD4+, CD20+ and CD83+ cells were found during CsA treatment (CsA-rej). This latter pattern resembled more the pattern seen in nonrejected tissues.

Also tissues identified as nonrejected were subdivided according to the type of immunosuppression (COS, CsA, no immunosuppression or LTS). The composition of the infiltrates was generally not different between groups, with two exceptions. Tissues of the noIS-no rej group (group 1c) contain more CD8+ T cells in the diffuse infiltrates (36%, p = 0.0178, Kruskal–Wallis test), as compared to the tissues of the other groups without rejection (groups 1a, b and d; 19%, 26% and 20%, respectively). The second exception are the tissues of the LTS group, which contain more CD20+ cells in nodular infiltrates (36%, p = 0.0203) than tissues from other nonrejecting animals (8%, 9% and 15%, groups 1a, b and c, respectively), but this number is also higher than tissue samples with rejection (Table 2). We conclude that the composition of the infiltrate, especially the diffuse infiltrate, is different when rejection is present or absent, but the composition is less influenced by the type of immunosuppression (COS, CsA, or no immunosuppression).

Regulatory T-cell markers inside the graft

Currently FOXP3 is considered the best marker for Tregs. We therefore investigated the percentage of cells positive for FOXP3 and its correlation with graft survival. Presence of other Treg markers (CD25, CTLA-4, GITR and CD103) was also investigated. FOXP3+ cells were observed in many biopsies, mostly in nodular infiltrates of rejected kidneys (Figures 2A and B). Tissues from kidneys without rejection (but with infiltrates!) show lower percentages FOXP3+ cells in nodular infiltrates (2.8% vs. 6.4%, p < 0.0001, Mann–Whitney U-test, Table 2 and Figure 2D) as well as in diffuse infiltrates (2.2% vs. 3.7%, p = 0.0091, Table 2 and Figure 2C). As the absolute cell density of diffuse infiltrates in nonrejected grafts is much lower than in the rejected grafts, the difference in absolute number of FOXP3+ cells per surface area between nonrejected and rejected tissues is even greater. The cell density of nodular infiltrates is not different between rejected and nonrejected tissues, only the surface area of nodular infiltrates increases upon rejection. Therefore, expressing the number of FOXP3+ cells per nodular surface area gives similar results as expressing the data as percentage of total infiltrating nodular cells. Furthermore, Spearman correlation coefficients of FOXP3 versus creatinine, Banff i score, or Banff t score also indicated that FOXP3 is positively correlated with rejection (Spearman r = 0.49, 0.47 and 0.42, respectively; all p < 0.0001). FOXP3 in nodular infiltrates was also positively correlated with CD8+ cells in tubuli (Spearman r = 0.63, p < 0.0001). However, the percentage of FOXP3 expressing cells was not correlated with the percentage of granzyme B+ cells in the tubuli (Spearman r = 0.15, p > 0.05), nor was the percentage of granzyme B+ cells correlated with the presence or absence of rejection (Table 2). We also determined the percentage of cells positive for CD25, CTLA-4, GITR and CD103 in nodular and diffuse infiltrates in these tissue samples. The percentages of CD25+ and CTLA-4+ cells followed similar patterns as the FOXP3 staining: highest in rejected and lowest in nonrejected samples. GITR staining did not differ between nonrejecting and rejecting animals. Percentages cells positive for these markers in diffuse infiltrates were less compared to percentages in nodular infiltrates. A different staining pattern was observed for CD103, as the highest percentage was found in diffuse infiltrates from rejecting animals (Table 2). No CD25, FOXP3, CTLA4 or GITR positive cells were found inside the tubules. CD103 was expressed in tubules and there the percentage was similar to the percentage of CD3+ cells in the tubules.

Figure 2.

FOXP3 staining. FOXP3 staining is present in rejected (A,B) and nonrejected tissues (C,D), both in interstitial (A,C) as well as nodular (B,D) infiltrates, although it is mainly expressed in nodular infiltrates.

Subsequently, we investigated the influence of immunosuppression on the percentages of cells positive for Treg markers in nodular infiltrates, since these displayed the highest percentages of these markers. CsA decreases FOXP3 mRNA and protein expression in vitro (21,23,24) and calcineurin inhibitors have also been reported to reduce FOXP3+ cells in the blood of renal transplant patients (25, 26). We could not find lower FOXP3 expression during CsA treatment neither in the absence (Table 3) nor in the presence of rejection (Table 4). An effect of CsA treatment was observed in group 2b, where the percentage CD103+ cells was significantly lower than in other groups with rejection (Table 4).

Table 3.  Analysis of nodular infiltrates in nonrejected tissue samples, divided by type of immunosuppression
 COS-no rej Group 1aCsA-no rej Group 1bnoIS-no rej Group 1cLTS Group 1d
  1. Numbers of cells positive for indicated markers in nodular infiltrates are expressed as percentage of the total number of infiltrating cells. Given is the mean ± SEM (number of samples analyzed).

CD254.7 ± 1.1 (18)4.2 ± 2.0 (9)3.6 ± 0.8 (9)1.3 ± 0.5 (3)
FOXP32.5 ± 0.5 (18)3.0 ± 0.8 (8)3.5 ± 1.1 (8)1.7 ± 0.2 (3)
CTLA4 14 ± 2.1 (17) 11 ± 3.0 (7) 13 ± 1.9 (8)9.3 ± 1.3 (3)
GITR6.1 ± 1.1 (17)2.7 ± 0.6 (9)4.2 ± 1.2 (9)1.5 ± 0.5 (3)
CD1031.1 ± 0.3 (14)2.8 ± 0.9 (6)3.8 ± 1.3 (8)1.0 ± 0.7 (4)
Table 4.  Analysis of nodular infiltrates in rejected tissue samples, divided by type of immunosuppression
 COS-rej Group 2aCsA-rej Group 2bnoIS-rej Group 2c
  1. 1CsA-rej significantly lower than COS-rej (p < 0.05) and than noIS-rej (p < 0.01).

CD2510.3 ± 4.4 (4)  7.1 ± 1.0 (11)9.7 ± 1.2 (23)
FOXP38.8 ± 0.4 (4)6.7 ± 1.0 (8)5.7 ± 0.8 (20)
CTLA4 28 ± 7.5 (2) 12 ± 2.7 (7) 21 ± 3.4 (23)
GITR3.6 ± 1.2 (3)1.8 ± 0.6 (4)4.0 ± 0.8 (20)
CD1033.8 ± 1.4 (4) 0.3 ± 0.21 (7)4.4 ± 1.3 (10)

Since the presence of Tregs has been associated with tolerance and drug-free graft survival, we investigated the expression of Treg markers in the LTS group. Although we only have material from four monkeys in the LTS group, in accordance with the finding that percentages of cells positive for FOXP3, CTLA-4 and CD25 was associated with rejection, the percentages cells stained positive for these markers was very low in the LTS group, but not significantly different from the other groups (Table 3). Figure 3 shows a longitudinal analysis of FOXP3, CD25 and CTLA-4 positive cells from two LTS monkeys, together with longitudinal data from two monkeys that were treated identical, but rejected their graft after cessation of treatment (9), demonstrating different patterns between the two rejecting animals versus the LTS monkeys.

Figure 3.

Longitudinal analysis of FOXP3, CTLA-4 and CD25 positive cells. Four animals treated with αCD40 +αCD86 for 56 days and CsA from day 42, with tapering levels until day 126 (9). The two monkeys represented in panels A and B rejected their grafts after cessation of treatment at 140 and 231 days, respectively. The two monkeys represented in panels C and D are alive and well, over 4 years after cessation of treatment, and are included in the LTS group (group 1d). Although incidental outliers are found (CTLA-4 expression panel C on day 308), rejection is characterized by an increase in the percentage positive cells for these markers, while expression is generally low in long-term surviving animals. Interestingly, the percentage positive cells of most markers are lower in these animals during CsA treatment (day 70 high dose and day 112, low dose) as compared to during costimulation blockade treatment (day 21 and day 42). This finding could not be confirmed in larger cohorts (Table 3 and 4).

Thus we found no differences to percentages of cells expressing Treg makers related and the type of immunosuppression, while in tissues with rejection percentages of CD25, FOXP3 and CTLA-4, positive cells were significantly higher than in samples without rejection.

FOXP3 mRNA expression in allograft tissue samples

Fontenot et al. reported that mRNA FOXP3 expression was not always concordant with protein FOXP3 expression (27). Hence, we also analyzed a number of tissue samples for mRNA expression of FOXP3 using qPCR to compare this with FOXP3 protein expression. We found that low protein expression correlates with low mRNA expression (Figure 4A; Spearman r = 0.54, p = 0.0004). Similar to FOXP3 protein, highest FOXP3 mRNA levels were found in rejected kidneys, while only low levels were found in nonrejecting animals (Figure 4B).

Figure 4.

Expression of FOXP3 mRNA in tissue samples. (A) A clear correlation (Spearman r = 0.54, p = 0.0004) was found between number of FOXP3+ cells in nodular infiltrates and FOXP3 mRNA as determined by quantitive PCR. (B) The biological significance of FOXP3 mRNA levels in our kidney tissue samples was further demonstrated by analyzing FOXP3 mRNA expression in tissue samples with and without rejection separately, which was compared with the protein FOXP3 expression, demonstrating that low protein expression correlates with low mRNA expression. In this small set of tissue samples, differences between rejection and no rejection reach near statistical significance for both FOXP3 mRNA expression (p = 0.0604) and FOXP3+ cells in nodular infiltrates (p = 0.0731).

Phenotype of FOXP3+ cells

FOXP3, CTLA-4, CD25 and GITR were all predominantly expressed in the nodular infiltrates. We therefore investigated the costaining of FOXP3 with CD4, CD8, CD25 and CTLA-4 in a representative subset of our allografted kidney tissue samples. Most FOXP3+ cells were also CD4+ (Figure 5A), while only very few FOXP3 CD8 double positive cells were detected (Figure 5B). Double staining with CD25 and CTLA-4 revealed costaining of FOXP3 with CD25 or CTLA-4, FOXP3 single positive cells, as well as CD25 or CTLA single positive cells (Figures 5C and D).

Figure 5.

Double staining of FOXP3 with CD4, CD8, CD25 and CTLA-4. A selection of tissue samples with high percentages of FOXP3+ cells was used to investigate double staining of FOXP3 with CD4, CD8, CD25 and CTLA-4. Almost all FOXP3+ cells are CD4+ (A), while almost none are CD8+ (B). Double staining with CD25 (C) or CTLA-4 (D) demonstrates that not all FOXP3+ cells are CD25 and CTLA-4 positive. In addition, we also find CD25+ and CTLA-4+ cells that are not FOXP3+, which could be activated cells.


The often serious side effects of life-long immunosuppression fuels the search for alternative approaches to prevent graft rejection. This search is hampered by the lack of reliable markers that predict maintenance of the unresponsiveness toward the graft after diminution or stopping immunosuppression. The easiest accessible compartment to measure immune parameters is the peripheral blood, but this may not always reflect the status of the immune system at the site of the immune response (28).

When investigating tissue samples in the rhesus monkey kidney allograft model, we observed the presence of nodular infiltrates as well as diffuse interstitial infiltrates. Development of nodular infiltrates into well-organized lymphoid follicles with defined T- and B-cell areas was apparent in many tissue samples. The development of these so-called tertiary lymphoid organs is a well-known phenomenon in transplanted organs as well as in other chronic inflammatory conditions (29). We found CD83+ DCs predominantly in these nodular infiltrates, suggesting that in this location antigen presentation takes place. It was therefore of interest to investigate these infiltrates in more detail with the aim to find possible evidence for T-cell regulation in accepted grafts. Several markers associated with Treg function have been described. CD25 is expressed on Tregs however, as also effector T cells express CD25, it seems a less reliable marker. In our study, we found CD25+ cells most abundantly in kidneys undergoing acute rejection. This may indicate that CD25 is indeed expressed on effector T cells. However, the percentages of CD25+ cells were relatively low and they were not observed inside the tubular epithelium, while CD4+, and more predominantly, CD8+ and CD103+ cells were seen in the tubules at the time of rejection (Table 2).

The high expression levels of CTLA-4 and GITR on T cells, combined with suppressive activity in vitro, suggests that these markers could be used to identify Tregs (30,31). As both CTLA-4 and GITR are, like CD25, upregulated on activated T cells, we investigated if these markers may be a better indicator of regulation inside the graft. We found that higher percentages of CTLA-4+ cells were present in rejected kidneys and that percentages GITR+ cells were similar in rejected and nonrejected kidneys, thereby excluding these markers for evaluation of tolerated grafts. The CD103 staining pattern was not correlated with tolerance; in contrast, CD103 was mainly present in diffuse infiltrates in rejected tissues, but even more in tubules of rejected kidneys. The stained cells are most likely CD8+CD103+ cytotoxic T cells (32), as we found equal amounts of CD8+ and CD103+ cells in the tubules, while the number of CD4+ T cells in the tubuli was much lower and equal in rejected and nonrejected tissues.

More recently, FOXP3 was identified as a Treg specific marker, exclusively associated with suppression (27,33) and FOXP3 mRNA can be found in tolerated skin and heart grafts (13,34). In addition, a regulatory phenotype is induced under various conditions, including allogeneic stimulation in both CD4+CD25 cells and CD4+CD25+ cells in vitro, which is accompanied by the induction of FOXP3 (35–39). However, induced FOXP3 expression does not always lead to induction of a regulatory phenotype (40, 41).

In our nonhuman primate kidney transplant model we found that the percentage of FOXP3+ cells is significantly higher in samples from rejected kidneys as compared to samples from nonrejected kidneys. We have previously described a similar finding in human cardiac allograft biopsies. High levels of FOXP3 mRNA were found during rejection (21), although expression of FOXP3 mRNA was similar between groups when expression was normalized against TCR mRNA. Staining for FOXP3 and other Treg markers in the LTS group is, in accordance with expression in other nonrejected tissues, very low. However, the number of animals in this group is only 4, and results found in these four animals may not hold in larger cohorts of long-term surviving animals, that may have other tolerance mechanisms.

We did not detect cells in the tubular epithelium that expressed Treg markers (FOXP3, CD25, CTLA-4 or GITR), which is in contrast to the FOXP3+CD4+ cells found inside tubuli of human allograft biopsies (42). Whether this is due to differences in staining methods (frozen sections vs. formalin fixed material), species differences, differences in immunosuppression at the time of tissue sampling or due to the different anti-FOXP3 Ab used is not clear and should be investigated further.

We have previously reported higher percentages of FOXP3+ and CTLA-4+ cells in day 21 kidney allograft biopsies of monkeys treated with costimulation blockade posttransplantation, as compared to biopsies of monkeys treated with costimulation blockade plus ATG induction (11). ATG treated animals rejected significantly earlier than costimulation blockade control treated animals. We conclude that ATG treatment reduces FOXP3+ cell percentages, in spite of subsequent rejection. At end-stage renal failure, ATG treated animals have high FOXP3 expression.

Decreased FOXP3 mRNA was found in PBMC of kidney allograft recipients with chronic rejection as compared to PBMC of tolerant patients or healthy individuals (43). Although this seems contradictory to our findings, it may also be possible that FOXP3 expression in the blood does not reflect the FOXP3 expression in the graft. FOXP3+ cells may home to the graft, where their action is needed. We have not determined the FOXP3 expression in PBMC, and the relation between the FOXP3 expression in the graft and in the PBMC should be investigated further. High FOXP3 and granzyme B mRNA were found in kidney biopsies with rejection, with a decreased FOXP3/granzyme B ratio (44). We have not found increased numbers of granzyme B+ cells, although the intensity of the staining increased with the severity of rejection.

How then can we explain the most abundant presence of FOXP3, CD25 and CTLA-4 positive cells at the time of graft rejection? We can assume that the presence of FOXP3+ Tregs in rejected kidneys is a physiological reaction to an inflammatory response, to downregulate the effector T cells, similar to several groups reporting Treg activity in response to immune activation (21,45,46). A second, not mutually exclusive explanation could be that FOXP3, like CD25 and CTLA-4, is not a marker for regulation, but only for activation, as was found in vitro (40,41).

Surprisingly, no effects on the percentages of Treg marker positive cells could be ascribed to the type of immunosuppression, not in presence or absence of rejection. We found a slight decrease of Treg markers during CsA treatment in the animals presented in Figure 3, but this could not be confirmed in our larger sample collection (Tables 3 and 4).

In conclusion, our data suggest that antigen presentation takes place in nodular infiltrates. T cells respond to donor Ag by upregulating activation and regulation markers. Destructive T-cell responses take place outside the nodular infiltrates, with high numbers of macrophages and CD103+ cytotoxic T cells in the diffuse infiltrates and tubules. We conclude that inside the graft, none of the percentages of cells expressing the Treg markers CD25, CTLA-4, GITR or FOXP3 correlate with the acceptation of kidney allografts in our rhesus monkey model. The composition of nodular and diffuse infiltrates is influenced only very little by the type of immunosuppression, but largely by the presence or absence of rejection.


The authors thank H. van Westbroek for help with preparing the figures, Dr. E. Remarque for help with the statistical analyses and Dr. B't Hart and Prof. dr. F. Claas for critical review of the manuscript.