Blood samples were collected at the Department of Nephrology, University of Heidelberg, Germany. A total of 139 stable KTR with no signs of rejection and 41 KTR diagnosed with biopsy-confirmed acute renal graft rejection at the time of enrolment were sampled. Participants had undergone a kidney transplant at least three months prior to enrolment and had no recent illness or autoimmune disease. Exclusion criteria for stable KTR were suspected inflammation, suspected deterioration of graft function (serum creatinine > 2 mg/dl), or a history of malignancy. The group of participants with acute biopsy-confirmed graft rejection included patients with both T-cell-mediated and antibody-mediated rejection. Allograft biopsies at the time of blood collection were evaluated and diagnosed by the Institute of Pathology of the University of Heidelberg using semi-quantitative histological scores. Clinical acute rejection was defined as a rejection episode associated with graft dysfunction based on a greater than 30% increase in serum creatinine from baseline values and confirmed by pathological analysis of the biopsies according to the updated Banff classification.

We then followed the occurrence of acute renal graft rejection in the healthy group at a median of 2.1 (0.5 - 2.7) years after enrolment. Five patients developed acute renal graft rejection during this period. Table 1 shows the clinical data of all participants at enrolment. The remaining 134 stable KTR did not develop acute renal graft rejection during this period.

The Regional Ethics Committee approved the study (reference number S-523/2012, 05.07.2018). All participants were fully informed about the study settings and aims. Written informed consent was received from all individuals.

Nine ml of venous blood was collected into ethylenediaminetetraacetic acid (EDTA)-containing tubes. Peripheral blood mononuclear cells (PBMCs) were then isolated by density gradient centrifugation using Lymphodex (Inno-Train Diagnostik GmbH, Kronberg, Germany). PBMCs (8 x 10⁶) were then surface stained with fluorochrome conjugated undiluted monoclonal antibodies, directed against CD8, CD127, CCR7, CD45RA and CD31 (for further details, see Supplementary Table S1 ). After 20 minutes, PBMCs were washed twice with 3 ml phosphate-buffered saline (PBS) and centrifuged at 483 g for 5 minutes. Intracellular FoxP3 was detected using an anti-human FoxP3 staining kit (clone PCH101, eBioscience, Frankfurt, Germany), according to the manufacturer’s instructions. Negative control samples were incubated with isotype-matched antibodies. Cells were analyzed using a FACS Canto cytometer (BD Biosciences). Dead cells and doublets were excluded using forward and side scatter characteristics (FSC and SSC). Statistical analysis was based on a cell count of 100.000 CD8^pos T cells for each patient using an appropriate stopping gate.

Supplementary Figure S1 shows the gating strategy for these measurements. We determined the percentage of CD8^pos T cells of all lymphocytes and separated them into CD8^posCD127^{low pos/neg} FoxP3^pos Tregs and CD8^posCD127^posFoxP3^neg Tresps ( Supplementary Figures S1A–C ). We then subdivided both CD8^pos Tregs and Tresps into CCR7^pos CD45RA^pos naïve, CCR7^pos CD45RA^neg CM, CCR7^neg CD45RA^neg EM, and CCR7^neg CD45RA^pos TEMRA Tregs/Tresps ( Supplementary Figures S1D, F ) and in CD45RA^posCD31^pos RTE, CD45RA^posCD31^neg resting MN, CD45RA^negCD31^pos memory and CD45RA^negCD31^neg memory Tregs/Tresps ( Supplementary Figures S1E, G ). Finally, we merged these two differentiation schemes to distinguish newly released antigen-unexperienced CCR7^posCD45RA^posCD31^pos RTE cells (RTEs) from CCR7^negCD45RA^posCD31^pos TEMRA cells (CD31^pos TEMRAs) and CCR7^posCD45RA^posCD31^neg resting MN cells (resting MNs) from CCR7^negCD45RA^posCD31^neg TEMRA cells (CD31^neg TEMRAs).

This allowed us to identify three distinct differentiation pathways of CD8⁺ RTE T cells. First, they can differentiate into CD31^neg memory cells via CD31^pos memory cells (pathway 1) and could produce CD31^pos TEMRA cells when this pathway is exhausted. Second, RTE T cells can proliferate directly into CD31^neg memory cells (pathway 2) and could generate CD31^neg TEMRA cells when this proliferation is exhausted. Thirdly, RTE T cells can proliferate into resting MN cells, which may subsequently convert into CD31^neg memory cells or proliferate into CD31^neg memory cells (pathway 3). It appears that CM cells arise predominantly via pathway 1, whereas EM cells are more likely to arise via pathway 2. Resting MN cells may be able to convert or proliferate into CD31^neg memory cells, producing both CM and EM cells.

To characterize CD8^posFoxP3^posCD127^{low pos/neg} Tregs as CD8^posCD25^posCD127^{low pos/neg} Tregs or CD8^posFoxP3^posCD25^pos Tregs nine ml of venous blood was collected into ethylenediaminetetraacetic acid (EDTA)-containing tubes. Peripheral blood mononuclear cells (PBMCs) were then isolated by density gradient centrifugation using Lymphodex (Inno-Train Diagnostik GmbH, Kronberg, Germany). PBMCs (8 x 10⁶) were then surface stained with fluorochrome conjugated undiluted monoclonal antibodies, directed against CD8, CD127, CD25, and CD45RA (for further details, see Supplementary Table S1 ). After 20 minutes, PBMCs were washed twice with 3 ml phosphate-buffered saline (PBS) and centrifuged at 483 g for 5 minutes. Intracellular FoxP3 was detected using an anti-human FoxP3 staining set (clone PCH101, eBioscience) according to the manufacturer’s instructions. Negative control samples were incubated with isotype-matched antibodies. Cells were analyzed using a FACS Canto cytometer (BD Biosciences). Dead cells and doublets were excluded using forward and side scatter characteristics (FSC and SSC). Statistical analysis was based on a cell count of 100.000 CD8^pos Tcells using an appropriate stopping gate.

For functional analysis of CD8^posCD25^posCD127^{low pos/neg} Tregs, we first isolated CD8^pos T cells by Magnetic-Associated Cell Sorting (MACS). An average of 75 ml of venous blood was collected in EDTA-containing tubes from 3 different healthy control subjects. Approximately, 13.2 x 10⁷ PBMCs per subject were obtained by density gradient centrifugation using Lymphodex (Inno-Train Diagnostik GmbH, Kronberg, Germany). The CD8^pos T cells were purified using the CD8^pos T cell Isolation Kit human (Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer’s instructions. On average, 2.1 x 10⁷ CD8^pos T cells were isolated per subject. CD8^pos T cells were then surface stained with fluorochrome conjugated undiluted monoclonal antibodies, directed against CD8, CD127 and CD25 (for further details, see Supplementary Table S1 ). After 20 minutes, PBMCs were washed twice with 3 ml phosphate-buffered saline (PBS) and centrifuged with 483 g for 5 minutes. Afterwards, CD8^pos Tregs were isolated by cell sorting using a FACS Aria II cell sorter (BD Bioscience, Heidelberg, Germany). In all experiments, dead cells and doublets were excluded by FSC-A versus FSC-H gating, while the remaining cells were sorted into CD8^posCD25^posCD127^{low pos/neg} Tregs and the remaining CD8^pos Tresps. An average of 61.000 CD8^posCD25^posCD127^{low pos/neg} Tregs were received per subject.

To analyze the suppressive activity of the isolated CD8^pos Tregs, 2 × 10⁴ CD8^pos Tresps were co-cultured with the purified CD8^posCD25^posCD127^{low pos/neg} Tregs in a ratio of 1:1 to 1:128 in 96-well v-bottom plates. These suppression assays were performed in a final volume of 100 μl/well of X-VIVO15 medium (Lonza, Verviers, Belgium). For T-cell stimulation, the medium was supplemented with 1 μg/ml anti-CD3 (eBioscience, Frankfurt, Germany). As controls, CD8^posCD25^posCD127^{low pos/neg} Tregs and CD8^pos Tresps alone were cultured both with and without any stimulus. Cells were incubated at 37 °C in 5% CO₂. After 4 days, 1 μCi ³H-thymidine (Hartmann Analytic, Braunschweig, Germany) was added to the cultures and the cells were incubated for a further 16 h. The cells were then harvested and ³H incorporation was measured by scintillation counting. The maximum suppressive activity (ratio of Tregs to Tresps 1:1 or 1:2) and the minimum ratio of Tregs to Tresps at which at least 15% suppression could be achieved were calculated.

We used linear regression to examine changes in the percentage of CD8^pos T cells, Tregs, Tresps, and their subsets over the course of life with separate models for each patient group. Age-independent group differences were examined using multiple regression analysis adjusted for the age variable (centered on the mean), including an interaction term of the age and the patient group. A p-value < 0.05 was considered significant. However, this research is an exploratory study in which the calculated p-values are descriptive, but not confirmatory. BiAS for Windows (version 10.06) was used for all statistical tests.

Figure 1 shows the CD8^pos Treg cell differentiation of 134 stable KTR with stable allograft function, 5 KTR who developed allograft rejection during the follow-up period (up to a maximum of 3 years), and 41 patients with impaired graft function and consecutive indication biopsy, confirming graft rejection at the time of enrolment.

In stable KTR, the percentage of naïve Tregs within the total Treg pool decreased with age, whereas the percentage of CM and EM Tregs increased significantly ( Figures 1A, B, D ). In contrast, the percentage of TEMRA Tregs remained unchanged ( Figure 1C ). Further subdivision of the naive Treg pool into RTE Tregs, MN Tregs, CD31^pos and CD31^neg TEMRA Tregs showed a significant decrease in RTE Tregs ( Figure 1F ), but an increase in resting MN and CD31^neg memory Tregs ( Figures 1I, J ), whereas both CD31^pos and CD31^neg TEMRA Tregs as well as CD31^pos memory Tregs did not change ( Figures 1E, G, H ). Therefore, our results suggest that in KTR with stable allograft function, RTE Treg differentiate via CD31^pos memory Tregs (pathway 1) or proliferate into CD31^neg memory Tregs (pathway 2). Thus, both CM and EM Tregs are increasingly produced with age, with resting MN Tregs also being enriched, presumably as a naive reserve population ( Figure 2A ).

Compared to stable KTR, those who developed a rejection during the follow-up period showed a trend towards an increased proportion of CM Tregs within the total Treg pool independent of age ( Figure 1B ). This is probably due to an increased differentiation of RTE Tregs via CD31^pos memory Tregs into CD31^neg memory Tregs ( Figure 2B , pathway 1), as RTE Tregs are reduced, independently of age, whereas CD31^neg memory Tregs and resting MN Tregs are almost significantly increased ( Figures 1F, I, J ). A reduction in TEMRA Tregs, particularly those expressing CD31 ( Figure 1H ), may suggest enhanced proliferation of RTE Tregs (pathway 2) in EM Tregs, which appears to be markedly age-dependent ( Figure 1D ). However, due to the small number of patients in this group, significance was not reached. With age, naive Tregs tended to increase, TEMRA Tregs and CM Tregs remained relatively constant, while EM Tregs decreased almost significantly ( Figures 1A-D ). Further analysis of age-dependent changes within the total Treg pool showed a tendency for CD31^pos TEMRA Tregs as well as CD31^neg memory Tregs to decrease ( Figures 1E, J ), while all other subsets remained constant ( Figures 1F-H ), except for resting MN Tregs, which increased significantly with age ( Figure 1I ). Apparently, in these patients, RTE Treg differentiation via CD31^pos memory Tregs into CM Tregs (pathway 1) is preserved with age, rather than their proliferation into EM Tregs (pathway 2), with CD31^neg memory Tregs likely to be less enriched than MN Tregs ( Figure 2A ).

In KTR with biopsy-proven rejection at the time of enrolment, the percentage of naïve Tregs in the total Treg pool was significantly reduced regardless of age, while that of TEMRA Tregs was significantly increased ( Figures 1A, C ). The percentage of EM Tregs also tended to be increased ( Figure 1D ) whereas that of CM Tregs remained unchanged compared to stable KTR ( Figure 1B ). Further subdivision of the naive Treg pool into RTE Tregs, MN Tregs, CD31^pos and CD31^neg TEMRA Tregs showed an almost significant decrease of RTE Tregs and a significant decrease of MN Tregs ( Figures 1F, I ), accompanied with a significant increase in CD31^pos memory Tregs and CD31^neg TEMRA Tregs ( Figures 1G, H ), while CD31^pos TEMRA Tregs and CD31^neg memory Tregs remained unchanged within the total Tregs ( Figures 1E, J ). Therefore, it seems that the exhaustion of RTE Treg differentiation via CD31^pos memory Tregs (pathway 1) and via their direct proliferation (pathway 2) was compensated by an increased differentiation of resting MN Tregs into CD31^neg memory Tregs, probably more by proliferation in EM Tregs than by conversion in CM Tregs (pathway 3), ( Figure 2B ). Thus, the proportion of CM Tregs within the total Treg pool was maintained, whereas that of EM Tregs was rather increased compared to stable KTR ( Figures 1B, D ). Compared to KTR with future rejection, CM Tregs decreased in KTR with current rejection ( Figure 1B ), also suggesting that resting MN Tregs are more likely to proliferate into CD31^neg memory Tregs ( Figure 2B ).

With age, these patients seemed to accumulate CD31^pos TEMRA Tregs and produced significantly fewer CD31^pos memory Tregs ( Figures 1E, G ), suggesting that RTE Tregs lose their ability to differentiate into CD31^neg memory Tregs via pathway 1. Furthermore, RTE Tregs decreased with age, whereas resting MN Tregs and CD31^neg memory Tregs increased almost significantly ( Figures 1F, I, J ). Therefore, it appears that only the ongoing differentiation of resting MN Tregs replenishes the CD31^neg memory Treg pool ( Figure 2A ). However, this means that the alternative differentiation of resting MN Tregs (pathway 3) in these patients cannot generate CM and EM Tregs as effectively with age as in stable KTR ( Figures 1B, D ).

Figure 3 shows the differences in CD8⁺ Tresp cell differentiation between the three patient groups. In stable KTR, there appeared to be an increased differentiation of naïve Tresps into TEMRA Tresps with age, as naïve Tresps decreased and TEMRA Tresps increased significantly, whereas CM and EM Tresps were maintained but did not increase significantly ( Figures 3A-D ). As both CD31^pos and CD31^neg TEMRA Tresps increased significantly with age ( Figures 3E, H ), pathways 1 and 2 appeared to be exhausted with age, so that RTE Tresps already differentiate via resting MNs into CD31^neg memory Tresps (pathway 3), replenishing the CM and EM Tresp pool ( Figure 2C ).

Compared to stable KTR, neither those who developed future rejection nor those who were diagnosed with current rejection showed age-independent differences in the differentiation of CD8^pos Tresps ( Figures 3A-J , 2D ). Only with age, patients with future rejection showed an increasing differentiation of RTE Tresps into CM Tresps ( Figure 3B ), presumably due to increased conversion of resting MN Tresps into CD31^neg memory Tresps ( Figure 2C ), whereas KTR with current rejection showed an increasing differentiation into EM Tresps ( Figure 3D ), presumably due to increased proliferation of resting MN Tresps into CD31^neg memory Tresps ( Figure 2C ).

We then investigated whether these differences in the differentiation of CD8^pos Tregs between the three patient groups had an impact on the composition of the total CD8^pos T cell pool with CD8^pos Tregs and CD8⁺ Tresps, their ratio to each other, and the proportion of total CD8⁺ T cells in all lymphocytes. In stable KTR, there was a significant increase in CD8⁺ Tregs with a decrease in Tresps and thus an increasing CD8⁺ Treg/Tresp ratio with age ( Figures 4A-C ), presumably due to an age-related stronger production of CD31^neg memory Tregs ( Figure 1J ) than CD31^neg memory Tresps ( Figure 3J ). In KTR with biopsy-proven rejection the same age-related changes were found, presumably because in these patients, the age-dependent differentiation that normally occurs in healthy individuals can be adequately replaced by differentiation of resting MN Tregs/Tresps.

Regardless of age, KTR developing future rejection showed a significantly increased percentage of CD8^pos Tregs and a complementary decreased percentage of CD8^pos Tresps, resulting in an increased Treg/Tresp ratio compared to stable KTR ( Figures 4A-C ). The percentage of total CD8^pos T cells did not change between the three groups ( Figure 4D ), indicating that there was indeed an increase in Tregs but decrease in Tresps.

To demonstrate that CD8^posFoxP3^posCD127^{low pos/neg} Tregs have suppressive capabilities, we used FACS analysis to show that these cells can also be reliably detected as CD8^posFoxP3^posCD25^pos Tregs or CD8^posCD25^posCD127^{low pos/neg} Tregs. In our experiments, we found an overlap of 100% for CD8^posFoxP3^posCD25^pos Tregs and 97.1% for CD8^posCD25^posCD127^{low pos/neg} Tregs with CD8^posFoxP3^posCD127^{low pos/neg} Tregs ( Figure 5 ). These analyses demonstrate the possibility of detecting CD8^pos Tregs without detecting the FoxP3 marker, which is an intracellularly expressed transcription factor, that requires cell fixation for intracellular staining. Therefore, we first isolated CD8^pos T cells by MACS technology and then used cell sorting to separate CD8^posCD25⁺CD127^{low pos/neg} Tregs from CD8^pos Tresps ( Figure 6A ).

To demonstrate the suppressive activity of CD8^posCD25^posCD127l^{ow pos/neg} Tregs we used suppression assays, in which CD8^posCD25^posCD127^{low pos/neg} Tregs and CD8^pos Tresps were co-cultured at ratios of 1:1 - 1:128. Maximum suppressive activity was calculated at a Treg/Tresp ratio of 1:1 or 1:2. In addition, the maximum titer of CD8^posCD25^posCD127^{low pos/neg} Tregs was determined at which a minimum suppressive activity of 15% was achieved. Figure 6B shows these results for three healthy controls. These experiments were performed twice in one individual (healthy control 1) to ensure reproducibility. Our results show that CD8^posCD25^posCD127^{low pos/neg} Tregs from healthy control 1 suppressed the proliferation of CD8^pos Tresps with a maximum suppressive activity of 29.8% and 21.1%, respectively, while the other individuals reached 52.8% and 83.8%, respectively. The titer up to which a minimum suppressive activity of 15% could be achieved was 1:16 in both experiments for healthy control 1 and 1:64 and 1:16 for healthy controls 2 and 3, respectively ( Figure 6B ).

Summarizing, our results reveal that in patients at higher risk of future rejection, RTE Tregs show enhanced differentiation into MN Tregs and CM-enriched CD31^neg memory Tregs. Conversely, in recipients with rejection-related graft impairment, this differentiation process appears to be exhausted, instead favoring the proliferation of MN Tregs into EM-enriched CD31^neg memory Tregs. This leads to an accumulation of CD31^neg TEMRA Tregs, indicating a loss of the proliferative capacity of RTE Tregs during rejection events.