Naive regulatory T cells: a novel subpopulation defined by resistance toward CD95L-mediated cell death

There is now clear evidence that CD4⁺ T cells contain a population of naturally immunosuppressive T cells characterized by constitutive expression of CD25, CTLA-4, and FOXP3.¹  Because not all potentially autoreactive T cells are deleted in the thymus,^2,3  peripheral control of T-cell responses by naturally occurring CD4⁺CD25⁺FOXP3⁺ immunoregulatory T cells (T_regs) is crucial to prevent autoimmunity.¹  Depletion of T_regs contributes to the induction of severe autoimmune diseases in animal models, and several studies have reported a defect of T_regs in various human autoimmune diseases.^1,4,5 

Despite extensive research to unravel the immunosuppressive function of T_reg, the exact molecular mechanism of immunosuppression is still elusive. Promising targets for pharmacologic mimicking of T_regs function remain undefined. Consequently, direct use of T_regs for therapy is currently under examination. Therapeutic expansion or depletion of T_regs with defined antigen specificity offers new treatment options for human diseases.⁶  Because accumulation of T_regs has been shown to be detrimental in cancer,⁷  new insights into mechanisms of T_reg homeostastis are required.

To gain new insight into the modulation of T_reg numbers as well as their antigen specificity, the development of natural T_regs during ontogeny and in the adult needs to be explored. The peripheral T_reg compartment consists mainly of thymus-derived T_regs.¹  However, under specific circumstances T_regs can also be generated out of conventional T cells (T_convs) (eg, if the antigen is targeted to immature dendritic cells [DCs]).^8-10  It has yet to be established how much “converted” T_regs contribute to the total number of T_regs in the periphery of adult mice and humans.

At a given time, the overall number of T_regs is defined by a balance of generation and demise of T_regs. At the end of an immune reaction, T cells are depleted by apoptosis. Activation-induced cell death (AICD) through CD95/CD95L has been described as such an apoptosis-inducing mechanism.¹¹  While naive T cells are CD95^– and resistant to apoptosis induction, activated T cells (CD45RO^hi) up-regulate CD95 and become sensitive to apoptosis. Upon T-cell receptor (TCR) restimulation, activated effector T cells (T_effs) up-regulate CD95L and induce AICD through crosslinking of CD95 by CD95L. AICD eliminates T_effs after an immune response and contributes to T-cell homeostasis. We have previously shown that most T_regs constitutively express CD95 and can easily be killed via crosslinking of this death receptor by CD95L.¹²  Together with the fact that most T_regs express high levels of CD45RO, we hypothesized that most T_regs resemble memory/recently activated T cells rather than naive T cells. Only very recently has the presence of naive T_regs of CD4⁺CD25⁺CD45RA^hi phenotype in addition to memory T_regs been appreciated in peripheral blood.^13-15  However, because naive T_regs express less CD25 (CD25⁺) than memory T_regs (CD25^hi) and because those studies did not include fluorescence-activated cell sorting (FACS) analysis of FOXP3 expression, a clear separation between CD25⁺ T_regs and CD25⁺ T_effs was not possible. Therefore, analysis of CD4⁺CD25⁺CD45RA^hi T cells might have included contaminating T_effs. Moreover, no functional differences between naive and memory T_regs were described; both cell types were reported to be immunosuppressive.^14,15  In contrast, memory T_convs are highly differentiated cells with distinct biologic properties from those of naive T_convs. Analagously, naive and antigen-primed/memory T_regs should exhibit different functional features.

Here we analyzed CD95 expression levels and apoptosis sensitivity of T_reg subpopulations. We showed that naive T_regs are resistant to CD95L-mediated apoptosis in contrast to memory T_regs. First, gene microarray analysis of CD95L-mediated apoptosis-resistant T_regs (rT_regs) revealed a naive T-cell signature of rT_regs. A 6-color FACS analysis was performed to detect FOXP3 and typical memory/naive T-cell markers on T_regs from adult peripheral blood. Naive T_regs constitute only 0.2% to 0.8% of total human CD4 cells, whereas in cord blood most T_regs belong to this naive apoptosis-resistant subpopulation. Short-term TCR stimulation in vitro converted these naive T_regs into CD95L-sensitive T_regs similar to adult memory (CD4⁺CD25^hi) T_regs. In vivo, recognition of self antigens by T_regs might lead to the conversion of most T_regs into memory cells, and previous studies focused on CD25^hi memory T_regs to avoid T_eff contamination of the T_reg pool. However, under pathogenic conditions like human autoimmunity we expect alterations of adult memory T_reg frequencies (eg, due to enhanced CD95L-mediated apoptosis as reported for [CD4⁺CD25^hi] T_regs derived from systemic lupus erythematosus [SLE] patients).¹⁶  In such situations, the analysis of the so far neglected naive T_reg pool might be of great importance.

Peripheral blood was obtained from healthy adult donors. Cord blood samples were obtained from the placentas of healthy full-term newborns delivered at the Department of Gynaecology and Obstetrics at the University of Heidelberg. All samples were obtained after informed consent and approval by the University of Heidelberg ethics committee according to the Declaration of Helsinki.

The monoclonal antibodies (mAbs) against human CD4, CD45RO, CD45RA, and CD31 were obtained from BD Pharmingen (Heidelberg, Germany) and αCD25 Ab from Miltenyi Biotec (Auburn, CA). CD95L was produced as a Leucin zipper-tagged ligand of CD95, which binds both murine and human CD95.¹⁷  The monoclonal αCD3 Ab (OKT3), αCD28 mAb, and the agonistic monoclonal αCD95 Ab (anti–Apo-1) were purified from hybridoma supernatants by protein A affinity purification as described.¹⁷  The monoclonal αFoxP3 mAb (clone PCH101) was obtained from eBioscience (San Diego, CA), annexin V Alexa Fluor 488 from Molecular Probes (Eugene, OR), and propidium iodide (PI) and protein A from Sigma-Aldrich (St Louis, MO).

Human peripheral blood lymphocytes (PBLs) were purified from peripheral blood by Biocoll (Biochrom, Berlin, Germany) gradient centrifugation followed by plastic adherence to deplete monocytes. Adult peripheral CD4⁺CD25⁺ cells were first enriched using magnetic-activated cell separation (MACS) beads (Miltenyi Biotec), and subsequently CD4⁺CD25^high cells were sorted with a FACS-Diva cell sorter (BD Biosciences, San Jose, CA). CD4⁺CD25^hi T_regs cells contained only the 1% to 2% of human CD4 cells with the highest CD25 expression, as previously reported.¹⁸  Similarly, CD4⁺CD25⁺ or CD4⁺CD25⁺CD45RO^lo/hi T cells were FACS sorted from cord blood–derived lymphocytes. FACS-sorted CD4⁺CD25^– cells were used as T_convs.

Dead cells were depleted using the MACS annexin V MicroBead kit (Miltenyi Biotec). Briefly, cells were incubated with annexin V microbeads and passed over a medium-size (MS) column to magnetically remove annexin V–positive cells. Purity of viable cells was determined by forward scatter/sideward scatter (FSC/SSC) analysis and propidium iodide staining.

Freshly isolated human T cells were cultured in IL-2 (100 IU) containing ex Vivo-15 medium (Cambrex, Verviers, Belgium) supplemented with 1% glutamine (Invitrogen, Karlsruhe, Germany). For apoptosis induction, T cells were stimulated with αCD95 Ab and 10 ng/mL protein A or different concentrations of CD95L (maximal 1:20 dilution) as previously described.¹²  Unstimulated cells were incubated with an isotype control Ab or CD95L-free control medium. To expand T_regs, cells were incubated with 0.1 μg/mL αCD3 Ab and 1 μg/mL αCD28 Ab in combination with irradiated JY feeder cells (kind gift from C. Falk, GSF-Institute for Molecular Biology, Munich, Germany) and 300 IU IL-2.¹²  Cell death was assessed by annexin V/propidium iodide costaining and forward to sideward scatter profiles. Specific cell death was calculated as described previously.¹² 

When expanded T_regs were used, they were washed extensively to remove IL-2, but no further resting of T_regs occurred prior to cell proliferation assays. T_convs (1 × 10⁴) were incubated in 96-well plates (Nunc, Wiesbaden, Germany) alone or in coculture with T_regs (1 × 10⁴ T_regs) together with irradiated T-cell–depleted PBMCs (1 × 10⁵) and were stimulated with soluble αCD3 mAb (0.2 μg/mL) and αCD28 mAb (0.005 μg/mL). Antigen-presenting cells (APCs) and T_convs were always derived from the same blood sample. After 4 days at 37°C in 5% CO₂, 1 μCi (0.037 MBq) [³H]-thymidine per well was added for additional 16 hours, and proliferation was measured in counts per minute (cpm) with a scintillation counter. Inhibitiory capacity (percentage) of T_regs in coculture experiments was defined as (1 – ³[H]-thymidine uptake [cpm] of coculture of T_regs with T_convs divided by cpm of T_convs alone) × 100%. Triplicate wells were used in all suppression experiments.

FOXP3 staining was performed according to the manufacturer's protocol. For 6-color FACS, cells were first stained with the surface mAb of interest followed by FOXP3 intracellular staining. FACS acquisition was performed immediately with a FACS Canto cytometer and analyzed with FACS-Diva software (BD Biosciences).

Total RNA was isolated using the Absolutely RNA Microprep kit (Stratagene, Heidelberg, Germany), and cDNA was prepared using random oligo(dT) primers (Invitrogen). The FOXP3 message was quantified by detection of incorporated SYBR Green using the ABI Prism 5700 sequence detector system (Applied Biosystems, Foster City, CA). The relative expression level was determined by normalization to GAPDH, and results are presented as fold expresssion of T_conv mRNA levels. FOXP3 primer sequences were 5′-AGCTGGAGTTCCGCAAGAAAC (forward) and 5′-TGTTCGTCCATCCTCCTTTCC (reverse). GAPDH primer sequences were 5′-GCAAATTCCATGGCACCGT (forward) and 5′-TCGCCCCACTTGATTTTGG (reverse).

Quality and integrity of 500 ng total RNA were controlled by running all samples on a 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). For RNA amplification, the first round was done according to Affymetrix without biotinylated nucleotides using the P1300 RiboMax Kit (Promega, Mannheim, Germany) for T7 amplification. For the second round of amplification, the precipitated and cleaned aRNA was converted to cDNA using random hexamers (Pharmacia, Freiburg, Germany). Secondstrand synthesis and probe amplification was performed as in the first round, with 2 exceptions: There was an incubation with RNAse H before first-strand synthesis to digest the aRNA, and T7T23V oligo was used for initiation of the synthesis of the second strand. The concentration of biotin-labeled cRNA was determined by UV absorbance. In all cases,12.5 μg of each biotinylated cRNA preparation was fragmented and placed in a hybridization cocktail containing 4 biotinylated hybridization controls (BioB, BioC, BioD, and Cre) as recommended by the manufacturer. Samples were hybridized to an identical lot of Affymetrix HG-U133A 2.0 for 16 hours. After hybridization the GeneChips were washed, stained with streptavidin-phycoerythrin (SA-PE), and read using an Affymetrix Gene-Chip fluidic station and scanner.

Analysis of microarray data was performed using the Affymetrix GCOS 1.2 software. For global normalization, all array experiments were scaled to a target intensity of 150, otherwise using the default values of the GCOS 1.2 software. Filtering of the results was done as follows: Genes were considered as regulated when their fold change was greater than or equal to 1.2 or less than or equal to –1.2 and the statistical parameter for a significant change (change call) was not NC (no change). The entire Minimum Information About a Microarray Experiment (MIAME)–compliant data set of the microarray experiments will be posted at the Gene Expression Omnibus (GEO) database.¹⁹ 

We have previously shown that most CD4⁺CD25^hi regulatory T cells (T_regs) are highly susceptible to CD95L-mediated cell death in contrast to conventional CD4⁺25^lo T cells (T_convs).¹²  Interestingly, a subpopulation of T_regs remained consistently apoptosis resistant to the treatment with CD95L. This apoptosis-resistant population (rT_regs) represents a minor subpopulation (10% to 30%) within the sorted CD4⁺CD25^hi T_regs.¹²  To further analyze this population, FACS-sorted CD4⁺CD25^hi T_regs from multiple donors were pooled and incubated with CD95L. Dead T_regs were removed by annexin V microbeads, and the cell-death–resistant rT_regs were recovered (Figure 1A). During a second challenge with CD95L (Figure 1B) or αCD95 mAb (data not shown) these rT_regs remained apoptosis resistant, confirming that they display a stable phenotype. As a control, CD4⁺CD25^hi T_regs were left without CD95L, incubated with annexin V microbeads, passed over MS columns, washed, and subsequently incubated with CD95L. As expected, specific cell death for these total CD4⁺CD25^hi T_regs remained high (Figure 1B).

To further characterize rT_regs, we pooled rT_reg cDNA derived from FACS-sorted CD4⁺CD25^hi T_regs of 8 healthy blood donors and performed gene chip microarray analysis. In this screen, we compared cDNA from rT_regs with cDNA derived from all CD4⁺CD25^hi T_regs. Interestingly, rT_regs showed a significant decrease of mRNA expression of genes related to T-cell memory like CTLA-4, various HLA genes, and CD103, whereas mRNA for molecules expressed mainly on naive cells such as CCR7 or CD127 (IL7R) was more prevalent in rT_regs than in total CD4⁺CD25^hi T_regs (Tables S1-S2, available on the Blood website; see the Supplemental Tables link at the top of the online article). This finding suggested that rT_regs are more naive in contrast to most T_regs found in the CD4⁺CD25^hi T-cell population, which are known to exhibit an antigen-primed/memory-like phenotype. To test this hypothesis, we incubated FACS-sorted CD4⁺CD25^hi T_regs with increasing concentrations of CD95L. The remaining viable rT_regs were then analyzed by 4-color FACS for the expression of the typical memory marker CD45RO ²⁰  (Figure 1C-E) and the naive marker CD31 ^21,22  (data not shown). In contrast to T_regs cultured in control medium (Figure 1C), rT_regs surviving CD95L stimulation were mostly CD45RO^lo (more than 85%) (Figure 1D). The number of CD45RO^hi cells progressively declined with increasing concentrations of CD95L (Figure 1E). The naive cell marker CD31 increased correspondingly on rT_regs (data not shown). We conclude that CD95L preferentially kills CD45RO^hi T_regs cells, leaving behind rT_regs with a naive phenotype and slightly lower CD25 expression levels. Furthermore, FACS analysis for intracellular FOXP3 expression demonstrated that all rT_regs were FOXP3⁺, and the anergy and suppressive capacity of in vitro–expanded rT_regs (data not shown) further supported their classification as T_regs.

Consequently, human T_regs should comprise a heterogeneous population with memory and naive T_regs. To prove that naive CD4⁺CD25^hi T cells are indeed T_regs, we performed FOXP3 FACS analysis in PBLs. In the absence of further highly specific T_reg markers, FOXP3 is considered to be the most reliable marker to identify T_regs in human peripheral blood to date.²³  Therefore, we established 6-color FACS analysis including mAbs against CD4, CD25, and FOXP3 for T_reg detection together with mAbs against CD45RO and CD45RA to distinguish memory and naive T_regs. Naive T_regs were initially considered as CD4⁺FOXP3⁺CD45RO^lo cells (R2 in Figure 2A, right panel), and memory T_regs were characterized as CD4⁺FOXP3⁺CD45RO^hi cells (R3 in Figure 2A, right panel). Using multicolor analysis one can ascertain that memory T_regs (blue) are mainly CD25^hi while naive T_regs (red) are mainly CD25^int and overlap with CD25^int T_effs. Accordingly, pure FACS-based isolation of T_regs is limited to the CD25^hi memory T_reg-enriched cell pool, because surface expression of CD25 is quite similar between naive T_regs and FOXP3^– CD4⁺CD25⁺ T_effs (Figure 2B). Naive T_regs showed a higher expression of CD45RA than memory T_regs (Figure 2C). Analysis of the naive cell marker CD31 also showed increased expression (data not shown). Comparison of CD95L-resistant cells in Figure 1D with the CD45RO^lo naive T_regs (red) from the 6-color staining (Figure 2B-C) revealed a similar phenotype, supporting that apoptosis-resistant T_regs represent naive T_regs in adult peripheral blood. Because sensitization toward CD95L-mediated cell death includes up-regulation of CD95, we compared CD95 expression of naive and memory T_regs by 6-color FACS for expression of CD4, CD25, FOXP3, CD45RO, CD45RA, and CD95 (Figure 2D). Indeed, naive CD4⁺CD25⁺FOXP3⁺CD45RO^lo T_regs (red) expressed very low levels of CD95 in comparison with the highly CD95⁺ memory T_regs (CD4⁺CD25⁺FOXP3⁺CD45RO^hi, blue, Figure 2D). Collectively, we define naive T_regs as CD4⁺CD25⁺FOXP3⁺CD45RO^loCD95^lo T cells and memory T_regs as CD4⁺CD25⁺FOXP3⁺CD45RO^hiCD95^hi T cells. Furthermore, we conclude that rT_regs represent naive T_regs in adult peripheral blood. The frequency of naive T_regs varied between 10% to 50% of total T_regs in adult peripheral blood and clearly declined with increasing age (Figure 2E).

Whereas the frequency of CD45RO^lo T_regs in adult peripheral blood is rather low, newborns should exhibit high numbers of naive T_regs. Therefore, we extended our functional analysis on cord-blood–derived T_regs. Nearly all CD4⁺CD25⁺ human cord blood T cells expressed FOXP3 as determined by FACS analysis (Figure 3A) and quantitative reverse transcriptase–polymerase chain reaction (RT-PCR) (data not shown). The percentage of FOXP3⁺ T cells within the CD4⁺ compartment was only slightly higher in cord blood (5.9% ± 3.1%, n = 6) than in adult peripheral blood (4.8% ± 1.7%, n = 6). T_regs were sorted from adult and cord blood according to CD25 expression as depicted in Figure 3B and showed significant suppressive capacity in vitro (Figure 3C). Thus, all CD4⁺CD25⁺ T cells in human cord blood constituted CD4⁺CD25⁺FOXP3⁺ T_regs similar to murine T_regs. In contrast, few CD4⁺CD25⁺ T cells in adult peripheral blood contained CD4⁺CD25⁺FOXP3⁺ T_regs (Figure 3A). Further, FACS analysis revealed that most cord blood T_regs showed low expression of CD45RO (Figure 4B) but high expression of CD45RA and CD31 (data not shown), confirming their naive status. Interestingly, most cord blood T_regs showed a low expression of CD95 (Figure 3D) similar to adult naive T_regs. High expression of CD95 was limited to the minor subpopulation of CD45RO^hi T_regs in cord blood (Figure 4B, arrow). To summarize, both naive T_regs from cord blood as well as naive T_regs from adult peripheral blood show a lower expression of CD95 than memory T_regs. In addition, both exhibit a similar level of CD25 expression (compare Figures 2B and 3A), further indicating that cord blood T_regs and adult naive T_regs represent a similar cell population.

Next we tested whether cord-blood–derived T_regs were resistant to CD95L-mediated apoptosis similar to adult naive T_regs. In contrast to freshly isolated adult CD4⁺CD25^hi T_regs, cord blood T_regs were nearly resistant toward CD95L-mediated cell death (Figure 4A). This correlates with the CD95^hi phenotype of memory T_regs in the adult CD4⁺CD25^hi T_reg fraction (Figure 2D) and the rather low presence of such memory T_regs among freshly isolated cord blood T_regs (Figure 4B, arrow). Interestingly, the few CD45RO^hi T_regs present in cord blood showed not only increased CD95 expression but also slightly enhanced sensitivity to CD95L-mediated apoptosis (Figure 4A, right).

During an immune response, naive T_convs are stimulated via their TCR. They clonally expand and differentiate into T_effs. One hallmark of peripheral T-cell depletion is that T_effs up-regulate CD95 and become sensitive to CD95L-mediated cell death to allow depletion of the expanded T-cell pool by CD95L.¹¹  We have shown that TCR stimulation of T_convs from adult peripheral blood with anti-CD3 mAb and IL-2 can be used to mimic this critical process in vitro.¹¹  Accordingly, we questioned whether naive T_regs could also up-regulate CD95 and be sensitized to CD95L-mediated apoptosis in response to TCR stimulation.

For this purpose, we stimulated freshly isolated cord blood T_regs with anti-CD3/CD28 mAbs and IL-2 and analyzed their phenotypes at various time points (Figure 5). Soon after TCR stimulation, CD95 was up-regulated on cord blood T_regs and reached a CD95 expression plateau after 3 to 4 days (Figure 5A-B). This was notable because adult human T cells take 5 to 6 days to reach a plateau of CD95 expression (data not shown). However, murine T cells were also sensitized toward CD95L faster than adult T cells after TCR stimulation,²⁴  which suggests that naive human cord blood T cells share similarities with murine T cells. Simultaneous FACS analysis of memory cell markers during the in vitro stimulation of cord blood T_regs revealed that CD45RO increased more gradually than CD95 (Figure 5B) and that CD45RA and CD31 were gradually lost until day 6 (data not shown). Thus, this in vitro–induced phenotype of cord blood T_regs (CD95^hiCD45RO^hiCD45RA^lo) was identical with the phenotype of most adult CD4⁺CD25^hi T_regs in vivo (Figure 2).

Most importantly, prestimulated cord blood T_regs were sensitive to CD95L treatment in contrast to freshly isolated cord blood T_regs (Figure 5C). To rule out outgrowth of a few preexisting cord blood CD45RO^hiCD95^hi T_regs, we aimed to exclude CD45RO^hiCD95^hi T_regs by FACS sorting. However, FACS sorting with anti-CD95 mAb could induce isolation-based apoptosis. Therefore, we defined an expression level of CD45RO that allowed separation of CD95^hi from CD95^lo cells without costaining for CD95 (dotted line in Figure 4B). This was performed using an aliquot of the FACS-sorting cells stained for CD45RO and CD95. After defining the CD45RO “cutoff,” the sorting cells were separated into CD45RO^hi (correspondingly CD95^hi) cells (Figure 4B, R1) and CD45RO^lo (correspondingly CD95^lo) cells (Figure 4B, R2). When we stimulated the CD45RO^loCD95^lo cord blood T_regs for 6 days, they still up-regulated CD95 and became sensitive toward CD95L-mediated apoptosis (Figure 5C). Therefore, outgrowth of CD45RO^hiCD95^hi T_regs could be ruled out.

Likewise, isolated rT_regs from adult blood up-regulated CD95 in response to TCR stimulation (anti-CD3/CD28 mAbs plus IL-2) and became sensitive to CD95L-mediated cell death (Figure 5E-F). Finally, freshly isolated adult CD4⁺CD25^hi T_regs, naive cord blood T_regs, and prestimulated cord blood T_regs showed similar immunosuppressive capacity in vitro (Figures 3C and 5D). In contrast, up-regulation of CD95 and sensitization toward CD95L-mediated cell death clearly distinguished memory T_regs from naive T_regs on a functional basis.

In our efforts to characterize CD95L apoptosis-resistant T_regs (rT_regs) within the population of adult CD4⁺CD25^hi T_regs, we found an increased expression of molecules defining naive T cells and a decreased expression of memory cell–associated molecules by microarray analysis. Further FACS analysis confirmed this finding and revealed that naive T_regs from adult and cord blood express low levels of CD95 and are resistant to CD95L-mediated cell death. This is in contrast to in vitro–prestimulated naive T_regs or in vivo–activated CD4⁺CD25^hiCD45RO^hi T_regs, which express high levels of CD95 and are highly sensitive to death receptor ligation. The number of T_regs in the organism is tightly regulated, and we propose that elimination of T_regs after an immune response contributes to the low frequency of T_regs in the adult CD4⁺ fraction.

CD95L-mediated apoptosis of T cells has mainly been studied in naive T_convs that were activated in vitro. Analysis of apoptosis sensitivity of in vivo–primed/memory T_effs has been hampered by their low frequencies in the healthy human individual. By contrast, continuous in vivo stimulation of naive T_regs with self antigens seems to keep the proportion of CD95^hi T_regs relatively high. In fact, incubation of naive T_regs with autologous APCs and IL-2 has been shown to stimulate naive T_regs in contrast to unreactive naive T_convs,¹⁴  and Romagnoli and coworkers have shown a high degree of self-reactivity within the T_reg population.²⁵  Thus, a constant activation of T_regs in vivo could conceivably lead to a very high proportion of memory T_regs in adults. This high frequency of memory T_regs allowed us to demonstrate the high apoptosis sensitivity toward CD95L of these antigen-experienced T_regs with a memory-like phenotype directly ex vivo.

Whereas naive cord blood T_regs are CD95^lo similar to naive T_convs, TCR stimulation in vitro revealed a rapid up-regulation of CD95. Simultaneously, CD45RO was up-regulated, whereas expression of the naive marker molecule CD31 declined in slower kinetics (data not shown). Six days after TCR stimulation in vitro, most cord-blood–derived naive T_regs had acquired the phenotype (CD95^hiCD45RO^hiCD45RA^lo) known from adult CD4⁺CD25^hi T_regs in peripheral blood and lymphoid tissue in vivo.¹⁵  In summary, naive T_regs show the same phenotype in cord blood and adult peripheral blood but are much more frequent in cord blood. By contrast, memory T_regs predominate in adult peripheral blood and are rare in cord blood. However, short-time TCR stimulation in vitro transformed cord-blood–derived naive T_regs into T_regs of a memory-like phenotype presented by most adult CD4⁺CD25^hi T_regs.

This in vitro induction of memory T_regs might mimic the interaction of naive T_regs with self antigens in vivo, which generates a pool of self-antigen–primed/memory-like T_regs very early in fetal ontogeny. T_regs can be detected as early as 14 weeks of human gestation in the periphery,²⁶  and T_regs in fetal lymph nodes (less than 17 weeks old) already show a higher expression of CD95 and CD45RO in comparison with T_convs.²⁷  Because the healthy fetus is usually not challenged with foreign antigens, activation of T_regs might reflect the early interaction of T_regs with self antigens in the pathogen-free organism. Interestingly, T_regs in the fetal and adult thymus express lower levels of CD25 and are of naive phenotype, while T_regs in the fetal/adult lymphoid organs as well as in the adult peripheral blood express higher levels of CD25 (CD25^hi) and are of memory phenotype (eg, CD95^hi, CD45RO).^15,27  Thus, after their emigration from the thymus, T_regs recognize self antigen in the periphery and convert to memory T_regs. During adult life ongoing self-antigen–specific activation of naive T_regs might trigger the immunosuppressive function of T_regs and thereby continuously supply the pool of memory CD4⁺CD25^hi T_regs. Up-regulation of CD95 and high sensitivity toward CD95L-mediated cell death would then allow for constant elimination of antigen-experienced T_regs. Both the high turnover of mouse self-specific T_regs^28-30  and the larger proportion of activated T_regs than T_convs in the human periphery²⁶  strongly support this model.

Previously, only CD4⁺CD25^hi T cells have been accepted to represent human T_regs, and many studies have focused on this population because pure isolation of FOXP⁺ T_regs is limited to this population in humans. With the availability of specific FOXP3 mAbs for flow cytometry, we now detected naive T_regs with lower CD25 expression in adult peripheral blood. This confirmed recent reports that suggested the presence of naive T_regs within the CD4⁺CD25⁺CD45RA^hi T-cell population of adult peripheral blood.^14,15  However, our multicolor FOXP3-FACS analysis also revealed that CD4⁺CD25⁺CD45RA^hi T cells consist of both FOXP3⁺ and FOXP3^– T cells (data not shown). Therefore, FACS sorting for CD4⁺CD25⁺CD45RA^hi T cells from adult peripheral blood yields a significant contamination with FOXP3^– T_convs. In fact, these contaminating T_convs might contribute to the reported relatively high proliferative potential of adult naive T_regs in contrast to the profound anergy of CD4⁺CD25^hi T_regs.¹⁴ 

It has long been appreciated that cord blood contains a distinct population of CD25⁺CD45RA^hi T_regs with suppressive function. Although PCR analysis of these CD25⁺CD45RA^hi cells has revealed FOXP3 expression,^31-33  it remained unclear whether all cord blood CD25⁺ cells express FOXP3 and are T_regs. Our flow cytometric analysis clearly established that essentially all cord blood CD25⁺ cells coexpress FOXP3. Thus, FOXP3⁺ cells can easily be isolated from cord blood on the basis of their CD25 expression. Cord-blood–derived T_regs are immunosuppressive and show proliferative anergy like adult CD4⁺CD25^hi T_regs.^15,31,33,34  Most importantly, expression of FOXP3 seems to be more specific for naive T_regs than for recently activated T_regs. Activation-induced up-regulation of FOXP3 mRNA has been reported for in vitro–stimulated T_effs and triggered a controversial discussion on FOXP3 specificity for human T_regs.^35-38  Interestingly, we also observe a slightly higher FOXP3 expression in memory T_regs than in naive T_regs, which might reflect the recent activation of memory T_regs by self antigen in vivo. However, naive FOXP3⁺ CD4⁺ T cells should unequivocally represent natural T_regs because T_effs with activation-induced FOXP3 expression (eg, peripherally generated T_regs) donot exhibit a naive phenotype. Furthermore, strict gating for CD4⁺CD25^hi T cells had led to false low T_reg numbers (1% to 2%) in human adult peripheral blood in many previous studies. With the detection of naive T_regs, total T_regs frequencies in mice (5% to 10%) and humans (3% to 7%) are actually comparable.

Modulation of the naive T_reg pool could be an interesting target for therapeutic intervention. Sereti et al recently showed reconstitution of CD4⁺CD25⁺CD45RO^– T cells in lymphopenic HIV patients after retroviral treatment. In response to IL-2 treatment, CD4⁺CD25⁺CD45RO^– T cells rapidly expanded.³⁹  Similarly, CD4⁺CD25⁺CD45RO^– T cells have been reported to expand during rheumatoid arthritis.⁴⁰  Although CD4⁺CD25⁺CD45RO^– T cells expressed high amounts of FOXP3, Sereti and coworkers considered these cells to be different from T_regs because they were CD45RO^lo, CD25^int, and also weak suppressors unlike CD4⁺CD25^hi T_regs. Moreover, CD4⁺CD25⁺CD45RO^– T cells were reported to be CD95^lo and resistant to CD95L-mediated apoptosis.³⁹  Because these properties are all features of naive T_regs, we consider it highly likely that the reported CD4⁺CD25⁺CD45RO^– T cells in treated HIV patients predominantly contained naive T_regs. Because FOXP3 was not analyzed at a single-cell level by flow cytometry, contamination of naive T_reg cells with T_convs within the analyzed CD4⁺CD25⁺CD45RO^– T cells could not be detected. Such a contamination with T_convs could explain the low IL-2 production and weak suppressive capacity of the CD4⁺CD25⁺CD45RO^– T-cell population derived from HIV patients that was observed.

Altogether, several lines of evidence support the existence of naive FOXP3⁺ T_regs (CD4⁺CD25⁺FOXP3⁺CD95^loCD45RO^loCD45RA^hi) and warrant thorough investigation of these cells in human disease. Naive and memory T_reg frequencies should be considered as a result of different regulating mechanisms including generation, elimination, and transformation of naive T_regs into memory T_regs. It is important to delineate the predominating processes for T_reg homeostasis in detail to allow a rational translation into human pathogenesis. This report provides a functional assay for the distinction between naive and antigen-primed T_regs, because resistance toward CD95L-mediated apoptosis clearly separates naive T_regs from their in vivo–preactivated CD4⁺CD25^hi T_regs counterparts. Pathogenetically this difference might be relevant, because Miyara et al recently showed that reduced T_reg frequencies in active SLE resulted from enhanced CD95-mediated cell death of SLE-derived CD4⁺CD25^hi T_regs.¹⁶  However, only CD25^hi T_regs, which are mainly composed of memory T_regs, were studied. Therefore, it remains possible that naive T_reg numbers are increased in SLE to compensate for a relative loss of memory-like CD4⁺CD25^hi T cells in SLE. As for other diseases with altered frequencies of CD4⁺CD25^hi T_regs, an analysis of naive T_regs should accompany that of the quantity and function of T_regs. Altered distribution of both cell types within the T_reg population might reflect not only age dependency but also fluctuating disease activity. Careful reanalysis of the previously neglected naive T_regs pool might help to optimize novel therapies based on the expansion of naive and apoptosis-resistant T_regs.

G. Faehnderich, C. Schneider, Dr A. Steinborn, and Prof Dr C. Sohn are gratefully acknowledged for making cord blood samples available to us. We thank K. Hexel and M. Scheuermann for FACS sorting and Dr A. Kuhn for assistance with recruitment of adult blood donors; T. Toepfer for excellent assistance in gene chip analysis; Dr R. Arnold, Dr K. Gülow, Dr M. Li-Weber, and S. Foehr for critical reading of the manuscript; and Prof Dr B. Arnold, Prof B. Wildemann, and Dr J. Haas for helpful discussion.