Whole-genome expression data from human and mouse CD4⁺CD25⁺ and CD4⁺CD25^- T cells, obtained using Affymetrix GeneChips (Affymetrix, Santa Clara, CA, USA), at the genomic scale were used to compile a primary list of genes involved in T_Reg cell function. CD4⁺ T cell subsets were isolated from either human peripheral blood or murine splenocytes and separated using FACS (fluorescence-activated cell sorting)-based cell sorting at purities consistently greater than 98%. Differential gene expression was determined using statistical parameters, as described under Material and methods, below. (For more detailed information, See Additional data file 1).

This primary data set from human T_Reg cells was extended for genes that were affected by FOXP3 overexpression in cultured human CD4⁺ T_hcell lines. To this end, different CD4⁺CD25^- derived T_h cell lines were generated by infection with retroviruses encoding for FOXP3 and GFP (green fluorescence protein) under the control of an internal ribosomal entry side (IRES) or with an empty control vector that contained only GFP. In these cells only FOXP3 overexpression could partially induce a T_Reg phenotype in vitro (data not shown). Using Affymetrix GeneChips, these genetically engineered cells were compared with cells infected with T_h GFP control vector. In addition, we also analyzed a human T_Reg cell line derived from human CD4⁺CD25⁺ T cells that maintained a regulatory phenotype in vitro and compared its gene expression profile with the control CD4⁺ T_h cell line. For the development of the Human T_Reg Chip we included those genes in our primary data set that were differentially expressed in both experiments by more than twofold. (For more detailed information, see Additional data file 2).

In additionally, T_Reg cell associated genes identified by literature search were also included (Additional data file 3). In summary, this resulted in the selection of 350 genes that were arranged on an oligonucleotide microarray. Furthermore, 45 control genes were included in the primary microarray design.

To obtain accurate and reliable transcription profiles, we validated the Human T_Reg Chip in terms of cross-platform comparability, sensitivity, and reproducibility of measurements. Relative expression data gained from the experiments investigating FOXP3 affected gene expression on Affymetrix GeneChips, as described above, were used as reference data in a cross-platform evaluation. Therefore, identical samples, obtained either from FOXP3 infected CD4⁺CD25^- T cells or GFP expressing controls, were also hybridized to the Human T_Reg Chip. Concordance of significantly regulated genes generated with the Human T_Reg Chip and the reference data was 81% (29/36; Figure 1a). Opposite regulation was observed only for a few marginally regulated genes (7/36). The Affymetrix GeneChip data for the 350 genes included in the Human T_Reg Chip is given in Additional data file 4). Furthermore, bacterial control genes at different concentrations were used to monitor microarray system sensitivity and the spectrum of linear signal measurement. A final concentration of 0.3 pmol/l was detectable, corresponding to approximately one transcript in 500,000 or approximately one copy per cell. Furthermore, we could demonstrate a linear regression between signal intensity and concentration covering more than three orders of magnitude (Figure 1b). To assess reproducibility, identical samples were applied to different Human T_Reg Chips and signal intensities were compared among each other (Figure 1c). The median correlation coefficient obtained from 52 log-log-plots was 0.98, which is well in line with commercially available microarray formats 2223 Finally, we determined the accuracy of measurements expressed as coefficient of variance calculated across eight replicates per gene. As depicted in Figure 1d, the vast majority of signal intensities (73%) calculated for the entire data set varied by less than 30%, reflecting the robustness of the applied microarray approach.

To obtain accurate and reliable individual transcription profiles we isolated CD4⁺CD25⁺ regulatory and CD4⁺CD25^- naïve T cells from peripheral blood of 11 healthy donors using MACS (Magnetic Cell Sorting) technology (Table 1). To estimate the fraction of T_Reg cells in the CD4⁺CD25⁺ cell population, we performed intracellular FOXP3 staining. Approximately 80% of the CD4⁺CD25⁺ T cells were FOXP3 positive and exhibited regulatory T cell function in vitro (Additional data file 5). Each sample was measured in at least two independent microarray experiments. Using Statistical Analysis of Microarrays (SAM) analysis, we identified 62 genes significantly differentially expressed in regulatory compared to naïve T cells. Based on Gene Ontology and references in the literature, genes were classified into functional categories such as cytokines/chemokines and their receptors (12 genes), cell cycle and proliferation (11), apoptosis (7), signal transduction (9), and transcriptional regulation (10). A detailed description of these genes is summarized in Table 2. Among them, LGALS3, CCR7, IL2RA (CD25), CTLA4, TRAF1, SATB1, and GZMK were additionally found to be affected by retroviral overexpression of FOXP3 in CD4⁺ T_h cells (Figure 1a).

Two-dimensional hierarchical clustering analysis was applied to arrange coexpressed genes and replicated experiments next to each other (Figure 2). The transcriptional pattern clearly separated CD4⁺CD25⁺ regulatory from CD4⁺CD25^- naïve T cells and distinguished between 32 upregulated and 30 downregulated genes.

Twenty-one of these 62 genes have already been described in the literature as being associated with T_Reg cells of both mouse and human origin, including FOXP3, CTLA4, IL2RA (CD25), and ITGB2 (Figure 3). Recovery of these 'old friends' confirmed our nonredundant microarray approach, including our cell separation strategy. Among the 62 genes, eight that were previously only implicated in murine T_Reg cell biology were also detected as being differentially expressed in human T_Reg cells (LGALS1, IL7R, GATA3, SATB1, TNFRSF1B, TNSF5, DGKA, and CCR5). Altogether, 15 genes were identified that were similarly regulated in mouse and human. Those genes at the intersection of both organisms reflect high levels of interspecies conservation during the evolutionary process, thereby lending credibility to their important role in T_Reg cell development and function (Figure 3). In addition to FOXP3, CTLA4 and IL2RA, we also found the chemokine receptor 7 (CCR7), the transferring receptor (TFRC) and integrin beta 2 (ITGB2) genes in this intersection group between mouse and human. Furthermore, six genes previously associated with human T_Reg cells were identified. Apart from the 'old friends', we identified 41 'new players' that have not previously been reported in the context of human T_Reg cells (Figure 3).

To verify the accuracy of our microarray data in more detail, real-time RT-PCR (reverse transcription polymerase chain reaction) was performed using the original samples. Referring to well characterized T_Reg cell genes (FOXP3, CTLA4, and CCR7), we were able to confirm our approach (Figure 4). This gave greater credence and reliability to the numerous additional genes that have not yet been reported in T_Reg cells. We selected three of these 'new players' (TNFRSF1B, TRAF1, LGALS3) and confirmed their T_Reg cell specific expression by quantitative real-time RT-PCR (Figure 5). As shown, in general PCR results correlated well with the differential gene expression data obtained by application of the Human T_Reg Chip. For a few donors variability in gene expression was observed between microarray and quantitative RT-PCR data, but the direction of change was consistent, lending confidence to the reliability of the Human T_Reg Chip results. Quantitative differences in fold changes have previously been described; in particular, an underestimation of real expression changes by microarray approach versus quantitative RT-PCR has been reported 2425.

To elucidate potential pathway modules implicated in T_Reg cell biology, we applied PathwayAssist, (Ariadne Genomics, Rockville, MD, USA), software to our unique expression dataset of human T_Reg cells from individual healthy donors. Mapping the 62 T_Reg cell specific genes yielded a network of 31 genes directly interacting with each other (data not shown). These 31 genes provided a comprehensive framework for further dissection into functional modules. These modules point to mechanisms controlling diverse cellular processes such as survival/apoptosis, T cell receptor signaling/activation/proliferation, and differentiation/maintenance of human T_Reg cells and are described in the following text.

Autoimmunity occurs as a consequence of self-tolerance breakdown, presumably resulting from a combination of inherited polymorphisms (or DNA variations), acquired environmental triggers, and stochastic events 59. Analyzing our transcriptional pattern of human T_Reg cells isolated from individual healthy donors, we found that 32 of the genes identified are involved in the pathogenesis of diverse autoimmune diseases (Table 2). We focus here on a few affected genes that are central to the functional modules discussed above and that might therefore influence disease pathogenesis.

We found TNFRSF1B to be 2.5-fold overexpressed in the naturally occurring T_Reg cells compared with their naïve counterparts. A single nucleotide polymorphism (SNP) in this gene was reported to influence susceptibility to multiple sclerosis, a severe inflammatory autoimmune disorder of the central nervous system 60. In addition, Sashio and coworkers 61 linked two other polymorphisms to the TNFRSF1B genelocus that increase susceptibility to Crohn's disease and ulcerative colitis, which are both chronic inflammatory diseases of the gastrointestinal tract. In Japanese patients, Morita and coworkers identified another SNP in the TNFRSF1B gene associated with systemic lupus erythematosus (SLE).

Type I diabetes is a T cell mediated inflammatory autoimmune disease of the endocrine pancreas, resulting in lack of insulin caused by β cell destruction. We found 18 genes in our human T_Reg cell signature that have been reported to contribute to pathogenesis of this disease, including granzyme A (GZMA) 63, the CD40 ligand (TNFSF5) 6465, CTLA4 66, and the T-cell specific transcription factor 7 (TCF7) 67. Furthermore, two polymorphisms in the HLA-DRB1 gene, which we found to be overexpressed in T_Reg cells, have been described to confer high-risk susceptibility 68.

Rheumatoid arthritis (RA) is a chronic inflammatory disorder that affects the joints and is probably caused by autoimmune mechanisms. Twenty-one T_Reg specific genes have been described as susceptibility genes for RA. For example, LGALS3 6970, GZMA 71, and the S100 calcium binding protein A4 (S100A4) 72 have been described as highly expressed in the synovial tissue and at sites of joint destruction contributing to the inflammatory process. The complex genetic component of RA etiology was further demonstrated by the discovery of multiple polymorphisms, for example in genes of the chemokine receptor 5 (CCR5) 73 and of HLA-DRB1 74, conferring high risk susceptibility.

In mice deficient for STAT4, a gene we found to be downregulated in our human T_Reg cells, RA is suppressed because of reduced levels of IL-12 and interferon interferon (IFN)-γ 75. Interestingly, STAT4^-/- mice were additionally almost completely protected from diabetes 76 and induction of experimental allergic encephalomyelitis 77, underlining the importance of STAT4 downregulation in T_Reg cells.

Because T_Reg cells are essential for the maintenance of self-tolerance, SNPs or mutations that affect genes expressed in T_Reg cells may result in the synthesis of aberrant mRNAs and proteins, which in turn could impair T_Reg cell function and/or development, leading to higher risks for autoimmunity. Additionally, failures in gene regulation resulting in inadequate protein amounts could disturb appropriate T_Reg cell activity, thereby probably contributing to the pathogenesis of autoimmune disorders.

Because most of the genes discussed here are central components of the functional modules discussed above, it is conceivable that the dysregulation of one or more of these genes affect T_Reg cell activity in terms of survival/apoptosis, differentiation, proliferation, and suppressor function, thereby promoting breakdown of self-tolerance and eventually leading to autoimmunity.

Blood samples were collected from 11 healthy donors after informed consent had been obtained, in accordance with institutional guidelines. The Ethics Committee of Hanover Medical School approved the study protocol. Basic characteristics of all donors are summarized in Table 1. None of the donors suffered from allergies or autoimmune disease and all were free from acute or chronic infections.

CD4⁺ T cells were prepared from peripheral blood of healthy donors by centrifugation over Ficoll-Hypaque gradients (Biochrom AG, Berlin, Germany) and MACS isolation using the CD4⁺ T cell isolation kit and AutoMACS technology (Miltenyi Biotech, Bergisch Gladbach, Germany). Subsequently, cells were separated into CD4⁺CD25^- and CD4⁺CD25⁺ T cells by either using sorting on a MoFlo (DakoCytomation, Fort Collins, CO, USA) to a purity in excess of 98% (for Affymetrix studies) or an AutoMACS using the regulatory human T cell isolation kit (Miltenyi Biotech). To increase purity of the CD25^- T cell fraction an additional separation step depleting remaining CD25⁺ T cells was added, if necessary. For studies on the Human T_Reg Chip purity of the enriched cell fractions was above 90%, as determined by flow cytometry (the remaining contaminating cells mainly represent CD16⁺/CD56⁺ natural killer cells and, at lower levels, CD8⁺ T cells, CD19⁺ B cells and CD14⁺ monocytes; Additional data file 6). Isolated cells were either directly used for RNA purification or pooled equivalently as indicated before RNA purification.

For Affymetrix GeneChip experiments, red blood cell depleted splenocytes from BALB/c mice were labeled with anti-CD4 and anti-CD25. Labeled cells were separated with a MoFlo and purity was in excess of 98%. Isolated cells were pooled equivalently (three independent individuals) and subsequently used for RNA purification.

CD4⁺CD25⁺ T_Reg cells were stimulated once with plate-bound anti-CD3 (TR66, 1 μg/ml), soluble anti-CD28 (CD28.2, 1 μg/ml; BD Bioscience, San Jose, CA, USA), and 50 U/ml recombinant human IL-2 (Proleukin; provided by P Wagner, Chiron Corporation, Emeryville, CA, USA), and thereafter weekly with irradiated allogeneic EBV-transformed B cells (LG2-EBV; provided by T Boon, LICR, Brussels, Belgium). CD4⁺CD25^- T cells were stimulated directly with irradiated LG2-EBV cells. Culture medium was Iscove's modified Dulbecco's medium, with 10% fetal calf serum,100 U/ml penicillin/streptomycin, and nonessential amino acids (PAA Laboratories, Linz, Austria). Human peripheral blood was obtained after informed consent had been obtained, in accordance with institutional guidelines. Antibodies for immunostaining were PE-, FITC-, APC-, and CyChrom-conjugated antibodies against CD4 (RPA-T4), CD25 (M-A251; all from BD Bioscience), and FOXP3 (PCH101; eBioscine Inc., San Diego, CA, USA) and respective isotype controls. Anti-CD3ε (TR66, produced from hybridoma supernatants) and anti-CD28 (CD28.2; BD Bioscience) were used for T cell stimulation.

The cDNA encoding human FOXP3 was amplified from cDNA of T_Reg cells using high fidelity PFU polymerase (Promega) and specific primers (FOXP3: 5'-GAC AAG GAC CCG ATG CCC A-3' and 5'-TCA GGG GCC AGG TGT AGG GT-3'). The PCR product was cloned into pCR4.1 TOPO (Invitrogen, Carlsbad, CA, USA), sequenced, and inserted into a pMSCV-based retroviral vector encoding an enhanced GFP under the control of an IRES sequence. The amphotropic PT67 packaging cell line (provided by M. Wirth, GBF, Braunschweig, Germany) was used for transfection. Filtrated (0.45 μm) virus-containing supernatant supplemented with 8 mg/ml sequabrene (Sigma-Aldrich, Munich, Germany) was applied to T_h cells at day 2 after allogeneic stimulation by centrifugation at 5000 × g for 60 minutes at room temperature. Cells were expanded thereafter with 50 U/ml IL-2, and GFP-expressing cells were sorted 1-2 weeks later using a FACS-Vantage (BD Bioscience).

To confirm purity of the separated cell fractions, regulatory and naïve T cells were analyzed by multicolor FACS using the following antibodies: anti-CD4-FITC and anti-CD25-PE (Miltenyi Biotec). Flow cytometry was done using a FACSCalibur applying CellQuest software (BD Bioscience).

CD4⁺CD25⁺ regulatory and CD4⁺CD25^- naïve T cells were isolated by MACS technology as described above. After cell lysis, RNA was extracted from both cell populations applying the RNeasy kit (Qiagen, Hilden, Germany). cDNA was synthesized using oligo(dT) primers and random hexamers by SuperScript II Reverse Transcriptase (Invitrogen, Karlsruhe, Germany). Quantitative real-time RT-PCR was performed in an ABI PRISM cycler (Applied Biosystems, Foster City, CA, USA) using a SYBR Green PCR kit from Stratagene (La Jolla, CA, USA) and specific primers optimized to amplify 90-230 base pair fragments from the different genes analyzed. A threshold was set in the linear part of the amplification curve, and the number of cycles needed to reach it was calculated for every gene. Relative mRNA levels were determined by using included standard curves for each individual gene and further normalization to RPS9 as a housekeeping gene. Melting curves established the purity of the amplified band. Primer sequences are summarized in Table 3.

A total of 395 oligonucleotides were deposited onto CodeLink activated slides (Amersham Biosciences, Freiburg, Germany) at a concentration of 25 μmol/l in 1.5× sodium phosphate buffer in a contact-dependent manner using a MicroGrid TAS II spotter (BioRobotics, Freiburg, Germany). All 50-mers were amino-modified at the 5'-end enabling covalent linkage to reactive ester groups provided by the glass surface. Coupling of DNA was ensured by overnight incubation in a saturated sodium chloride chamber, and blocking residual reactive groups was done as recommended by the manufacturer 78. Until used, slides were maintained in a desiccated environment. To ensure complete spotting, SYBR-Green staining of three randomly selected Human T_Reg Chips of each printing batch was performed as previously described 79.

Each probe in our microarray consists of a single 50 mer oligonucleotide, because utility and performance of 50 mer oligonucleotide microarrays was previously established 80. The Human T_Reg Chip consists of 350 oligonucleotides probing genes specific for T_Reg cells and 31 oligonucleotides representing housekeeping genes consulted for normalization. Furthermore, many control oligonucleotides are included: two 5'-3' controls to ensure RNA integrity, four bacterial hybridization controls to examine a linear hybridization process, five spike-in controls to check sample preparation, one positive control (Arabidopsis thaliana) for simpler grid finding, and finally 32 negative controls to calculate the background level. Altogether, we immobilize eight replicates per oligonucleotide, split into two separated arrays per slide, each containing 1,600 spots. Genes probed on the Human T_Reg Chip were selected by extensive analyses of literature and previously conducted Affymetrix microarray experiments. Design and synthesis of the oligonucleotides were performed by MWG using the Affymetrix probe sets as reference. Our Human T_Reg Chip will be made available to the scientific community on our website 81.

Quality and integrity of the total RNA isolated from 1-2 × 10⁵ CD4⁺CD25⁺ and CD4⁺CD25^- T cells was controlled by running all samples on an Agilent Technologies 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). Samples were prepared by applying a double-linear amplification method in accordance with the Eberwine protocol82 and modified by Affymetrix. Briefly, the first round of RNA amplification was performed without biotinylated nucleotides using the Promega P1300 RiboMax Kit for T7 amplification (Promega, Mannheim, Germany). After clean up of the precipitated aRNA synthesis of second round, first-strand cDNA was done using random hexamers (Pharmacia, Freiburg, Germany). Subsequent second-strand cDNA was prepared as in the first round but integrating an additional RNAse H incubation step to digest the aRNA before annealing of the T7T23V primer. The second round of RNA amplification was performed as an in vitro transcription assay in the presence of biotinylated UTP using the GeneChip^® Expression 3'-Amplification Reagents Kit for IVT Labeling (Affymetrix). The concentration of the obtained biotin-labeled cRNA was determined by ultraviolet absorbance and its quality as means of product length distribution was again checked using the Agilent Bioanalyzer. In all cases, 15 μg of each biotinylated cRNA preparation was fragmented and placed in a hybridization cocktail containing four biotinylated hybridization controls (BioB, BioC, BioD, and Cre). Samples were hybridized to individual Human T_Reg Chips for 16 hours at 42°C using a Lucidea Slidepro (Amersham Biosciences). After hybridization the microarrays were washed as recommended in the manufacturer's instructions (CodeLink Expression Bioarray System; Amersham Biosciences), stained with Cy5-streptavidin (Amersham Biosciences), and read using an arrayWorX^e scanner (Applied Precision, Issaquah, WA, USA).

Samples were amplified for GeneChip analysis according to the recommended protocols by the manufacturer. In all cases, 10 μg of each biotinylated cRNA preparation was fragmented and placed in a hybridization cocktail containing four biotinylated hybridization controls (BioB, BioC, BioD, and Cre), as recommended by the manufacturer. Samples were hybridized to an identical lot of Affymetrix GeneChips for 16 hours. After hybridization the GeneChips were washed, stained with SA-PE, and read using an Affymetrix GeneChip fluidic station and scanner.

Differentially expressed genes between CD4⁺CD25⁺ and CD4⁺CD25^- measured on Affymetrix GeneChips were selected according to predefined categories deduced from three parameters calculated by MAS 5 software: fold change (FC), change p value (pValue), and signal intensity difference (SID). Category A is defined as an FC above 2, pValue <0.001 (for increased) or >0.999 (for decreased), and SID above 200. Category B is defined as FC above 2, pValue <0.01 (for increased) and >0.99 (for decreased), and SID above100. Category C is defined as FC above 1.5, pValue <0.001 (for increased) and >0.999 (for decreased), and SID above 40.

The likelihood of a significant regulation decreases from category A to C. Preferentially, most of the selected genes collected for the Human T_Reg Chip are categorized as A. Selection was performed by collecting genes that were significantly regulated in human cells, genes that were similarly regulated between mouse and human, genes that were found to be regulated only in mouse cells and referenced in the literature, and genes that were significantly affected by FOXP3 overexpression in cultured T_h cell lines. Also considered were genes known for their impact in mouse and human regulatory T cell development.

Signal intensities were qualified and quantified by means of Imagene software v5.5.2 (BioDiscovery, Los Angeles, CA, USA). Spots of poor quality (flag = 3) were excluded from further analysis. To adjust arrays from different experiments, data normalization based on median signal intensities of the housekeeping genes was carried out as proposed using the following formula:

Where SI_normalized is the normalized signal intensity, I_n is the mean signal intensity of gene n, B_n is the mean background intensity of gene n, and <ln house> is the median signal intensity from housekeeping genes expressed as ln (logarithm naturalis).

Differences in gene expression among CD4⁺CD25⁺ regulatory and CD4⁺CD25^- naïve T cells were determined statistically by corrected t test analysis using the SAM tool 83. Differentially expressed genes were defined using the following SAM parameters: delta = 2.46 and median FDR (false discovery rate) = 0.48. For two-dimensional hierarchic clustering analysis Genesis software v1.4.0 was applied 84.

The entire data sets are deposited in a MIAME compliant format at Gene Expression Omnibus (GEO) 85. Data derived from the Human T_Reg Chip are available under the series accession number GSE3882 (platform ID, GPL3110).

Data derived from Affymetrix GeneChip system and used as reference and selection data sets are published at GEO under series accession number GSE4527 (FOXP3 and GFP transduced CD4⁺ T_h cells) and GSE4571 (representing data from CD4⁺CD25⁺ and CD4⁺CD25^- T cellsisolated by cell sorting from human peripheral blood and CD4⁺CD25⁺ and CD4⁺CD25^- T cells isolated by cell sorting from spleen prepared from BALB/C mice).