Quantitative DNA Methylation Analysis of FOXP3 as a New Method for Counting Regulatory T Cells in Peripheral Blood and Solid Tissue

Regulatory T cells (Treg) are an essential component of the immune system, balancing necessary aggressiveness against foes with tolerance for self-constituents (1). Pathologic imbalances of Treg are found in two major disease classes, i.e., autoimmune diseases and cancers. Whereas in autoimmune diseases, low Treg levels contribute to the failure to protect the body from its own defense mechanisms (2–4), recent studies indicate an overrepresentation of Treg in the peripheral blood of patients with solid tumors (5–10). Furthermore, an increased accumulation of Treg in the tumor mass goes along with an adverse prognosis (11, 12). Those observations not only contradict the concept of immuno-ignorance, but they suggest an alternative functional explanation for tumor immunity: tumor-induced dysregulation would exhibit tolerance, impeding a self-protective immune response. The tumor development thus would be promoted in part by the immune system itself, i.e., the suppressive function of Treg. Thus, Treg could be an important variable to be measured in diagnostic and patient monitoring settings.

An increasingly explored option for the therapy of solid tumors and other diseases is the modulation of the immune system. For instance, patients with melanomas and kidney cancer are treated by the administration of an immunostimulatory cytokine interleukin 2 (IL-2), which is thought to activate natural killer (NK) cells and T cells. Recent reports suggest that Treg proliferation is also triggered by IL-2, which would constitute an adverse effect for the intended immune activation of the therapy (13–16). A converse approach is pursued in immunosuppression therapies for transplantations, in which the intention is to abrogate host versus graft and graft versus host reactions. For that, T lymphocytes are blocked or eliminated. Modern therapies use antibodies, such as thymoglobuline (rabbit ATG) and basiliximab. rATG is a polyclonal antibody reactive against a variety of T-cell antigens and has been shown to efficiently eradicate T-lymphocytes (17). Basiliximab is a chimeric mouse human monoclonal antibody directed against the IL-2Rα (CD25; ref. 18). Binding of basiliximab to its target effectively inhibits T-cell signaling. In order to judge the efficacy of those therapies, it is important to measure the actual cellular effects of both stimulation (such as IL-2 treatment) and immunosuppression (basiliximab and rATG). Accordingly, immunomonitoring strategies are a crucial diagnostic and analytic component during the development or even in the clinical routine application of those therapies. Within the monitoring concepts, it is of particular value to quantitatively determine Treg because—in healthy individuals—they are naturally responsible for the regulation of an adequate immune response. However, precise Treg measurement has thus far remained awkward. Initially, CD4⁺ Treg were characterized by high expression levels of CD25, which is a component of the high-affinity IL-2 receptor (1). Although CD25 has always been known to be a rather unspecific activation marker, the identification of the forkhead box transcription factor (FOXP3)—as a constitutively expressed master switch for Treg development—has provided a powerful marker and an essential tool for analyzing and understanding the role of Treg (19–22). The specificity of FOXP3 expression in Treg was shown in the murine immune system using transgenic animals harboring FOXP3 promoter–driven reporter genes (23). High levels of FOXP3 expression were also found in human natural CD25^highCD4⁺ Treg. As a marker for human Treg, however, FOXP3 mRNA expression and protein synthesis is of doubtful value, due to the transient expression of FOXP3 in activated nonregulatory effector T cells (24–26). Thus, in the absence of a specific Treg marker, the interpretation of FOXP3⁺ and/or CD25⁺ Treg remained ambiguous because the measurements employed determine a mixture of activated conventional T cells and natural Treg. In addition, quantification of Treg by measuring FOXP3 and CD25 in solid tissues using immunohistochemical methods remains experimentally demanding, error-prone, and difficult to objectively standardize (27). Far-reaching clinical decisions within routine applications in such a setting are not in sight at present.

We have previously shown, both in mice and men, that stable expression of FOXP3 in naturally occurring Treg requires epigenetic, i.e., DNA methylation–based regulation (28, 29). DNA methylation is a biologically and chemically stable epigenetic modification, locking long-term gene expression patterns (30–32). Demethylation at a highly conserved region within the human FOXP3 gene (Treg-specific demethylated region, TSDR; ref. 29) was found to be restricted to Treg when tested against all major peripheral blood cell types and a selection of non–blood cells. In addition to the high specificity for Treg, it was also observed that FOXP3 TSDR demethylation occurred only in natural Treg, but not in recently activated effector T cells transiently expressing FOXP3 (28, 29). It was further associated with stable FOXP3 expression upon Treg expansion in vitro. These data indicated that epigenetic modifications in the FOXP3 TSDR serve as a valuable marker for the identification of T cells with a stable Treg phenotype.

Here, we established a quantitative real-time PCR-based methylation assay, which enables the specific and sensitive determination of Treg numbers by measuring demethylated FOXP3 TSDR DNA. We addressed the question of whether the methylation assay reliably reports Treg numbers in diagnostic or therapeutic settings, such as IL-2 therapy for patients with melanoma and rATG or basiliximab treatment for transplantation patients. For the former, we used peripheral blood from patients with melanoma before and after completion of an IL-2 cytokine treatment. For the latter, we analyzed peripheral blood of patients prior to and after kidney or simultaneous kidney/pancreas transplantation, with the antibodies administered perisurgically. We further conducted a blinded study testing the notion that an increase of FOXP3 TSDR demethylation in the peripheral blood would serve as an indicator for the establishment of malignancies in patients with manifested cancers compared with healthy controls. Finally, we conducted studies to evaluate this real-time PCR tool in its use to quantify the Treg number in formalin-fixed paraffin-embedded (FFPE) solid tissue samples, including both healthy and tumorous tissues.

Healthy donors. Peripheral blood samples were obtained from healthy donors after informed consent in accordance with local ethical committee approval. For the sorting of major peripheral blood leukocyte populations, samples were treated according to Baron and colleagues (29). Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation using Ficoll-Hypaque (Sigma-Aldrich). CD4⁺ T cells were isolated from PBMC by using anti-CD4 microbeads and the AutoMACS magnetic separation system (Miltenyi Biotec). All microbeads were purchased from Miltenyi Biotec and all antibodies for cell surface stainings were from BD PharMingen. Intracellular FOXP3 staining was performed with the PE anti-human FOXP3 staining set (eBioscience) and used according to the manufacturer's instructions. MACS-sorted CD4⁺ T cells were stained using anti–CD45RA⁻FITC and anti–CD25⁻APC. Cells were sorted into CD25^highCD45RA⁻ Treg and CD25⁻CD45RA⁺ naïve T cells by fluorescence-activated cell sorting (FACS; Aria, BD Bioscience). An aliquot of the CD4 population was used to determine the content of FOXP3⁺ cells by flow cytometry; here, the analyzing gates resemble the sorting gates. Cytometric analysis was performed as previously described (33) using a FACSCalibur (BD Biosciences) and the FlowJo software (Tree Star).

For cell spiking, 10,000, 5,000, 2,500, 1,250, and 625 purified CD25^highCD4⁺ cells were each mixed with 2 million isolated granulocytes corresponding to a Treg percentage of 0.5%, 0.25%, 0.13%, 0.06%, and 0.03%, respectively. Cell mixtures were subjected to genomic DNA isolation, followed by bisulfite treatment and methylation-sensitive FOXP3 TSDR real-time PCR (see below). Template input per reaction corresponds to 56, 28, 14, 7, and (approximately) 3.5 copies of Treg DNA.

Patients with melanoma following IL-2 therapy. Patients with melanoma (five men and one woman; mean age, 52 years; ranging from 33 to 75) with stage IV metastatic disease, received IL-2 by s.c. administration with or without the addition of histamine dihydrochloride as described previously (34). After informed consent according to the local ethical guidelines, heparinized blood samples were drawn before and 2 weeks after the completion of a 4-week schedule of IL-2 therapy. PBMCs were isolated by density gradient centrifugation using Ficoll-Hypaque 1.077 (Biochrom or Sigma-Aldrich). For extracellular staining, the following surface monoclonal antibodies were used: anti–CD3-PercP (clone SK7), anti–CD25-PE (M-A251), and anti–CD4-FITC (B9.11). Nuclear anti–FOXP3-APC staining was performed according to the manufacturer's instructions using anti–FOXP3-APC (PCH101; eBioscience). Data acquisition was performed on FACSCalibur (BD Bioscience).

Transplantation patients under immunosuppressive therapy. Between October 2007 and May 2008, transplantation patients were recruited from the Department of Surgery and Nephrology, Virchow Klinik, Universitätsmedizin Charité, Berlin, and received a non-HLA identical allograft from a cadaveric donor. Transplant patients (three male patients; mean age, 48 ± 13) were first transplanted (kidney/pancreas) and received an initial 1.5 mg/kg body weight i.v. thymoglobulin (Genzyme) starting at day 0 followed by 4 further consecutive days. Patients for basiliximab therapies (n = 2 male, n = 4 female; mean age, 52 ± 11) were also first transplanted (kidney) and received two doses of basiliximab (20 mg i.v. 2 h before reperfusion and at day 4 posttransplantation). Maintenance immunosuppression in both groups consisted of tacrolimus, mycophenolate mofetil, and steroids. PBMCs were isolated by density gradient centrifugation using Ficoll-Hypaque 1.077 (Biochrom or Sigma-Aldrich). All experiments using human material were approved by the local ethical committee.

Peripheral blood analysis of cancer patients and healthy donors. Samples were obtained from the sources listed below and sample numbers received from each source are indicated in parentheses: Proteogenex (93), Universitätsklinik des Saarlandes (n = 1), Universitätsklinikum Schleswig-Holstein, Campus Lübeck (n = 2), Medizinische Hochschule Hannover (n = 18), and Universitätsklinikum Münster (n = 2). Written informed consent was obtained from all study participants adhering to the local ethical guidelines. Blood was collected in one or more 10 mL BD vacutainer tubes (BD Vacutainer Plus blood collection tube; Becton Dickinson) and each tube immediately inverted ∼10× to avoid blood clotting. The blood collection tubes were centrifuged at 1,500 × g for 10 min at 4°C with the centrifuge brake turned off. After centrifugation, ∼0.5 to 1 mL of fluid including the cellular layer between plasma and erythrocytes were transferred into a prelabeled 2 mL Cryo vial (Fisherbrand 2 mL round style bottom cryogenic storage vials; Fisher Scientific). Samples were frozen at −80°C within 4 h of the blood draw and stored at this temperature until shipped on dry ice. The sample collection consisted of 22 patients with breast cancer (22 women; mean age, 60; 18 patients with stage I to II disease and 4 patients with stage III or IV disease), 27 patients with colorectal cancer (13 men and 14 women; mean age 64; 13 patients with stage I or II and 14 patients with stage III or IV disease), 24 lung cancer patients (11 male and 13 females; mean age, 62; 9 patients suffering from stage I and II, 9 patients with stage III or IV, and 6 patients with no disease stage information), 23 prostate cancer patients (23 male; mean age, 65; 9 with stage I/II disease, 1 with stage III/IV disease, and 13 without stage information) and 20 normal controls (11 male and 9 female; mean age, 55).

FFPE tissues. FFPE samples from 15 patients with primary colorectal cancer and 5 patients with stages II to IV primary lung cancer were retrieved from the Pathology Department of Charité, Campus B.F., Berlin. Colorectal cancer tissues had been already analyzed in a previous study for FOXP3 expression using immunohistochemistry (35). FFPE control samples were either from adjacent healthy colon (five samples) and lung (one sample) tissue from the tested patients or healthy tissue from other colon (five samples) and lung (two samples) patients obtained during surgery. All patients with colorectal cancer (mean age, 55 years; range, 36–73) and controls (mean age, 58 years; rang, 48–76) were female. All lung controls (n = 3) were male (ages 61, 65, and 67), one lung cancer patient was female and four were male (mean age, 63 years; range, 52–74). For DNA isolation from FFPE tissues, up to eight tissue slices were used. Slices were incubated overnight in 50 mmol/L of Tris/HCl (pH 8), 1 mmol/L EDTA, 0.5% Tween 20 containing 2 mg/mL of proteinase K with subsequent incubation at 90°C for 10 min and DNA was purified with Qia-AMP DNA FFPE tissue kit (Qiagen) according to the manufacturer's protocol.

Genomic DNA was isolated using the DNeasy blood and tissue kit (Qiagen). For PBMC and sorted blood cells, the protocol for cultured cells was followed. For DNA isolation from buffy coats, the whole blood protocol was used. Bisulfite treatment of genomic DNA was performed according to Olek and colleagues (36) with minor modifications.

Real-time PCR was performed in a final reaction volume of 20 μL using Roche LightCycler 480 Probes Master (Roche Diagnostics) containing 15 pmol each of methylation or non–methylation-specific forward and reverse primers for TSDR, 5 pmol of hydrolysis probe, 200 ng of lambda-DNA (New England Biolabs), and 30 ng of bisulfite-treated genomic DNA template or a respective amount of plasmid standard. Each sample was analyzed in triplicate using a LightCycler 480 System (Roche). Cycling conditions consisted of a 95°C preheating step for 10 min and 50 cycles of 95°C for 15 s followed by 1 min at 61°C. Methylation-specific and demethylation-specific amplification primers for TSDR map to the following chromosomal positions: NCBI36:X:49004163-49004190:1 (forward primers) and NCBI36:X:49004227-49004251:1 (reverse primers). Hydrolysis probes map to chromosomal position NCBI36:X:49004200-49004222:1 (demethylation-specific probe) and NCBI36:X:49004200-49004217:1 (methylation-specific probe).

PCR products were generated with methyl– and non-methyl–specific primers for FOXP3 TSDR using genomic, bisulfite-treated DNA from sorted naïve and regulatory T cells. DNA fragments were cloned into pCR2.1-TOPO vector, using TOPO TA cloning kit (Invitrogen) according to the manufacturer's instructions and verified by sequencing. Plasmids were purified with Qiagen Plasmid Midi Kit, the concentration was determined by Qubit fluorometer (Invitrogen) and diluted to obtain final concentrations of 100, 10, 1, and 0.1 fg representing 20,000, 2,000, 200, and 20 plasmid copies as standard for quantitative PCR (qPCR) reactions, each for methylated and nonmethylated FOXP3 qPCR assay.

Amounts of methylated and unmethylated FOXP3 DNA were estimated from calibration curves by linear regression on crossing points from the second-derivative maximum method (37). The median was used to aggregate triplicate measurements of the buffy coat samples. The proportion of unmethylated DNA was computed as the ratio of unmethylated FOXP3 TSDR-DNA and the sum of methylated and unmethylated FOXP3 TSDR-DNA. For female patients, this ratio was corrected with a factor of 2 due to the fact that one of the two TSDR alleles is methylated as a result of X-inactivation.

For the quantitation of the FOXP3 methylation-based discrimination between peripheral blood of cancer patients and normal controls, the area under the receiver operating characteristics (ROC) curve (AUC) value was used. Here, it is equivalent to the probability that a randomly selected cancer patient shows higher FOXP3 methylation in peripheral blood than a randomly selected normal control. No discriminative ability corresponds to an AUC of 0.5. A perfect discrimination has an AUC of 1.0. AUCs were estimated by means of the trapezoidal rule. The Wilcoxon-rank sum test was used to compare demethylation rates between normal and cancer patients. The Wilcoxon signed-rank test was used to compare demethylation rates pre– and post–IL-2 treatment. For correlation analysis, Pearson's product moment coefficient and t test statistics were used. All P values are two-sided.

Immunohistochemical staining was performed as previously described (35). In brief, slides were incubated with the rat monoclonal antibody against human FOXP3 protein (PCH101, 1:200; eBioscience), followed by biotin-conjugated rabbit anti-rat and the EnVision peroxidase kit (Dako). Ten randomly selected high-power fields (1 HPF = 0.237 mm²) were analyzed for FOXP3⁺ cell infiltration in tumor tissue and matched normal colonic mucosa/lung parenchyma, and 10 HPF were averaged in each case.

Methylation-sensitive FOXP3 TSDR real-time PCR system. In order to quantitatively analyze the specificity of FOXP3 TSDR demethylation to Treg, we established a real-time PCR system (qPCR). This system consists of methylation-dependent amplification primers and hybridization probes for both bisulfite-converted methylated and unmethylated DNA of the FOXP3 TSDR (Fig. 1A). We tested this system on plasmids cloned from bisulfite-converted Treg and naïve T cell DNA. These clones contain sequences corresponding to unmethylated or methylated bisulfite-converted FOXP3 TSDR. We showed high specificity, linearity of amplification over six orders of magnitude, and no cross-reactivity with the opposite template species, even at unphysiologically high concentrations (ranging from 20 to 200 million copies of plasmid DNA; Fig. 1B).

With technical variables specified, we compared the performance of conventional FACS analysis for Treg detection with the FOXP3 TSDR demethylation assay. For this, we used sorted CD25^highCD4⁺ T cells and CD25⁻CD4⁺ conventional T cells from healthy male and female donors. Flow cytometry analysis revealed FOXP3 expression in >95% of CD25^high Treg and up to 4.3% in CD25⁻CD4⁺ conventional T cells (Fig. 2A). Demethylation analysis confirmed these findings with ∼93% of the CD25^high Treg showing an unmethylated FOXP3 TSDR in male donors and 45.9% in female donors (∼92% of the cells, when corrected with a factor of 2 owing to inactivation and methylation of one X-chromosome). Among CD4⁺CD25⁻ cells, <1% were demethylated in the FOXP3 TSDR (Fig. 2B). Treg content in CD4⁺CD25^high as determined by TSDR analysis is somewhat lower than observed with FOXP3 expression analysis, suggesting superimposition by activated non–Treg in the latter procedure. Equally, TSDR demethylation is lower in naïve T cells than determined by FACS analysis, suggesting that impurities retained during CD25⁻ sorting mainly originated from CD25^medium cells, which may express FOXP3 to some extent but are not exhibiting bona fide natural Treg with the CD25^high, FOXP3 TSDR demethylated Treg phenotype. However, altogether, a good correlation between both detection methods was observed.

Next, the assay sensitivity was tested in an in vitro system. For this, we added a gradually decreasing amount of sorted Tregs into a constant number of control cells (granulocytes), which have been shown to be completely methylated in the FOXP3 TSDR (ref. 29; see also next paragraph). We robustly detected 0.5%, 0.25%, 0.13%, and 0.06% Treg. The detection limit was at 0.03% Treg corresponding to approximately 3.5 copies of unmethylated template DNA in 12,000 copies of methylated background DNA (Fig. 2C).

FOXP3 TSDR demethylation in sorted blood cell subtypes. For the presented assay, it is a prerequisite for Treg measurement in full blood that all non-Treg blood fractions are overwhelmingly methylated in the FOXP3 TSDR. Thus, to characterize biological specificity, we analyzed the methylation level in various leukocyte fractions. Results are summarized in Table 1. Isolated CD14⁺ monocytes, CD15⁺ granulocytes, CD56⁺ NK cells, and both CD8⁺ memory (CD27⁺CD45RA⁺) and CD8⁺ naïve (CD27⁺CD45RA⁻) cytotoxic T cell fractions showed <0.5% TSDR demethylation. CD19⁺ memory and naïve B cell fractions comprise <2% demethylated TSDR. CD4⁺CD27⁺CD45RA⁺ naïve T cells and CD4⁺CD27⁺CD45RA⁻ memory T cells contain a demethylated fraction of 6.2% and 10.86%, respectively. Both fractions have previously been described to contain approximately 10% of CD25^high FOXP3⁺ Treg. Therefore, our measurements meet the expected rate of Treg in those fractions. As shown above in Fig. 2B, the selective removal of CD25^high cells from the total CD4⁺ cell fraction leads to the depletion of all cells with a demethylated FOXP3 TSDR. This fact once more suggests the specificity of TSDR demethylation to CD25⁺⁺ cells. Taken together, it seems that only Treg are demethylated at the TSDR and a specific Treg measurement based on the TSDR demethylation should be feasible.

FOXP3 TSDR demethylation in patients with melanoma treated with IL-2. Next, we studied the effect of IL-2 therapy on the frequency of CD4⁺CD25^high Foxp3⁺ Treg in the PBMCs of six patients with metastatic melanoma using both FACS and TSDR demethylation analyses (Fig. 3A). Prior to therapy, FACS measurements of PBMC showed an average frequency of CD4⁺CD25^high FOXP3⁺ cells of 4.5% (range, 1.9–8.9%; SD, 2.4%). After 4 weeks of an IL-2 therapy schedule followed by 2 weeks of reconvalescence, the levels of FOXP3⁺ cells were measured again. Compared with pretherapy levels, FACS data showed a significant increase with an average of 11.6% of FOXP3 expressing cells (range, 5.1–17.3%; SD, 4.6%; P = 0.03). When Treg frequency was determined using the FOXP3 TSDR demethylation assay in the same samples, the average demethylation ratio prior to IL-2 therapy was at 3.4% (range, 2.4–4.2%; SD, 0.7%). After IL-2 therapy, Treg levels were significantly increased showing an average demethylation level of 8.8% (range, 6.9–11.5%; SD 1.6%; P = 0.03). From two of these patients, CD4⁺CD25^high Treg were isolated by magnetic beads (MACS) before and after IL-2 therapy. In both patients, CD4⁺CD25^high cells showed similar demethylation values (from 87.5% to 88.3% and from 85.2% to 93.9%; data not shown). Taken together, cell numbers as measured by FACS and FOXP3 TSDR demethylation correlate well (Pearson correlation R = 0.65, P = 0.023) and they seem to detect largely congruent cell populations (i.e., Treg). The previously reported increase of Treg upon IL-2 therapy was confirmed by two independent technological approaches.

FOXP3 TSDR demethylation in immunosuppressed, solid organ–transplanted patients. We counted Treg levels in transplantation patients treated with the immunomodulatory therapeutic antibodies thymoglobulin (rATG; three patients) and basiliximab (six patients) applying the FOXP3 TSDR demethylation assay (Fig. 3B). When demethylation prior to and after ATG treatment was analyzed, we found that Treg levels decreased in all three patients. Pretherapy levels ranged from 0.6% to 7.6% and dropped to values <0.25% in each case. Statistical significance could not be reached due to the low number of patients (n = 3) included. When demethylation prior to and after basiliximab therapy was analyzed, we found a significant decrease of Treg levels (P = 0.03) for all samples with an average pretherapy demethylation level of ∼1.6% (range, 0.6–4.3%), and an average posttherapy demethylation level of 0.66% (range, 0.02–2.2%). Taken together, we see a significant decline of Treg levels upon the described immunosuppressive therapies.

FOXP3 TSDR demethylation in peripheral blood of patients with cancer and healthy donors. Next, we tested if an up-regulation of Treg numbers was observed in the peripheral blood of patients suffering from solid tumors using the FOXP3 TSDR demethylation assay. For this, we measured the methylation status of the FOXP3 TSDR in the leukocyte fraction of a total of 115 blood samples from healthy controls, colorectal, lung, prostate, and breast cancer patients in a blinded study (for sample preparation and characteristics, see Materials and Methods). For healthy donors, the average Treg frequency was at 1.4%, ranging from 0.4% to 2.9%. For patients with colorectal cancer, the Treg concentration was at 2.3%, ranging from 0.6% to 3.5% with one outlier at 19%. Excluding the outlier, the average Treg frequency was at 1.6%, only slightly higher than for healthy donors. Treg concentration in lung cancer patients ranged from 0.5% to 4.4% with an average of 2.3%. Although results for patients with colorectal cancer did not significantly differ from normal controls, Treg frequencies in patients with lung cancer were significantly elevated (P = 0.003). ROC curves for colon and lung cancer show an AUC (see Materials and Methods: Statistical Analysis) of 0.66 for colon and an AUC of 0.75 for lung cancer (Fig. 4A). A separate analysis of Treg levels in male and female lung cancer patients and comparison with gender-matched controls shows similar AUC values (male lung cancer AUC = 0.73; female lung cancer AUC = 0.74); however, comparisons are no longer significant (male lung cancer P = 0.076; female lung cancer P = 0.071). When comparing Treg frequencies in breast and prostate cancer patients to healthy controls of the same sex, the average Treg concentration was elevated to 2.4% for both cancer types, compared with 1.7% in female and 1.4% in male normal controls. The changes, compared with normal individuals, were statistically significant for prostate cancer (P = 0.03) and slightly below significance for breast cancer (P = 0.08). ROC curves showed an AUC of 0.71 for discriminating between normal females and those with breast cancer and an AUC of 0.73 for discriminating between normal males and those with prostate cancer (Fig. 4B).

FOXP3 TSDR demethylation in FFPE tissue samples. Finally, we asked if measurement of FOXP3 TSDR demethylation might resolve the technical issues encountered with current methods to detect Treg in solid tissues. We selected FFPE samples from non–small cell lung and colon carcinoma and corresponding adjacent normal tissue. On the one hand, samples were measured using immunohistochemistry, detecting the number of FOXP3⁺ cells in tissue sections per HPF (Fig. 5A). On the other hand, we used material from the same paraffin blocks for Treg quantification by measurement of FOXP3 TSDR demethylation. When using immunohistochemistry, higher concentrations of Treg were found in colorectal cancer tissue (average immunohistochemistry cell number, 15.3 Treg/HPF; range, 4.4–52 Treg/HPF) when compared with normal adjacent colon tissue (average immunohistochemistry cell number, 0.3 Treg/HPF; range, 0.1–0.6 Treg/HPF). When the concentration of Treg in colorectal cancer tissue was analyzed using FOXP3 TSDR demethylation assay, an average demethylation of 6.3% was observed, ranging from 1.9% to 12.3%. In healthy control tissue, on average, 1.5% Treg were measured ranging from 0.8% to 2.8%. For both technologies, the tumor-specific increase of Treg infiltration is statistically highly significant (P < 0.001). For lung tissue, immunohistochemistry measurements showed an increase of Treg in tumor tissue (average immunohistochemistry cell number, 6.8 Treg/HPF; range, 3.9–10.1 Treg/HPF) compared with normal lung tissue (average immunohistochemistry cell number, 0.2 Treg/HPF; range, 0.1–0.4 Treg/HPF). Measured by FOXP3 TSDR demethylation analysis, average Treg levels were at 3.8%, ranging from 1.6% to 6.4% for cancerous tissue and average demethylation levels were at 0.7% for healthy tissue, ranging from 0.5% to 0.8%. Despite the limited amount of lung tissue samples included in this study, the observed changes are statistically significant (P = 0.036 for immunohistochemistry and FOXP3 TSDR demethylation). Further statistical analysis showed that the determination of FOXP3⁺ cells by immunohistochemistry and FOXP3 TSDR demethylation analyses are highly correlated (Pearson correlation R = 0.80; P < 0.001; Fig. 5B). To compare the performances of the technologies, we used the nonparametric test of DeLong (38). Immunohistochemical analysis showed a perfect separation between normal and cancerous tissues (AUC = 1.0). FOXP3 TSDR methylation analysis showed an almost perfect separation (AUC = 0.96). Based on these two AUC values, there is no significant difference between the immunohistochemical and the FOXP3 TSDR demethylation methods when analyzing normal control versus cancer tissue distinction (P = 0.17). Thus, the tissue data show that both measurement methods are equivalent as far as distinction is concerned.

Over the last decade, Tregs have become a central focus of immunologic research and an important player in tumor immunity (1, 2, 15). In order to understand the role of Treg in disease settings, it is of paramount importance to be able to specifically detect and quantify the number of Tregs present in peripheral blood or disease-inflicted tissue. Once possible, Treg counts might serve as an important diagnostic and prognostic variable. However, current detection methods for Treg do not allow for exact quantification due to the absence of cell type specificity in the FOXP3 and CD25 mRNA and protein markers used (26, 27). All evidence from our previous studies (28, 29) suggested that demethylation of the FOXP3 TSDR is exclusive for stable bona fide Treg. Thus, its measurement seemed highly suitable for the establishment of a quantitative Treg assay. When transformed onto a qPCR platform (Fig. 1), FOXP3 TSDR demethylation showed high biological specificity to Treg when all major leukocyte fractions were analyzed. Residual FOXP3 TSDR demethylation in cell types other than Treg, including B cells, is most likely attributed to incomplete purification by the flow cytometric sorting schemes used (Table 1). Physiologic Treg levels are assumed to range between 0.5% and 5% in the peripheral blood and at ∼0.1% in healthy other tissues (27, 39, 40). Thus, the observed detection limit of the FOXP3 TSDR demethylation assay (0.03%) qualifies for accurate Treg measurements in all clinical settings (Fig. 2). Although our analyses prove specificity within the blood components, eventually, it will be necessary to analyze the epigenetic profile of further, nonhematopoietic cell types to show the universal applicability of this system. This is particularly significant because recent data indicate that FOXP3 mRNA expression has been observed to act as a tumor suppressor in non-Treg in solid tissues (41), a fact that might or might not be reflected in the epigenetic pattern. The universal applicability of this assay might also be impaired by differences between male and female donors. The straight correction by a factor of 2, which we used to compensate for the X-inactivation, does not take into account possible biological differences between the two sexes, such as increased tolerance of foreign antigens during pregnancy, which might require elevated Treg levels (42). In awareness of the specifications and limitations referred to, we intended to use the newly developed assay in a number of possible routine applications in clinical settings.

We measured blood samples from melanoma patients that had received IL-2 treatment. Although aiming at enhancing NK and T cell responses, recent observations suggested a significant increase of Treg numbers following IL-2 treatment (13, 16). These observations ignited a debate as to whether this treatment would be beneficial (due to the intended enhancement of NK and T effector cells) or adverse (due to enhancement of the immunosuppressive activity of Treg) to cure. Complicating this debate is the question of whether the IL-2–induced FOXP3⁺ T cells are bona fide Tregs. Data about their suppressive function are inconsistent (14–16) and transient expression of FOXP3 is also an established property of activated effector T cells (24). Our data now confirm that cells with multiple Treg characteristics including CD25 and FOXP3 expression and FOXP3 TSDR demethylation are up-regulated after IL-2 therapy (Fig. 3A). Although still falling short of a functional test, these findings reiterate those concerns that IL-2 therapies stimulate natural Treg (16).

A converse treatment strategy is used for immunosuppression of transplantation patients. Transplantation patients receiving rATG showed a dramatic decrease of Treg (Fig. 3B), as would be expected from a therapy that unselectively depletes T cells. We also see a—less pronounced, but consistent for all tested patients—decrease of Treg following basiliximab therapy (Fig. 3B). The latter observation is unexpected because previous studies do not report a Treg decrease. Basiliximab's mode of action has been described as solely blocking the IL-2α receptor (CD25) without initiating cell depletion (43, 44). We cannot pinpoint the observed decrease of Treg on basiliximab exclusively because the therapy regimen includes other potent immunosuppressive factors, such as calcineurin inhibitors (tacrolimus), antiproliferate agents (mycophenolate mofetil), and corticosteroids (prednisone). At this point in time, we do not speculate on effects leading to or caused by the observed decrease of Treg during basiliximab treatment. Instead, our rationale behind the analyses performed was to exemplarily show that the qPCR assay is capable of monitoring the patients' Treg status. Taken together, we show that the effects of both, immunostimulatory and immunosuppressive therapy on Treg levels can be monitored using the presented qPCR assay.

We then investigated the idea that the FOXP3 TSDR demethylation assay might provide a new, early indicator for the body's response to the existence of solid tumors, which can be analyzed in peripheral blood. We tested this hypothesis by conducting a blinded study that included patients with breast, prostate, colon, and lung tumors and compared those with healthy donors. In line with recent reports (39, 40), we showed that in blood from patients with lung and prostate cancers, significantly higher signals of Treg, i.e., demethylated FOXP3 TSDR, were found compared with healthy controls. For patients with breast cancer, the Treg frequency determined by the demethylation assay was scarcely nonsignificant (P = 0.08) but still distinctly higher than in controls (Fig. 4). The results of FOXP3 TSDR demethylation analysis in the peripheral blood of patients with colon cancer differ from previous results obtained by FACS, which showed an increase of FOXP3⁺ cells compared with normal controls (45, 46). These differences might reflect the fact that the demethylation assay exclusively detects natural Treg. This is in contrast with the FACS analysis, which also detects other fractions of T cells transiently expressing FOXP3. Thus, if we speculate that all tumors take advantage of some suppressive immune function, it may be that the FOXP3 TSDR demethylation assay falls short of identifying important other (transiently) FOXP3-expressing suppressive cells. Those cells may govern tumor immunity in colorectal cancer. Further research on both natural Treg and those other suppressive cell types is required. The search for epigenetic detection systems for these other cell types, including Tr1 cells, is ongoing in our laboratory.

Analysis of a larger number of patients and controls may clarify its significance to breast cancer and reconfirm those for lung and prostate cancer. Because it is assumed that markers which identify tumor patients with an AUC value of >0.7 are considered as potential diagnostic tools (47), FOXP3 TSDR demethylation could serve as an early detection marker for breast, lung, and prostate cancers. In our view, two major conclusions can be drawn from the results presented here: (a) this study encourages the idea to monitor Treg levels as variables for malignant tumor development in the body or the risk thereof; (b) further extensive studies are required for the various cancer entities and therapeutic approaches in order to validate FOXP3 TSDR demethylation as a marker in the various settings.

We also investigated the applicability of the FOXP3 TSDR demethylation assay as an alternative for the laborious and sometimes imprecise and biased immunohistochemistry analyses currently used for the measurement of Treg in fixed tissue samples (Fig. 5). Our analyses showed that the performance of both technologies is equivalent as far as distinction of cancer compared with healthy tissue is concerned. When taking into account that FOXP3 TSDR demethylation is fully quantitative, performed without field/slide selection bias, is technically less demanding and independent of antibody performances, it seems to be the adequate tool for routine applications. Consequently, we believe that the low technical demands of the assay here will allow a significant acceleration of research and standardization of routine testing of tumor-infiltrating Treg. Biologically, our data obtained with both FOXP3 TSDR demethylation and immunohistochemistry assays confirm an accumulation of Treg in lung and colon carcinoma tissue compared with their healthy adjacent tissues. This supports previous findings which suggest that an increased prevalence of Treg in the microenvironments of solid epithelial tumors (48) as reported for invasive breast, pancreas (6), hepatic (5), and lung (49) tumors. The implementation of a precise and routine testing of solid tissues for Treg becomes particularly important because there are various reports which suggest Treg as an important prognostic factor for disease outcome, as shown for ovarian cancer or breast cancer (11, 12).

In conclusion, we believe that the presented assay for quantitative Treg detection will, in the future, lead to better interstudy comparability, and faster and more efficient analysis even of FFPE tissues. Furthermore, our findings render the measurement of Treg frequency by analysis of the FOXP3 TSDR methylation status a prime target not only for early tumor recognition but also for monitoring antitumor strategies.

U. Baron, U. Hoffmüller, and S. Olek: Ownership interest, Epiontis. A. Hamann: Ownership interest, Charité. The other authors disclosed no potential conflicts of interest.

Grant support: Investitionsbank Berlin; the Wilhelm Sander Foundation and by the Deutsche Forschungsgemeinschaft (SFB633 and SFB650; J. Huehn).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Alexander Hellwag for technical assistance, Michael Weber for advice, and Dirk Nargorsen for providing sample information. We owe Alex Olek for making our concept thinkable.