A unique subset of CD4+CD25highFoxp3+ T cells secreting interleukin-10 and transforming growth factor-beta1 mediates suppression in the tumor microenvironment.

Purpose: Immunosuppression, including that mediated by CD4+CD25highFoxp3+ regulatory T cells (Treg), is a characteristic feature of head and neck squamous cell carcinoma (HNSCC). Tregs with a distinct phenotype in tumor-infiltrating lymphocytes (TIL) contribute to local immune suppression. Experimental Design: The frequency and phenotype of Treg in TIL and/or peripheral blood mononuclear cells (PBMC) in 15 HNSCC patients and PBMC in 15 normal controls were compared. Single-cell sorted CD4+CD25high T cells were tested for regulatory function by coculture with carboxyfluorescein diacetate succinimidyl ester–labeled and activated autologous CD4+CD25− responder T cells. Transwell inserts separating Treg from responders and neutralizing interleukin-10 (IL-10) or transforming growth factor-β1 (TGF-β1) antibodies were used to evaluate the mechanisms used by Treg to suppress responder cell proliferation. Results: In TIL, CD25+ cells were enriched in the CD3+CD4+ subset (13 ± 3%) relative to circulating CD3+CD4+ T cells (3 ± 0.7%) in HNSCC patients (P ≤ 0.01) or normal controls (2 ± 1.5%; P ≤ 0.001). Among the CD3+CD4+ subset, CD25high Treg represented 3 ± 0.5% in TIL, 1 ± 0.3% in PBMC, and 0.4 ± 0.2% in normal controls. Tregs in TIL were GITR+, IL-10+, and TGF-β1+, although circulating Treg up-regulated CD62L and CCR7 but not GITR, IL-10, or TGF-β1. Treg in TIL mediated stronger suppression (P ≤ 0.001) than Treg in PBMC of HNSCC patients. The addition of neutralizing IL-10 and TGF-β antibodies almost completely abrogated suppression (5 ± 2.51%). Transwell inserts partly prevented suppression (60 ± 5% versus 95 ± 5%). Conclusions: Suppression in the tumor microenvironment is mediated by a unique subset of Treg, which produce IL-10 and TGF-β1 and do not require cell-to-cell contact between Treg and responder cells for inhibition.

The role of tumor-infiltrating lymphocytes (TIL) in patients with cancer has been debated for many years. A considerable body of evidence suggests that TIL contribute to tumor escape from the immune system (1 -3). In many cases, however, TIL accumulations have been linked to improved patient outcome and TIL-T recognized as effectors of antitumor immune responses (4 -8). Recent studies have revealed that a subset of CD4 + T cells, referred to as CD4 + CD25 high Foxp3 + naturally occurring regulatory T cells (nTreg), may accumulate in the tumor environment and suppress tumor-specific T-cell responses, thereby hindering tumor rejection. In patients with cancer, accumulations of Treg cells in the tumor were associated with a significant reduction in survival (9,10).
To date, three types of CD4 + Treg cells have been partly characterized in humans: (a) CD4 + CD25 -IL-10 + Foxp3 low type 1 T regulatory (Tr1) cells, which arise in the periphery on encountering antigen in a tolerogenic environment via a process that is interleukin (IL)-10 dependent (11,12); (b) naturally occurring CD4 + CD25 high Foxp3 + T cells (nTreg), which arise directly in the thymus and have the ability to suppress responses of both CD4 + CD25and CD8 + CD25 -T cells in a contact-dependent, cytokine-independent, and antigennonspecific manner (13 -15); and (c) Th3 cells, which are dependent on IL-4 for functional differentiation (16). We have previously described an enrichment of CD4 + CD25 high T cells (Treg) among tumor-infiltrating (17) as well as circulating lymphocytes in patients with head and neck squamous cell carcinoma (HNSCC; refs. 18,19). The presence of Treg at the tumor site may be detrimental to the host defense against the tumor. It could inhibit antitumor immune responses and interfere with immunotherapeutic strategies.
The inconsistent reports in the literature about the role of TIL in the control of cancer progression in humans seem to result from the fact that phenotypic rather than functional attributes of these cells are used for their identification and characterization. Predictions of tumor behavior solely based on numbers and tumoral or peritumoral localization of TIL are likely to be inaccurate, as it is the functional potential of immune cells that determines their effectiveness as antitumor effectors in situ. Functions of TIL are difficult to assess in vitro because of the necessity for their isolation from tissues and purification, which are labor intensive and require large tumor specimens. For these reasons, studies of TIL, including Treg subpopulations, have been largely confined to in situ examinations generally involving immunocytochemistry (4,20).
Treg contributions to shaping of the tumor microenvironment and to regulation of antitumor responses in situ can only be evaluated after their isolation from tumors. To this end, we describe here phenotypic and functional characteristics of Treg isolated from TIL in HNSCC. We show that Treg accumulating in tumors express a unique phenotype distinct from that of Treg in peripheral blood mononuclear cells (PBMC), and they are more suppressive. Furthermore, their presence in the tumor is linked to a poor prognosis in subjects with HNSCC.

Materials and Methods
Patients and tumor tissues. Patients with HNSCC included in this study (n = 24) were initially seen at the Outpatient Otolaryngology Clinic at the University of Pittsburgh Medical Center. The patients who underwent surgery signed the Institutional Review Board -approved informed consent releasing the use of specimens (tissues and blood) for research purposes. Tumors were obtained from 15 patients undergoing surgery for treatment of primary disease between September 2000 and March 2001. The tissues were processed and banked at the University of Pittsburgh Cancer Institute Tissue Procurement Facility. Samples of PBMCs were obtained before surgery from five of these patients. In addition, PBMCs were donated by nine patients seen between January and October 2006 who had active disease (six had primary and three had recurrent disease). The patients who donated PBMC were males (n = 6) and females (n = 3) with a mean age of 60 years (range, 23-82). The oncologic therapy, if given, was terminated from 3 weeks to 2 months at the time of phlebotomy for this study. In addition, samples of PBMC were obtained from 15 age-matched normal controls. The age, sex, and clinicopathologic characteristics of the patients whose TILs were studied are listed in Table 1.
Collection of PBMC. Peripheral venous blood (20-30 mL) was drawn into heparinized tubes. The samples were hand carried to the laboratory and immediately centrifuged on Ficoll-Hypaque gradients. PBMCs were recovered, washed in AIM-V medium (Invitrogen), counted in the presence of a trypan blue dye to evaluate viability, and either immediately used for experiments or cryopreserved.
Isolation of TIL. TIL were isolated at the University of Pittsburgh Cancer Institute Tissue Procurement Facility according to a standard operating procedure as described previously (21). Briefly, after removal of fat, blood, or necrotic areas, tumor tissues were washed in RPMI 1640 containing 50 Ag/mL gentamicin (Invitrogen) and cut into 1 mm 3 pieces in a Petri dish while covered with RPMI 1640 plus antibiotics. Tumor fragments were transferred to flasks and dissociated using 0.05% collagenase (type IV; Sigma) and 0.02% DNase (type I; Sigma) in the same medium supplemented with 5% (v/v) fetal bovine serum (Invitrogen). The digest was passed through 90-Am and 50-Am nylon mesh to remove clumps, and the filtrate was washed two to three times in medium followed by centrifugation at 350 Â g for 10 min. To separate tumor cells from lymphocytes, the cell suspension was layered onto a discontinuous Ficoll-Paque gradient of 75% over 100% Ficoll-Paque in medium and centrifuged for 800 Â g for 20 min at room temperature (21). Lymphocytes were collected from the interface between 75% and 100% Ficoll-Paque and washed twice before further use.
Cryopreservation of TIL and PBMC. TIL and autologous or allogeneic PBMCs were cryopreserved at a cell concentration of up to 20 Â 10 6 cells/mL in a freezing medium consisting of RPMI 1640 supplemented with 20% (v/v) human AB serum (Nabi) and 10% (v/v) DMSO (Fisher Scientific). Cells were cooled at a rate of 1jC per min to -80jC (Cryomed) and stored in liquid nitrogen. Before use, cells were thawed at 37jC and immediately resuspended in an excess of warm AIM-V medium. Cells were washed twice in the same medium equilibrated for 2 to 4 h, adjusted to the concentration of 5 Â 10 6 /mL, and used for experiments.
Antibodies. The following anti-human monoclonal antibodies Before use, all mAbs were titrated using normal resting or activated PBMC to establish optimal staining dilutions.
Surface and intracellular staining. To determine the frequency of CD4 + CD25 high T cells and the ''Treg'' surface marker profile, TIL or PBMC (at least 0.2 Â 10 6 cells/tube) were stained with mAbs in the above-described panel for 15 min at 4jC. Appropriate isotype antibody controls were used for each sample. Cells were washed and examined by four-color flow cytometry. Intracellular staining for Foxp3, CD152 (CTLA-4), and GITR was done as described previously (22). Intracytoplasmic expression of transforming growth factor-h1 (TGF-h1) and IL-10 was assessed before and after stimulation of PBMC for 4 h with phorbol 12-myristate 13-acetate (20 Ag/mL) and ionomycin (1 Amol/L/mL). Briefly, samples were first incubated with mAbs against surface markers CD4, CD3, and CD25. After extensive washing, cells were fixed with 4% (v/v) formaldehyde in PBS for 20 min at room temperature, washed once with PBS containing 0.5% (v/v) bovine serum albumin and 2 nmol/L EDTA, permeabilized with PBS containing 0.5% bovine serum albumin and 0.1% (v/v) saponin, and stained with anti-CTLA-4-PE, anti-Foxp3-FITC, anti-TGF-h1-FITC, or anti-IL-10-PE mAb for 30 min at room temperature. Cells were further washed twice with PBS containing 0.5% bovine serum albumin and 0.2% (v/v) saponin, resuspended in fluorescence-activated cell sorting flow solution, and immediately analyzed by flow cytometry. Appropriate isotype controls were included for each sample.
Flow cytometry. Flow cytometry was done using a FACScan flow cytometer (Beckman Coulter) equipped with Expo32 software (Beckman Coulter). The acquisition and analysis gates were restricted to the lymphocyte gate as determined by characteristic forward and side scatter properties of lymphocytes (Fig. 1A). Forward and side scatters were set in a linear scale. For analysis, 1 Â 10 5 lymphocytes were acquired. The analysis gates were restricted to the CD3 + CD4 + (e.g., Fig. 1B), CD4 + CD25 + , and CD4 + CD25 high T-cell subsets (e.g., Fig. 1C), as appropriate. Cells expressing Treg markers were acquired and analyzed in the FL1 or FL2 logarithmic scale using the set gates.
Suppression experiments. Single-cell sorting of CD4 + CD25 high T-cell populations was done as described previously (22). Treg obtained from TIL or PBMC were tested for regulatory function by coculture analysis at the suppressor to responder ratios of 1:2 or 1:5 with at least 0.5 Â 10 5 carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled autologous CD4 + CD25 -T responder cells per well. Soluble OKT3 (10 Ag/mL; American Type Culture Collection) and soluble anti-CD28 mAb (1 Ag/mL) were used for stimulation in the presence of 150 IU/mL IL-2 for 5 days.
CFSE labeling of responder T cells was done as described previously (22). Briefly, CD4 + CD25 -T cells separated by single-cell sorting were stained with 1.5 Amol/L CFSE (Molecular Probes/Invitrogen) for 10 min at room temperature. The CFSE label was quenched by the addition of an equal volume of FCS (Invitrogen), and then cells were washed extensively with PBS.
T-cell populations were classified as suppressive if they inhibited proliferation of autologous CD4 + CD25responder cells in the coculture assay and if decreasing the number of CD4 + CD25 + T cells relative to the number of CD4 + CD25responder cells in coculture restored proliferation. CD4 + CD25 + T-cell populations that satisfied both of these criteria were classified as suppressor T cells. These criteria were applied to all populations tested.
All CFSE data were analyzed using the ModFit software provided by the Verity Software House. The percentages of suppression were calculated based on the proliferation index of responder cells alone compared with the proliferation index of cultures containing responder plus Treg. The program determines the percentage of proliferating cells within each peak, and the sum of all peaks in the control culture (no Treg) is taken as 100% of proliferation and 0% of suppression.
Transwell assays. To assess whether cell-to-cell contact was necessary for Treg to mediate suppression, polycarbonate 24-well Transwell inserts (0.4 Amol/L; Corning Costar Corp.) were used. At least 5 Â 10 4 CFSE-labeled CD4 + CD25 -T cells were stimulated with soluble OKT3 and soluble anti-CD28 mAb (each at 1 Ag/mL) in the presence of 150 IU/mL IL-2 in the lower chambers of the plates. Autologous Treg or control cells (CD4 + CD25 -) were added to the upper chambers at the suppressor to responder ratio of 1:2. To analyze whether TGF-h1 and IL-10 are required for Treg-mediated suppression, neutralizing IL-10 antibody (2 Ag/mL; R&D Systems) and/or TGF-h1 antibody (20 Ag/mL; R&D Systems) or nonblocking IgG anti-mouse control antibody were added to Treg in the upper chambers. Cells were cultured for 5 days and tested for suppression of responder cell proliferation.
In the case of Foxp3, IL-10, or TGF-h1, sections were incubated with the secondary antibodies under the same conditions. Primary antibodies were omitted in all negative controls. Sections were incubated in a medium with 4 ¶,6-diamidino-2-phenylindole (DAPI; Vector Laboratories) to trace cell nuclei. Slides were evaluated in a Nikon Eclipse E-800 fluorescence microscope under Â200 magnification. For digital image analysis, the software Adobe Photoshop 6.0 version was used. Statistical analysis. Differences between groups were assessed using Wilcoxon test with P values adjusted by resampling (2,000 permutations). Two-tailed adjusted P values of <0.05 were considered to be significant.

Results
Phenotypic characteristics of TIL and PBMC in HNSCC patients. Flow cytometry analysis of TIL isolated by enzymatic digestion from tumor tissues showed that the CD3 + CD4 + T-cell subset accounted for 44 F 15% (mean F SD) of all T cells, with a range of 30% to 60%. The gating strategy used to identify this cell subset is shown in Fig. 1A to C. Among these CD4 + T cells, an average of 13 F 3% was CD25 + (Fig. 1D). Immunohistochemistry of HNSCC tissues (n = 15) confirmed the prominent presence of CD4 + CD25 + T cells in tumor infiltrates (Fig. 2). This enrichment in CD4 + CD25 + T cells was significant (P V 0.001) compared with circulating CD3 + CD4 + T cells (3 F 0.7%) in HNSCC patients or in normal controls (2 F 1.5%). We also determined the proportions of CD25 high cells (Treg) present in the CD4 + CD25 + subset in TIL (3 F 0.5%; n = 15), and in PBMC (1 F 0.3%; n = 14) of HNSCC patients as well as normal controls (0.4 F 0.2%; n = 15), as illustrated in Fig. 3. Our data showed that the mean percentages of CD4 + CD25 + as well as CD4 + CD25 high T cells were significantly higher in TIL and in PBMC of HNSCC patients (P V 0.001 and 0.01, respectively) compared with the mean percentage of these cells in PBMC of normal controls.
For this analysis, CD4 + T cells with a mean fluorescence intensity (MFI) of CD25 expression z120 were classified as CD25 high (Fig. 1C). This cutoff point was selected based on our earlier observation that only CD4 + CD25 + T cells with MFI z120 had suppressor activity after single-cell sorting of normal control PBMC. CD4 + CD25 + T cells with MFI <120 for CD25 did not suppress autologous or allogeneic responder populations and probably consisted of activated T effector cells. An average MFI for CD4 + CD25 high cells in TIL was 160 F 25, whereas that for CD4 + CD25 high cells in PBMC was 125 F 12. CD4 + TIL expressed higher levels of CD25 than did circulating CD4 + T cells.
Phenotypic characteristics of CD4 + CD25 high T lymphocytes in TIL and PBMC. We next analyzed T-cell activation and Treg markers as well as IL-10 and TGF-h1 (membrane bound and intracellular) on gated CD4 + CD25 high T cells in isolated TIL and PBMC of HNSCC subjects or normal controls by flow cytometry. No significant differences were observed in percentages of CD4 + CD25 high T cells positive for various Treg markers between the PBMC autologous with five TIL samples and allogeneic PBMC (n = 9) in HNSCC patients with active disease  (P z 0.05; data not shown). Therefore, we were able to compare the Treg phenotype in TIL (n = 15) with that in normal controls (Fig. 4) or in combined autologous and allogeneic PBMC (n = 14) of HNSCC patients (Fig. 5).
Backgating on CD4 + CD25 high T cells, we observed that these cells in TIL and PBMC of HNSCC patients were largely Foxp3 + and CTLA-4 + , with the majority expressing Fas (CD95; Fig. 5A). Expression of Foxp3 on CD4 + CD25 + TIL was also determined on frozen sections of tumors by immunofluorescence (Fig. 6). Staining in situ confirmed the presence of numerous Foxp3 + CD25 + T cells in the tumor (Fig. 6A). Several other T-cell markers were differentially expressed on CD4 + CD25 high T cells in TIL compared with this subset of T cells in PBMC of HNSCC patients. For example, most CD4 + CD25 high T cells in TIL were GITR + , FasL + , IL-10 + , and TGF-h1 + (Figs. 5A and C and 6) relative to CD4 + CD25 high T cells in PBMC (P V 0.0015), only few of which were positive for GITR and all of which were negative for FasL, IL-10, and membrane-bound or intracellular TGF-h1 (Fig. 5). Intracytoplasmic expression of IL-10 in CD25 + TIL (Fig. 6A) or TGF-h1 in CD4 + CD25 + TIL (Fig. 6B) in situ was clearly shown by immunofluorescence. Tumor sections were also stained for CD11c, CD4, or CD25 and IL-10 to determine whether IL-10 was expressed by tumorassociated macrophages/dendritic cells or Tregs. We observed that most IL-10 + cells were also CD4 + CD25 + but not CD11 + (data not shown). Further, as shown in Fig. 6B, TGF-h1 + TILs  were also CD4 + CD25 + . These in situ data corroborate flow cytometry results with isolated TIL indicating that 80% to 100% of CD4 + CD25 high cells were positive for intracytoplasmic IL-10 and TGF-h1. As shown in Fig. 5C, TIL expressed IL-10 and TGF-h1 without in vitro activation, although, following stimulation with phorbol 12-myristate 13-acetate and ionomycin, expression of these two cytokines intensified. A significantly higher percentage of CD4 + CD25 high T cells was positive for the migration and lymph node homing receptors CCR7 and CD62L in PBMC of HNSCC patients than in TIL (P V 0.0015; Fig. 5A). Another chemokine receptor, CCR4, was found on nearly all CD4 + CD25 high T cells in both PBMC of HNSCC patients and TIL (Fig. 5A) but on only f40% of CD4 + CD25 high T cells in normal controls (Fig. 4). CD4 + CD25 high T cells were CD45RO + in TIL and PBMC of HNSCC patients (Fig. 5A) and normal controls (Fig. 4), indicating that Treg in humans largely belong to the memory T-cell compartment as suggested previously (23). Few CD4 + CD25 high T cells were positive for CD45RA (5.7 F 2.3%) or CD127 (12 F 15%; data not shown).
Because only 60% of CD25 high Treg were positive for CD45RO, and the remaining 40% cells did not express Fig. 6. Immunofluorescence for CD25, Foxp3, IL-10, GITR, or TGF-h1in sections of HNSCC. A, TIL positive for CD25 and Foxp3, IL-10, GITR, orTGF-h1on cryostat sections of a representative tumor (out of 15 examined). a, negative control using DAPI and secondary mouse anti-biotin FITC-labeled antibody alone. Nuclei are stained with DAPI. b, CD25 + (red) and Foxp3 + (green) cells with DAPI; note Foxp3 expression in CD25 + TIL. c, CD25 + (green) and IL-10 + (red) cells with DAPI; note coexpression of CD25 and IL-10 in numerousTIL (yellowish). d, CD25 + (red) and GITR + (green) cells with DAPI. Coexpression of CD25 and GITR in manyTIL (yellow). Magnification, Â500. B, a,TIL positive for CD4 + (green) and CD4 + CD25 + (yellow); b, same tumor area stained forTGF-h1. Nuclei are stained with DAPI. c, an overlay of a + b with CD4 + CD25 + TGF-h1 + cells stained pink. Magnification, Â400. CD45RA, we suspected that CD25 high T cells (Treg) might selectively down-regulate CD45RA. Therefore, gating on CD4 + CD45RO low , CD4 + CD45RO high , CD4 + CD45RA low , and CD4 + CD45RO high T-cell subsets, we analyzed CD25 expression in these subsets. Interestingly, all CD4 + CD45RO low T cells were also CD45RA low (MFI V 50) and CD25 low/interm (MFI V 50). In contrast, most cells in the CD4 + CD45RO high subset were CD25 high (MFI V 165) but expressed no CD45RA. All of the CD25 high T cells (Treg) were negative for CD45RA, whereas the CD25 low/interm T cells were CD45RA low (MFI V 80). These results suggest that (a) activated (CD25 + ) non-Treg gain CD45RO expression and down-regulate CD45RA and (b) close to half of the CD25 high Treg loose CD45RO expression but do not gain CD45RA. Whether this loss of CD45RO expression by Treg is related to the state of their differentiation or activation at the tumor site or to their sensitivity to apoptosis is presently unknown.
Compared with CD4 + CD25 high Treg present in the peripheral circulation of normal controls, those in TIL and PBMC of HNSCC patients had significantly increased percentages of Foxp3, CTLA-4, and Fas (P V 0.001; Figs. 4 and 5). FasL was absent from the surface of PBMC in patients (Fig. 5) and normal controls (Fig. 4). Although GITR was expressed on f80% of CD4 + CD25 high cells in TIL, in PBMC of HNSCC patients, CD4 + CD25 high T cells expressed little GITR (Fig. 5). In addition, in TIL (Fig. 5), close to half of CD4 + CD25 high Treg up-regulated CD132 (IL-2Rg). However, CD132 was not upregulated on CD4 + CD25 high T cells in PBMC of HNSCC patients (Fig. 5).
These data indicate that considerable phenotypic differences exist between the CD4 + CD25 high populations in TIL and PBMC of HNSCC patients. Only TILs express GITR, secrete IL-10 and TGF-h1, and up-regulate CD132. Likewise, circulating CD4 + CD25 high Treg in PBMC of normal controls are phenotypically distinct from those in PBMC as well as TIL in patients with cancer.
Suppressor activity of CD4 + CD25 high T cells isolated from TIL or PBMC of HNSCC patients. To evaluate suppressor activity of CD4 + CD25 high T cells in TIL or PBMC, single-cell sorting was used for their isolation followed by cocultures with CFSElabeled autologous CD4 + CD25 -T cells as responders. The cocultures were set up at the suppressor to responder ratios of 1:2 or 1:5 using responder proliferating in response to OKT3 and anti-CD28 antibodies in the presence of IL-2. To analyze proliferation of responder cells, gating was set to include CD4 + CFSE + T cells (Fig. 7A). The percentage of inhibition measured varied depending on the suppressor to responder ratios used and the Treg origin. Strong suppression of proliferation (P V 0.001) was observed using suppressor to responder ratios of 1:2 for Treg isolated from TIL or PBMC of HNSCC patients (Fig. 7B). At the suppressor to responder ratio of 1:5, only Treg isolated from TILs were suppressive (Fig. 7B). In contrast, CD4 + CD25 high T cells isolated from the circulation of normal controls only minimally suppressed proliferation (Fig. 7B). This result confirms that the CD25 + T-cell subset in normal controls is mainly composed of activated CD25 + non-Treg as previously reported by us (22). On the other hand, CD4 + CD25 high TILs are highly suppressive, and suppression mediated by Treg present in TIL is significantly higher (P V 0.044) than that mediated by the CD4 + CD25 high subset in PBMC of patients with HNSCC.
Treg in TIL mediate suppression of autologous T cells via IL-10 and TGF-b1. A large majority (75-96%) of CD4 + CD25 high Foxp3 + T cells in TIL isolated from HNSCC patients were positive for intracytoplasmic IL-10 and TGF-h1 by flow cytometry and in situ by immunofluorescence (Figs. 5 and 6). As described above, these cells mediated stronger suppression relative to Treg isolated from PBMCs that were negative for IL-10 and TGF-h1. Therefore, we hypothesized that Treg function in TIL was mediated via IL-10 and TGF-h1 secretion and did not require cell-to-cell contact. To test this hypothesis, CD4 + CD25 + Tregs isolated from TIL were coincubated at the suppressor to responder ratio of 1:2 with CFSE-labeled autologous CD4 + CD25 -T cells responding to OKT3 and anti-CD28 antibodies in the presence or absence of Transwell inserts that permit diffusion of soluble factors but prevent cell-to-cell contact.
Coincubation in the presence of Transwell inserts resulted in significant reduction of suppression (P V 0.03) as shown in Fig. 8A. Nevertheless, in the presence of Transwell inserts, suppression levels remained at 60%, suggesting that cell-to-cell contact was only partly necessary. When, however, Treg in the upper compartment were treated with both neutralizing antibodies to IL-10 and TGF-h1, suppression was completely abrogated (P V 0.0001; Fig. 8A). Further, the addition of neutralizing anti-IL-10 or TGF-h1 antibodies to Treg in the absence of Transwell inserts did not reduce suppression beyond the level seen with the insert alone (i.e., f50%; Fig. 8B). These results suggest that not only cell-to-cell contact but also soluble factors (cytokines) are involved in suppression. However, in the absence as well as presence of cell-to-cell interaction, blocking of both IL-10 and TGF-h1 was necessary to completely abrogate suppression ( Fig. 8A and B). In aggregate, these data suggest that suppression is mediated by soluble factors produced by Treg and able to diffuse through the membrane when cellto-cell contact is abrogated. However, cell-to-cell contact contributes to suppression mediated by Treg-derived cytokines.
The role of CD4 + CD25 high T-cell subsets in tumor progression. Most patients (n = 12) whose tumors were used for TIL isolation had advanced disease (stage T 3 /T 4 ). The tumors were infiltrated by CD3 + CD4 + CD25 + Foxp3 + T cells (Figs. 3 and 5) as determined by immunofluorescence of frozen tumor sections. These Treg were strongly suppressive and produced IL-10 and TGF-h1. Thirteen of 15 of the HNSCC patients died of their disease within 3 to 4 years of surgery. Our observations suggest that the presence at the tumor site of CD4 + CD25 high Foxp3 + T cells with strong suppressor functions seems to be a characteristic feature of T 3 /T 4 stage tumors and might be associated with a poor prognosis.

Discussion
Several studies have reported that accumulation of TIL at the tumor site is associated with a better prognosis (5 -8). However, other studies show that infiltrating CD4 + T cells can favor the tumor to the detriment of the host (10). Specifically, the subpopulation of CD4 + CD25 high Foxp3 + T cells referred to as Treg may impair the ability of the host immune system to defend against tumor progression (10,24). An improved understanding of the phenotype and function of Treg within TIL could help establish a more reliable prognostic role of TIL in patients with solid tumors. Further, such an understanding is necessary to foster the development of targeted approaches that would eliminate Treg and perhaps specifically augment antitumor immune responses. Indeed, our study suggests that the type and functional attributes of tumor-infiltrating CD4 + T cells, but not their quantity or localization, may prove to be the most critical prognostic determinants.
In contrast to our results, Badoual et al. (4) recently showed that the percentage of CD4 + Foxp3 + T cells positively correlated with a better prognosis in patients with HNSCC. In addition, recent studies in follicular lymphoma (25,26) and Hodgkin's lymphoma (27) showed that the presence of tumor-infiltrating CD4 + Foxp3 + Treg cells was associated with a better patient survival. Results of these studies are counterintuitive, and it is important to note that, in these studies, only the in situ frequency of CD4 + Foxp3 + T cells was correlated with clinical outcome (4,25,27). None examined the phenotypic characteristics and suppressor functions of CD4 + CD25 high Foxp3 + T cells in TIL.
We show for the first time that Treg localizing to tumor tissue in HNSCC comprise a unique subset of CD4 + CD25 high T cells, which secrete IL-10 and TGF-h1 and mediate strong suppressor function. Intratumoral CD4 + CD25 high Foxp3 + T cells have phenotypic and functional characteristics that differ from those of circulating CD25 high Treg. Phenotypically, Treg found in TIL combine characteristic features of both nTreg and Tr1 Fig. 8. Potential mechanisms used to suppress proliferation of responder cells by CD4 + CD25 high Tcells inTIL of HNSCC patients. Autologous CD4 + CD25 -Tcells were labeled with CFSE and stimulated with OKT3 and anti-CD28 antibody in the presence of IL-2 as described in Materials and Methods. CD4 + CD25 high Tcells from TIL were added at the start of the culture (suppressor to responder ratio of 1:2). In some experiments,Transwell inserts were used to separate responder fromTreg. Neutralizing anti-IL-10 and/or anti-TGF-h1antibody or nonblocking IgG anti-mouse antibody were added to theTranswell and non-Transwell assays. Flow cytometry was done on day 5. A, left, suppression of responder proliferation separated from Treg byTranswell inserts (TR) compared with assays that are not separated by inserts at the suppressor to responder ratio of 1:2; right, suppression of responder proliferation separated fromTreg byTranswell inserts in the presence of nonblocking IgG anti-mouse antibody or neutralizing anti-IL-10 + anti-TGF-h1antibody at the suppressor to responder ratio of 1:2. B, suppression of responder proliferation in the absence of cell-to-cell interactions and the presence of nonblocking IgG anti-mouse antibody, neutralizing anti-IL-10 or anti-TGF-h1antibody, or neutralizing anti-IL-10 + anti-TGF-h1antibody at the suppressor to responder ratio of 1:2. Results of three representative experiments done withTIL of different HNSCC patients. cells; on the one hand, they express the phenotype consistent with activated memory T cells (CD25 high CD45RO + CD45RA -) and up-regulate Foxp3, CTLA-4, and CD95 as previously described by us for circulating nTreg in normal donors (22) and HNSCC patients (19). On the other hand, Treg in TIL upregulate IL-10 and TGF-h1 and express CD132 (IL-2Rg), which are features is characteristics of tumor-induced Tr1 cells (28). These novel findings introduce several questions about the origin of Treg at the tumor site and interactions between the host and human Treg subsets. Whether the regulatory cells are recruited from the periphery and accumulate at the tumor or whether the tumor microenvironment converts CD4 + CD25 -T cells present in situ to CD4 + CD25 high Treg is currently unclear. Treg in TIL might also comprise a subset of ''transition'' cells (i.e., nTreg activated by tumor antigens in the tumor microenvironment and converted into Tr1 T cells). Recent observations suggest that nTreg cells are involved in the differentiation of IL-10 -secreting Tr1 T cells in vitro (29) and in vivo (30). Most circulating Treg are positive for the lymphocyte homing markers CD62L and CCR7 as also previously reported (22). This finding indicates that they can migrate to lymph nodes as well as to the tumor. Notably, close to V90% of Treg in TIL are GITR + , although only few circulating Treg express GITR, and this suggests that Treg at the tumor site acquire GITR expression perhaps as consequence of local activation. Thus, considerable reshaping of infiltrating lymphocytes seems to be ongoing in the tumor microenvironment.
We also report for the first time that virtually all tumorinfiltrating Treg are positive for the chemokine receptor CCR4. We and others have previously shown that Treg in PBMC from patients with cancer and from normal controls up-regulate CCR4 (31). The functional significance of CCR4 expression on tumor-infiltrating Treg remains uncertain. Curiel et al. (10) recently showed that the chemokine CCL22 (CCR4L), which is abundantly expressed in ovarian cancer tissues and tumor cells in ovarian ascites, preferentially attracts the peripheral CD4 + CD25 + Foxp3 + T cells to the tumor. Several recent reports associated this expression of CCR4 with protection of T cells from apoptosis (32). Because TIL are especially sensitive to apoptosis in the tumor microenvironment (33), this possibility is intriguing and should be further investigated.
Our phenotypic analysis of the CD4 + CD25 + T-cell subset in TIL indicates for the first time in humans that close to half of the CD25 high Treg in TIL and PBMC down-regulate CD45RO expression but do not concomitantly gain CD45RA. This loss of CD45RO expression might be related to apoptosis sensitivity of CD25 high Treg. It is well documented that CD45RO high memory T effector cells (CD25 low/interm ) are highly sensitive to activation-induced cell death, whereas CD45RA + naive T cells are protected (34). Induction of apoptosis in human T cells by ligation of CD45 has been described (35). The loss of CD45RO on Treg may increase their resistance to apoptosis relative to tumor-specific CD45RO + T effector cells that are sensitive to apoptosis, contributing to an increased Treg compartment in patients with cancer.
Cytokines, and in particular IL-10 and TGF-h1, play major roles in the generation of nTreg (36 -38) and function of Tr1 cells (11,12), although their individual contributions to differentiation of Treg subsets and suppressor function are still debated. In the initial studies, using in vitro assay systems, it was shown that strong suppressive activity exerted by CD4 + CD25 high Foxp3 + Treg requires cell-to-cell contact (13,14), and it was thus considered likely that the suppression was dependent on ligand-receptor interactions at the cell surface. However, more recent reports suggest that some subsets of human CD4 + CD25 high cells produce IL-10 and/or TGF-h1 (39 -41). We show that CD4 + CD25 high T cells in TIL isolated from HNSCC patients or in situ produce IL-10 and TGF-h1 and that neutralizing antibodies to these cytokines completely restored T responder cell proliferation in an autologous system. This suggests that soluble factors play a central role in immune suppression mediated by Treg in TIL. Interestingly, however, our data also show that Transwell inserts that prevent cellto-cell contact between suppressor and responder also significantly block suppression (50%). The fact that the combination of neutralizing antibodies to IL-10 and TGF-h1 completely abrogated suppression in the presence or absence of the Transwell insert, whereas Transwell inserts in the absence of the blocking antibodies reduced suppression by 50%, suggests that these cytokines might be required for activation of surface molecules on Treg that are involved in suppression mediated by cell-to-cell contact. Thus, it is possible that, although cell-tocell contact contributes to suppression, it is dependent on Tregderived cytokines, which regulate/induce mechanisms directly responsible for blocking of responder cell functions. Our data suggest that mechanisms involving cytokines and also requiring cell-to-cell contact contribute to suppression mediated by Treg.
Our data also suggest that suppressor functions of Treg present in TIL may be related to poor survival in HNSCC patients with advanced disease. Frequently, newly diagnosed HNSCCs are locoregionally advanced (stages III and IV) at the time of diagnosis (42). Unfortunately, current oncologic therapy does not provide complete tumor control for advanced disease, and prognosis for advanced local HNSCC remains disappointing (43). It has been previously reported that advanced primary tumors of ovary are infiltrated by suppressor CD4 + CD25 + Foxp3 + T cells (10), whereas early tumors are characterized by a significantly lower Treg infiltration (44). These results indicate that Treg infiltrates into the tumor might reflect disease progression and that the nature of the tumor microenvironment, including accumulation of Treg, may influence the outcome of patients with cancer. Our data with TIL obtained from tumors of patients with advanced disease stages fit well with this explanation.
This study contributes to a better characterization of CD4 + T cells present in the tumor tissues and peripheral circulation of patients with HNSCC by analysis of the surface markers as well as cytokine profile and function, thus providing a broader basis for the development of more reliable prognostic factors. It also emphasizes the fact that in patients with cancer the tumor induces tolerance via activation and/or induction of Treg both locally and systemically. Additionally, the observation that Treg in TIL and PBMC have a different phenotype and use different mechanisms to suppress T-cell proliferation might have strong implications for the design and implementation of future immune therapies.