Tumors induce immunologic tolerance by several mechanisms, involving tolerogenic antigen presenting cells, Foxp3+CD4+ CD25+ regulatory T cells (Treg) and soluble immunoregulatory factors.1–4 An increased frequency of Treg was first observed in the peripheral blood of patients with bronchial carcinoma when compared with healthy individuals.5 Then similar findings were reported in patients with a variety of cancer types. This was followed by several reports indicating that Treg can be actively recruited and expanded in the tumor microenvironment.4, 6–9 Treg that are found within the tumor microenvironment are highly suppressive and abrogate effector function of cytotoxic T cells as well as NK cell-mediated cytotoxicity.1, 7, 10–13 These data have important clinical implications, as targeting Treg cells can enhance therapeutic antitumor effect in both humans and animal models.14–18 Interestingly, depletion of Treg cells led to crossreactive tumor immunity against tumors of diverse origins.19 Factors that have been described to be important for Treg induction within the tumor include TGF-β, IL-10, H-Ferritin, IDO and Prostaglandin E2 (PGE2).9, 20–22 These findings highlight the importance of understanding the mechanisms of immune-escape in the tumor microenvironment.
Stimulation of the T-cell receptor (TCR) or priming with dendritic cells can induce FoxP3 expression and the acquisition of Treg activity in CD4+CD25− T cells from normal PBMC.23–25 Yang et al. extended these findings and demonstrated that tumor infiltrating T cells in follicular lymphoma (FL) -involved lymph nodes could also be induced to express FoxP3 through TCR stimulation. Furthermore, they showed that CD70+ malignant B cells could facilitate this conversion of conventional T cells to Treg in FL.26 In this study, we investigated whether Treg induction in FL is a tumor-specific phenomenon attributed to either unique properties of tumor B cells or a specific program in tumor-infiltrating T cells that are destined to acquire a regulatory phenotype.
Material and methods
With informed written consent, specimens were obtained from patients with untreated (n = 5)/relapsed FL (n = 5) or reactive follicular hyperplasia of the tonsil (from sleep apnea patients, n = 12). All specimens were processed into a single-cell suspension and cryopreserved in fetal calf serum with 10% DMSO in liquid nitrogen.
Cell isolation and sorting
PBMC or FL samples were enriched for CD4+ cells using antiCD4 magnetic beads and the autoMACS system (Miltenyi Biotec, Auburn, CA). CD4+ cells were then stained with antiCD25 PE and antiCD4 APC, after which CD4+CD25− and CD4+CD25+ cells were isolated on a MoFlow cell sorter (Becton Dickinson, Mountain View, CA). B cells from PBMC or FL were obtained through enrichment using antiCD19 magnetic beads (Miltenyi Biotech) or the B cell Negative Isolation Kit according manufacture's instructions (Invitrogen, Carlsbad, CA). All isolated cell populations were at least 95% pure.
CFSE-labeling and T-cell proliferation assay
Sorted CD4+CD25− cells (Tresp) were labeled with CFSE as described previously.7 The suppressive activity of Treg was assayed by adding them to cocultures of Tresp and pooled allogeneic antigen presenting cells. Cocultured cells were harvested on Day 5 and analyzed for CFSE intensity on a modified dual laser LSRScan (BD Immunocytometry Systems, San Diego, CA) in the Shared FACS Facility at Stanford University using FlowJo software (TreeStar, Ashland, OR) for data analysis. The proportion of proliferating cells was measured by the percentage of CFSEdim cells.
Chemokine secretion assays
To quantify secretion of CCL17 and CCL22 by B cells, we first isolated B cells from lymphoma specimens or tonsils using the Dynal B cell negative Isolation Kit (Dynal Biotech Inc), which generally yielded purity greater than 95%. The B cells were resuspended at a concentration of 106/ml and were cultured in RPMI + 10% fetal calf serum with or without soluble CD40 ligand (a gift from Seattle Genetics) at 0.7 μg/ml for 4 days. Subsequently, the supernatant was removed and subjected to ELISA assays to quantify CCL17 and CCL22 using Quantikines kits (R&D systems, Minneapolis, MN).
CD4+CD25− and CD19+ cells from PBMC and/or FL were cocultured in flat-bottomed, 96-well plates and were separated by a 3 μm membrane or nonseparated as indicated for the individual experiment. Cultures contained IL-2 (100 IU/ml) and cells were plated at a density of 2 × 105 cells/well for each cell type added. Culture medium was X-vivo (Cambrex, Promega BioSciences, San Luis Obisbo, CA) media supplemented with L-glutamine (2 mM), penicillin (100 U/ml), streptomycin (0.1 mg/ml) and 10% BSA. After 96–120 hr, cultured cells were harvested and analyzed by FACS for CD4-FITC, FOXP3-APC and intracellular cytokine expression. Intracellular staining of Foxp3, IFN-γ and IL-2 was performed according to manufacturer's instructions (eBioscience, San Diego, CA).
The frequency of regulator T cells in FL, tonsil and PBMC was analyzed using ANOVA (GraphPad Software, San Diego, CA). A p value < 0.05 was considered statistically significant.
The frequency of Treg in follicular lymphoma-involved lymph nodes
We examined the frequency of Treg cells and found that the percentage of CD4+ cells expressing FoxP3 in FL-involved lymph nodes was significantly higher than that in benign reactive tonsils and in PBMC with a median frequency of 39.5% (range 18–49%) in FL, 9.6% (range 3–13%) in tonsils and 8.9% in PBMC (range 5–12%) (P < 0.0001) (Fig. 1a). Furthermore, Treg from FL were able to suppress the proliferation of autologous CD4+CD25− cells in a dose-dependent fashion, demonstrating their regulatory function (Fig. 1b). These findings are consistent with previous reports7, 27
Analysis of Treg-trafficking-related chemokines and chemokine receptors in follicular lymphoma and benign tonsils
Given that the frequency of Treg was increased in FL when compared with that in reactive tonsils, we hypothesized that this might be due to preferential Treg trafficking to the tumor or due to conversion of conventional T cells to Treg in the tumor microenvironment. It has been reported that chemokines CCL17 and CCL22 and their receptors, CCR4 and CCR5, are involved in Treg trafficking.7, 28, 29 Thus, we examined expression of CCR4 and CCR5 on CD4+FoxP3+ Treg cells isolated from FL or tonsils. In addition, we quantified CCL17 and CCL22 production by B cells isolated from either tonsils or FL. As shown in Figure 2a, we observed no difference in expression of CCR4 and CCR5 on Treg between tonsils and FL. Although B cells from tonsils appeared to produce more CCL22 in the absence of CD40 ligand stimulation (Fig. 2b), this would be unlikely to account for less-concentrated Treg in tonsils. With CD40 ligand stimulation, we saw no difference in the amount of CCL17 and CCL22 produced by either tonsils B cells or FL B cells (Fig. 2b). Therefore, it seemed unlikely that preferential Treg trafficking was the mechanism that could account for an increased frequency of Treg in FL.
Conversion of conventional T cells to regulatory T cells by tumor B cells but not normal B cells
To investigate whether Treg could be converted from conventional T cells in FL, we cocultured sorted CD4+CD25− cells from FL specimens with purified autologous tumor B cells. We found that the FoxP3 expression was induced in the coculture of FL B cells and their autologous conventional T cells. In contrast, the FoxP3 expression was not induced in the coculture of B cells and autologous conventional CD4+CD25− T cells from PBMC (Fig. 3a). Furthermore, FoxP3 induction in CD4+CD25− T cells was not achieved when CD4+CD25− cells were separated from the tumor B cells in transwell assays, indicating that cell–cell contact was required in this process (Fig. 3b). In all experiments shown in Figure 3, cells were cocultured for 96–120 hr, after which they were stained with CD4-FITC and FoxP3-APC followed by FACS analysis. Dead cells were excluded from analysis by their small forward scatter. Cells were gated on CD4 staining and subsequently on FoxP3 staining. In Figures 3a and 3c, where FoxP3 induction occurred, we observed a large percentage of CD4 cells shifted upwards and rightwards, corresponding to the induction of FoxP3 expression in CD4+ cells. For the purpose of quantification, a CD4+FoxP3+ gate was drawn based on the isotype controls of FITC and APC. The cells within this gate represented the percentage of CD4+ cells expressing FoxP3+ at the end of 96–120 hr coculture. Data collected from repeated experiments were summarized in Supporting Information Table I.
These results led us to ask whether the conversion of Treg was a property of the tumor B cells or of the tumor infiltrating T cells. To address this question, we cocultured purified B cells from normal PBMC with sorted CD4+CD25− cells from a FL specimen and vice versa. As shown in Figure 3c, the Treg conversion was dependent on tumor B cells but independent of the source of T cells. Thus, we demonstrated that FL-derived B cells possessed unique properties that could promote Treg conversion.
It has been observed that FoxP3 induction in human CD4 T cells does not necessarily correlate with acquisition of regulatory function. This discordance has been reported for human CD4+ T cells activated by TCR stimulation with antiCD3 and costimulation with CD28.30, 31 In accordance with their lack of regulatory function, these FoxP3+ CD4+ T cells do produce Th1 cytokines such as interferon-γ and IL-2 upon activation with PMA/ionomycin, in contrast to Treg, which do not.30, 31 Thus, the cytokine profile of induced FoxP3 expressing T cells can serve as a surrogate parameter for their regulatory activity.
Next, we tested cytokine profile of converted Treg as a surrogate marker for Treg activity. The cocultures of CD4+CD25− T cells and B cells were subsequently stimulated with PMA and ionomycin, after which the expression of FoxP3, IFN-γ and IL-2 was examined in CD4+FoxP3+ cells. As shown in Figure 4, the percent of FoxP3+ cells was increased in the FL coculture but not in that of PBMC. However, the production of IL-2 and IFN-γ by the FoxP3+ cells in the FL mixture remained at the background level, suggesting that the converted FoxP3+ cells in the FL mixture were not activated effector T cells. Of note, the production of IL-2 and IFN-γ by the FoxP3-negative T cells was used as an internal control of cytokine production.
There has been increasing evidence that Treg play an important role in fostering an immune privileged tumor microenvironment thereby promoting tumor progression.32 In support of this, studies have shown that the numeration and distribution of Treg in the tumor microenvironment are correlated with clinical outcome.4 Treg can be recruited, expanded or converted in the tumor microenvironment.4, 6–9 However, it is still not clear whether Treg conversion in the tumor microenvironment is a property of the tumor cells or that of the tumor-infiltrating T cells.
In this study, we have shown that FL B cells can convert conventional T cells to Treg. Our results confirmed the observation of other investigators.27 In addition, we demonstrated that Treg conversion in FL is clearly dictated by tumor cells. We observed, for the first time, that in contrast to malignant B cells, normal B cells did not have the ability to convert Treg in our experimental setting. Furthermore, tumor-infiltrating T cells were as susceptible to Treg conversion as normal T cells from PBMC (Figs. 3a and 3c). FL has distinct clinical and pathological properties, such that immunological factors of both the host and the tumor microenvironment have been linked to clinical outcome.33–37 Our results demonstrated how tumor B cells can impact the immunological factors in the tumor microenvironment.
Previous studies have established that artificial TCR stimulation by antiCD3 along with costimulation by CD28 can convert conventional T cells to Treg. Under these experimental conditions, the presence of FL-derived malignant B cells is not necessary but helpful to Treg conversion.23, 26 In contrast to these studies, Treg conversion in our experiments was achieved without artificial TCR stimulation. Instead, the stimulation of conventional T cells was solely provided by normal or malignant B cells. This observation is important for 2 reasons: First, this observation suggested that only malignant B cells are uniquely equipped with the ability to modify the tumor microenvironment, which may play a role in tumor progression. Second, it is possible that TCR stimulation by natural antigens, such as tumor-associated antigens, presented by malignant B cells may be the key factor leading to Treg conversion. The mechanism of the B and T-cell interaction in Treg conversion is the focus of our future studies.
Treg conversion observed in our experiment described a phenomenon, where the percentage of CD4+FoxP3+ cells increased during the coculture of malignant B cells with tumor-infiltrating conventional T cells (CD4+CD25−). The tumor-infiltrating conventional T cells were sorted to at least 90% purity before cocultured with B cells. We and others have observed that 5–10% tumor-infiltrating conventional T cells from FL do express FoxP3 (unpublished data and Ref. 26). During the 96–120 hr coculture, we observed 2-fold increase of the total cell number (data not shown). The question arises whether Treg conversion we described was due to preferential proliferation of the CD4+CD25−FoxP3+ cells over the conventional T cells (the expansion scenario), or FoxP3 expression was induced in true conventional,CD4+CD25−FoxP3−, T cells (the conversion scenario) or the combination of both. Our data showed that both the total cell number and the frequency of FoxP3 expressing CD4 T cells increased during the coculture period, which is in concordance with the observation from other investigators.27 As we could not label the 5–10% of the CD4+CD25−FoxP3+ cells separately from the true conventional T cells (CD4+CD25−FoxP3−) and observe their behavior independently in the coculture, we were not able to distinguish the conversion from the expansion scenario. However, we believe that the pure expansion scenario is unlikely, given that the total cell number doubled during the coculture period, but the percentage of CD4+FoxP3+ cells increased from 5–10% to 50–90% (Supp. Info. Table I). In support of this reasoning, it should be mentioned that true human Treg (CD4+CD25+FoxP3+) are difficult to expand in vitro, because even a minute amount of conventional T cells mixed in the sorted Treg can outgrow the Treg in expansion cultures. As such, Treg expansion in vitro usually requires robust stimulation of TCR and special conditions such as addition of Rapamycin to preferentially suppress proliferation of conventional T cells (Ref. 38 and unpublished results by Hou and coworkers). Taken together, we believe that the increased frequency of CD4+FoxP3+ cells during the coculture of conventional T cells with malignant B cells likely represents Treg conversion and perhaps expansion of these Treg as well.
It has been shown that FoxP3 can be transiently expressed in T cells activated by artificial TCR along with costimulation of CD28.30, 31 Although it is unknown whether the disconnection between FoxP3 expression and acquisition of regulatory function is also present in Treg induced by antigen presenting cells, we felt that it was important to determine whether converted Treg indeed had regulatory activity. Gavin et al. have shown that Th-1 production can distinguish FoxP3-expressing effector T cells from FoxP3+ Treg cells. Thus, single-cell cytokine profiles can serve as an ideal surrogate for Treg activity.31 This FACS-based single-cell analysis is easy to do on a small number of converted Treg available from our conversion experiment. On the basis of this, we analyzed the production of IFN-γ and IL-2 in converted Treg by flow cytometry and showed they indeed did not produce Th-1 cytokines, suggesting that they have acquired Treg activity in addition to expression of FoxP3. Using a T-cell suppression assay, Mittal et al. have shown that Treg converted by FL B cells have suppressive function.27 Thus, our assays were complementary to these investigators' in demonstrating the regulatory function of converted Treg.
Preferential trafficking is another potential mechanism by which Treg are accumulated in the tumor microenvironment. Chemokines and chemokine receptors that are involved in Treg trafficking are CCL17, CCL22 and CCR4, CCR5, respectively. CCL17 and CCL22 are secreted by macrophages, and CCR4 and CCR5 are expressed on CD4 T cells, including Treg. Normal and malignant B cells can secrete CCL17 and CCL22 when stimulated by soluble CD40 ligand or viral antigens39–41; however, it is not clear whether malignant B cells can secrete more such cytokines that can account for the high frequency of Treg in tumors. For this reason, we compared the production of CCL17 and CCL22 by purified B cells and expression of CCR4 and CCR5 on Treg from FL specimens and benign tonsils (Fig. 2). To our surprise, with soluble CD40 ligand stimulation, there was no difference in CCL17 and CCL22 production between B cells from tonsils and B cells FL. Furthermore, without soluble CD40 stimulation, tonsils B cells appeared to produce more CCL22, which would be unlikely to account for less frequency of Treg in tonsils. Our results indicated that preferential trafficking of Treg would be unlikely the mechanism for increased frequency in FL when compared with that in tonsils.
It is important to recognize that malignant B cells do produce CCL17 and CCL22 (Fig. 2b and Ref. 7), as the production of these cytokines may play an important role in architectural localization of Treg within the tumor microenvironment. The architectural localization of Treg appeared to be clinically important, as a recent study demonstrated that the number and the location of Treg relative to follicles in a lymph node were correlated with the risk of histological transformation in FL.42 It should be mentioned that we did not measure the secretion of CCL17 and CCL22 by tumor infiltrating macrophages, because they generally do not survive the freeze-and -thaw process (data not shown).
In summary, our results provided evidence that malignant B cells possess intrinsic properties to modify the tumor microenvironment through the induction of Treg. Studies are underway to identify molecules that are essential in Treg conversion. To this end, we used the antiTGF-β blocking antibody in the conversion experiment, but it failed to inhibit Treg conversion. Similar results were reported by other investigators.27 However, some studies suggested that CD70 and PD-1 on B cells may play a role in Treg induction.26, 27 Identification of the critical molecules involved in Treg induction by lymphoma tumor cells may lead to specific interventions to prevent the negative immune regulatory effects of the lymphoma cells.
This study was supported by grants from the National Institute of Health (to R.L. and R.N.) and the Doris Duke Charitable Foundation (to R.N.). W.A. is a Lymphoma Research Foundation Fellow and a recipient of the American Society of Clinical Oncology Young Investigator Award. R.Z. is supported by the Deutsche Krebshilfe, Germany. R.L. is an American Cancer Society Clinical Research Professor.