Follicular B Lymphomas Generate Regulatory T Cells via the ICOS/ICOSL Pathway and Are Susceptible to Treatment by Anti-ICOS/ICOSL Therapy Free

Follicular lymphoma (FL) is usually considered as an indolent disease, but some patients present resistance to treatment and/or transformation into aggressive lymphoma (1). The clinical outcome of FL patients was shown to be influenced by tumor-infiltrating macrophages and T cells (2–5), including CD4⁺ regulatory T cells (Treg).

Tregs comprise natural and inducible Tregs, that can modulate immune responses by suppressive mechanisms (6). Accumulation of Tregs was found in various hematologic and nonhematologic malignancies (6). In contrast with solid tumors, in which Tregs suppress antitumor immunity and promote tumor progression (7–10), high amounts of intratumoral FoxP3⁺ Tregs were associated with a favorable outcome in FL (11–13). This result appears paradoxical, because Tregs from FL tissues were shown to inhibit the function of antitumor CD4⁺ and CD8⁺ T cells (14–16). Of note, a similar favorable influence of Tregs has been also reported in classical Hodgkin lymphoma a lymphoma subtype in which the microenvironment is known to play a key role (17). Nonetheless, the favorable prognostic value of Tregs in FL remains controversial because it was not clearly confirmed in recent studies (18–20).

Tregs lack CD127 (IL7 receptor α chain) and usually express CD25 (IL2 receptor α chain), GITR (glucocorticoid-induced TNFR-related protein), CTLA-4 (cytotoxic T lymphocyte associated antigen 4), and CD45RO (21). Expression of the transcription factor FoxP3 (forkhead box P3) is considered as a crucial feature for the development and function of Tregs (21). Lymphoma B cells produce the CCL22 chemokine inducing Treg recruitment (15) and converting conventional T cells (Tconv) to Tregs (22–24). Cell contacts were shown to be required for this conversion, which involves CD70, CD80, and CD86 expression on B cells (22, 24).

ICOS (inducible T cell costimulator, CD278)/ICOSL pathway–driven generation of Tregs has been recently demonstrated in the periphery (25) and microenvironment of solid tumors (26–28). ICOS, a costimulatory molecule of the CD28 family, is expressed on activated T cells. ICOS is involved in T-cell responses upon engagement with its ligand, ICOSL (ICOSLG, B7-H2, CD275), which is normally expressed on B cells, dendritic cells, and monocytes (29). ICOS expression confers an activated phenotype and a strong suppressive capacity to intratumoral Tregs (30). Particular Tregs displaying strong expression of ICOS and CXCR5 and inhibiting CD4⁺ effector T cells were recently identified in FL tissues, and called follicular regulatory T cells (T_FR; ref. 31). This subset is reminiscent of T_FR observed in normal follicles in mice (32, 33). Nonetheless, the hypothesis of a direct suppressive function of Tregs on FL B cells has not been fully investigated. These observations prompted us to clarify the role of Tregs in FL, using a combination of flow cytometry, cell sorting, and IHC on tissue samples. As a control group, we analyzed tissue samples from not only benign lymphadenitis, but also lymphoma subtypes in which the microenvironment is suspected to play a significant role, especially cHL, and to a lesser extent diffuse large B-cell lymphoma (DLBCL).

Fresh biopsy samples from 46 patients were collected at diagnosis. Samples were classified as FL (n = 20), DLBCL (n = 10), cHL (n = 9), and reactive lymphadenitis (rL; n = 7) according to the World Health Organization classification (IARC, Lyon, 2008). DLBCL cases were considered as de novo cases, due to the lack of previous history of FL and absence of FL histologic features. All patients gave informed consent and the study was approved by the ethical board of the Paoli-Calmettes and Albert Bonniot institute.

Burkitt lymphoma cell line Raji and COS-7 cells were obtained from American Type Culture Collection (ATCC). Raji and COS-7 cells were respectively cultured in RPMI 1640 Medium (Life Technologies) supplemented with 10% fetal calf serum (FCS; Lonza) and in DMEM (Life Technologies) supplemented with 10% FCS. These cell lines were not authenticated at the time we performed the study.

Sample tissues were mechanically disrupted and passed through a nylon filter (BD Bioscience) to obtain mononuclear cells (MC), which were frozen or used immediately for functional assay.

B cells and CD4⁺ T cells were obtained from MCs by negative selection (Stemcell Technologies). To isolate Tregs and Tconv, purified intratumoral CD4⁺ T cells were labeled with anti-CD4-PC7, anti-CD25-APC, anti-CD127-FITC, and a viability marker (live/dead aqua) and sorted by FACs (BD Biosciences). To separate ICOS⁺ and ICOS^neg CD4⁺ T cells, intratumoral CD4⁺ T cells were labeled with anti-ICOS-PE and live/dead aqua.

Antibodies are detailed in Supplementary Table S1. To analyze intracellular markers like FoxP3, Helios, Bcl-2, and Ki67, cells were fixed/permeabilized after surface staining and then incubated with specific antibodies using Foxp3 staining buffer set (eBioscience). Data were acquired on LSRII (BD Biosciences) and analyzed using FACS Diva and FlowJo softwares.

Naïve wild-type FVB/N (Charles River Laboratory) were used at 6 weeks of age. Mice were maintained in the pathogen-free animal facility "AniCan" at the Cancer Research Center of Lyon. Experiments were conducted in accordance with the European and French laws and were validated by the local animal ethical evaluation committee (CECCAP ethical committee agreement no. CECCAPP-2013-010 and registered at the MESR, #02312.01).

Mice were injected with either anti-mICOS mAb (17G9, rat IgG2) or isotype control (BioXcell; 100 μg/mL) by i.p. route and sacrificed after 24 hours or 48 hours. Blood was collected by intracardiac function, and lymph node and spleen were subjected to mechanical disaggregation to obtain homogenous single-cell suspension. Membrane multiparametric flow cytometry stainings (CD45-AF-700, CD3-APC-H7, CD19-BV605, ICOS-L-PE, and live/dead Aqua) were performed to evaluate ICOS expression on CD19⁺ B cells after gating on viable cells (live/dead Aqua^neg). A multistaining including an IgG2a-PE in place of anti-ICOSL Mab (FMO) was used as a negative control. Flow cytometry analyses were performed on an LSR-II (Becton Dickinson), and results were analyzed using FlowJo software.

To study proliferative capacity, CellTrace violet (Life Technologies)–labeled Tregs and Tconv were activated by CD3/CD28 dynabeads (Life Technologies; 1 bead for 1 cell) in the presence of recombinant human IL2 (rIL2, 100 IU/mL; Chiron) for 5 days. To evaluate the suppressive effects of Tregs on CD4⁺ Tconv, CellTrace violet–labeled Tconv were activated by CD3/CD28 dynabeads in the presence of rIL2 (100 IU/mL) and Tregs for 4 days. Cells were harvested and stained with anti-CD2-PECF-594 and live/dead near-IR. Proliferation was evaluated by Division index (DI) using the proliferation tool of FlowJo software. Supernatants of coculture were collected and IFNγ was further measured by ELISA (BD Biosciences).

FL B cells purified from tissues were stained with CellTrace violet and were cultured with autologous or allogeneic Tregs (1/1 ratio) in the presence of rIL2 (100 IU/mL) for 5 days. Anti-IgA/G/M (10 μg/mL; Jackson ImmunoReseach) and CpG-B (2 μg/mL; Invivogen) were added to stimulate FL B cells as previously described (34). To study T-cell–dependent suppression of Tregs, an equal number of Tconv and/or Tregs were added to stimulated B cells. Supernatant of coculture was collected to measure IL6 by ELISA (R&D Systems).

Purified FL B cells or total MCs were cultured in culture medium. Cells were harvested and stained with live/dead aqua, anti-CD19-FITC, anti-ICOSL-PE, and anti-CD86-APC before and during culture. FL B cells were cultured overnight until ICOSL was expressed. Then, different ratios of intratumoral ICOS⁺ or ICOS^neg CD4⁺ T cells were added on the B cell culture. In some experiments, anti-ICOS–blocking antibody (314.8 clone, produced in our laboratory) or isotype control (Ctrl) IgG1 (Sigma) were added at 10 μg/mL.

CellTrace violet-labeled intratumoral CD4⁺ T cells were cultured alone or with autologous or allogeneic FL B cells expressing ICOSL at 1.5/1 ratio in the presence of rIL2 (100 IU/mL). Anti-ICOS (314.8 clone) or anti-ICOSL (MIH.12 clone, eBioscience) blocking antibodies or isotype Ctrl IgG1 were added at 10 μg/mL for each. After 5 to 6 days, cells were harvested and surface stained with anti-CD4-PC7, anti-CD19-PECF-594, anti-CD25-APC, anti-ICOS-PE, and live/dead near-IR before fixation/permeabilization and intracellular staining with FoxP3-FITC.

Cells harvested after 5 to 6 days of coculture were restimulated with PMA (10 ng/mL; Sigma) and Ionomycin (1 μg/mL, Sigma) for 4 hours. After stimulation, cells were stained with anti-CD4-V450, anti-CD19-PECF-594, anti-CD25-PC7, anti-FoxP3-PE, anti-IFNγ-FITC, anti-IL2-APC and live/dead near-IR.

IHC was performed on whole sections of paraffin blocks used for diagnosis. After dewaxing and pressure-cooker antigen retrieval, mAbs directed against ICOS (clone sp98; Abcam), and FoxP3 (clone 236A/E7; eBioscience) were incubated in an automated immunostainer (Dako) using a standard avidin biotin peroxidase technique. Double staining experiments were performed using the K5361 Envision G/2 Doublestain System (Dako) according to the supplier's instructions. ICOSL IHC was performed on frozen tissue sections using the Envision-Flex K8000 kit (Dako) and the Autostainer Plus device (Dako). The anti-ICOSL mAb HIL-270 (provided by Pr. Richard Kroczek) was diluted at 10 μg/mL and incubated for 45 minutes at room temperature. As a control for the specificity of the HIL-270 ICOSL mAb, paraffin-embedded cell blocks were constructed using ICOSL-transfected COS-7 cells and parental COS-7 cells as negative control. Briefly, COS-7 cells transfected for 48 hours with βDNA4 vector containing full-length human ICOSL (produced in our laboratory) using X-tremGENE 9 DNA Transfection reagent (Roche) according to the manufacturer's instructions. After 48 hours, transfected and parental COS-7 cells were detached and prefixed in 4% formaldehyde solution and then suspended in agarose 1% LMT to form a cell-agarose gel, which was fixed in 4% formaldehyde solution for 24 hours. After fixation, the cell-agarose gel was fixed, dehydrated, paraffin impregnated on a Leica ASP300S processor (Leica Biosystems), and finally embedded in paraffin using a Leica Histo-Embedded processor. The method for ICOSL IHC on paraffin embedded cell blocks was similar to that described for ICOS and FoxP3. As expected, positive HIL-270 immunostaining was present only in ICOSL-transfected cells. For all antibodies, including HIL-270, a control was also done by omitting the primary mAb on serial sections of frozen tissues analyzed, which resulted in a lack of positive signal.

Total RNA was extracted from purified FL B cells and Burkitt lymphoma cell line Raji with TRIzol reagent (Life Technologies) according to the manufacturer's protocol and was then reverse transcribed into cDNA using M-MLV Reverse Transcriptase (Life Technologies). Quantitative PCR was performed using Taqman Universal PCR Master Mix and primers as follows: glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Unigene reference: Hs.544577) and human ICOSL (Hs. 14155; Life Technologies). Gene expression was measured using the 7900HT Fast Real-Time PCR System (Life Technologies). For each sample, mRNA expression was normalized to GAPDH as a reference gene.

Quantitative variables were expressed as mean ± SEM. Statistical analysis was performed with GraphPad Prism 5 software (GraphPad software) using the Mann–Whitney and Wilcoxon nonparametric t test (*, P < 0.05; **, P < 0.01).

Coexpression of CD25^high and CD127^low/neg was used to identify and isolate Tregs in FL tissue samples (Fig. 1A). CD25^highCD127^low/neg cells expressed high levels most Treg markers such as FoxP3 (86% ± 9.1%, mean ± SD), Helios (84% ± 7.6%; Fig. 1B) and strongly expressed other Tregs markers such as GITR (72 ± 13.4), surface CTLA-4 (32.8% ± 11%), and CD39 (80.2% ± 16.1%; Fig. 1B). Most CD25^highCD127^low/neg Tregs expressed CD27 (99.2% ± 1.1%) and were CD45RA-negative (3.2% ± 1.6%; Fig. 1B), suggesting a memory T-cell phenotype. Tconv were identified as CD25^neg/low among viable CD4⁺ T cells (Fig. 1A). When compared to Tconv, purified CD25^highCD127^low/neg Tregs had a low proliferation rate in response to CD3/CD28 stimulation (87% and 10% for Tconv and Tregs, respectively; Fig. 1C). Tconv proliferation and IFNγ production was strongly inhibited when CD25^highCD127^low/neg cells were added in culture (Fig. 1D). Taken together, these results demonstrate that CD25^highCD127^low/neg T cells have the phenotypic and functional profile of effector Tregs.

The mean percentage of Tregs within CD4⁺ T cells was significantly increased in FL tissues as compared to rL (7.5-fold, P = 0.0005), cHL (2.6-fold, P = 0.003) and DLBCL (2.3-fold, P = 0.01; Fig. 2A). Both Tregs/Tconv and Tregs/TCD8⁺ ratios were significantly higher in FL than in other samples (Fig. 2B). ICOS was mostly expressed in CD4⁺ T cells, especially in Tregs (65.3% ± 14.4%), as compared with other tumor infiltrating lymphocytes (TIL) like CD8⁺ T cells, γδ T cells, and natural killer cells (Fig. 2C). Accordingly, the percentage of ICOS⁺ Tregs among CD4⁺ T cells was significantly increased in FL tumors (Fig. 2D). Like the majority of Tregs in FL, ICOS⁺ Tregs expressed CXCR5 (mean ± SD: 83% ± 7.5%; Fig. 2E). When compared to ICOS^neg Tregs, ICOS⁺ Tregs in FL tissues displayed significant upregulation of surface CTLA-4 (40.1% ± 10.1%), GITR (83.6% ± 10.2%), CD39 (85% ± 15.9%), PD1 (90.4% ± 7.9%) as well as higher Ki67 expression (19.1% ± 7.2%; Fig. 2E). These data indicate a specific accumulation of activated and expanding ICOS⁺ Tregs in the FL microenvironment.

To identify a putative cell population responsible for ICOS⁺ Tregs accumulation in fresh FL tissues, we assessed the expression of ICOSL and CD86 on FL B cells, reactive T cells, myeloid dendritic cells (mDC) and plasmacytoid DC (pDC). FL B cells were identified as CD3^neg/CD16^neg/CD56^neg/CD20⁺, because most of these cells in FL samples were shown to be neoplastic B cells due to coexpression of CD10 and Bcl-2 (Supplementary Fig. S1A). pDCs were identified as lineage^neg/HLA-DR⁺/CD33^neg/+/CD123⁺ and mDCs as lineage^neg/HLA-DR⁺/CD33⁺⁺/CD123^neg (Fig. 3A). CD86 was strongly expressed on mDCs (90.6% ± 10.1%), pDCs (21.8% ± 15.9%), and FL B cells (52.6% ± 27.3%), but ICOSL was weak on all cell populations (Fig. 3B) including reactive T cells (Supplementary Fig. S1A). Surprisingly, only ICOSL expression was rapidly upregulated on FL B cells after a few hours of culture in vitro of purified FL B cells (Fig. 3C). ICOSL expression was weak on FL B cells after purification (1.7% ± 1.6%), but it was induced by short in vitro culture (15.1% ± 18.0% after 2 hours of culture) and increased during culture (45.1% ± 34.6% at 48 hours). Furthermore, ICOSL upregulation on FL B cells was also observed in vitro culture of total MCs (Supplementary Fig. S1B). We then examined the expression of ICOSL transcripts on purified FL B cells by quantitative RT-PCR. The Burkitt lymphoma cell line Raji was used as a positive control due to its high ICOSL expression detected by flow cytometry (Supplementary Fig. S1C). ICOSL transcripts were detectable on purified FL B cells and Raji cell line but its mRNA level was 2-fold higher on FL B cells than on positive control Raji (Fig. 3D). In contrast to FL B cells, ICOSL was expressed on lymphoma B cells from some DLBCL samples, whereas ICOS was absent on TILs (Supplementary Fig. S1D).

ICOSL IHC was performed on 16 frozen samples, including human tonsils (n = 2), reactive lymph node (n = 1), low-grade follicular lymphoma (n = 10), and DLBCL (n = 3). ICOSL immunostaining was absent or faint in neoplastic follicles in the 10 FL cases analyzed (Fig. 4). Internal positive controls were mainly located outside the B-cell follicles and often displayed morphologic features of macrophages with large cytoplasm and oval nucleus (Fig. 4). Capillary venules also displayed positive signals of variable intensity (Fig. 4) in accordance with a previous report (35). A few ICOSL⁺ cells are also present in the interfollicular areas of benign tonsils and lymphadenitis, whereas normal B cells in reactive follicles appeared mainly negative. In contrast, DLBCL cells displayed heterogeneous ICOSL positivity in two of three analyzed cases (data not shown).

FoxP3 and ICOS IHC were performed on 6 FL and DLBCL cases that were also analyzed by flow cytometry. IHC results confirmed flow cytometry data, because numerous ICOS⁺ were present in FL cases, with a predominant intrafollicular localization (Fig. 4). Variable amounts of FoxP3⁺ cells were also observed within neoplastic follicles of FL cases (Fig. 4). In addition, ICOS⁺ /FoxP3⁺ TILs could be detected in FL cases using double staining experiments and were located within the follicles (Fig. 4), which implies vicinity and possible interaction with neoplastic B cells.

ICOSL downregulation through ICOS binding was previously described in normal murine B cells (36) and in human pDC (27). We evaluated the impact of FL T cells on ICOSL expression by cocultures of conditioned FL B cells with either autologous intratumoral ICOS⁺ or ICOS^neg CD4⁺ T cells at different ratios during 24 hours. An important decrease of ICOSL expression on FL B cells (about 67% at 1:1 ratio after 24 hours of coculture) was observed in the presence of ICOS⁺ CD4⁺ T cells (Fig. 5A). A high level of ICOSL expression could be restored on FL B cells using an antagonist anti-ICOS antibody (314.8; Fig. 5B). Similar results were also observed in both in vitro culture of total MCs and in vivo using normal murine model (Supplementary Fig. S2). In contrast, no alteration of CD86 expression on FL B cells was observed (Fig. 5B). These data suggest that ICOS/ICOSL interaction is involved in downregulating ICOSL expression on FL B cells.

We next explored the possible causes of Treg accumulation in FL tissues. We cocultured purified FL B cells conditioned to express ICOSL by in vitro culture, with intratumoral autologous or allogeneic CD4⁺ T cells in the presence of rIL2 for 5 to 6 days. Intratumoral CD4⁺ T cells were labeled with CellTrace violet prior coculture to evaluate the division rate. We first observed an enhancement of CD25 and FoxP3 expression on CD4⁺ T cells (data not shown). Consequently, the number of Tregs, defined by CD25 and FoxP3 coexpression, was significantly increased in the presence of FL B cells (median, 33.5% versus 20.0%; n = 6; P = 0.03; Fig. 6A, left). These Tregs comprised two distinct subsets of CD25⁺FoxP3⁺ and CD25^highFoxP3^high Tregs (Fig. 6A, right). Median percentages of CD25⁺FoxP3⁺ Tregs were 29.2% following coculture with FL B cells, versus 19.9% (P = 0.03) without FL B cells. The number of CD25^highFoxP3^high Tregs was virtually undetectable without FL B cells, but clearly increased during coculture. Their median percentage was 3.0% and 9.4% per total CD4⁺ T cells and per total Tregs, respectively (P = 0.03). Proliferation of Tregs in culture with FL B cells was significant, reaching 47.9% and 20.5% of CD25^highFoxP3^high Tregs and CD25⁺FoxP3⁺ Tregs, respectively (Fig. 6B, top). ICOS expression was markedly higher on CD25^highFoxP3^high Tregs than on CD25⁺FoxP3⁺ Tregs [mean fluorescence intensity (MFI): 1451 vs. 573, respectively; Fig. 6B, bottom]. After restimulation of cell suspension using PMA/ionomycin, intracellular cytokines were detected (Fig. 6C). FoxP3^neg CD4⁺ T cells strongly secreted IL2 (median 23%) and IFNγ (median: 13.4%). In contrast, both putative Treg subsets showed weak IL2 and IFNγ secretion. Median percentages among CD25⁺FoxP3⁺ Tregs were 5.3% and 1.9% for IL2 and IFNγ, respectively. Median percentages among CD25^highFoxP3^high Tregs were 4.6% and 1.1% for IL2 and IFNγ, respectively. These data indicate that both cell populations were indeed Tregs rather than activated T cells. Thus, FL B cells expressing ICOSL could favor intratumoral enrichment of ICOS⁺ Tregs.

When neutralizing anti-ICOS and anti-ICOSL mAbs were added in coculture, there were not significant change in the percentage of CD25⁺FoxP3⁺ Tregs among CD4⁺ T cells (Fig. 6D, left), whereas both mAbs induced a significant reduction in the percentage of CD25^highFoxP3^high Tregs (Fig. 6D, right). No impact was observed after the addition of an isotype control IgG1. These results indicate that generation of CD25^highFoxP3^high Tregs expressing ICOS is strongly dependent on ICOSL expression by FL B cells.

Our observation of Tregs overrepresentation in FL tumors prompted us to test their capacity to inhibit FL B cells. We selected IL6 as one of the main B cell–produced cytokine (37). We first tested whether Tregs could act on B cells activated in a T-cell–independent way. FL B cells were activated by CpG-B and anti-IgA/G/M prior to culture with Tregs at 1:1 ratio. We observed a significant attenuation of CD80 and CD86 expression on FL B cells. This was evidenced by the percentage for CD80 (approximately 30% of inhibition, P = 0.0078, n = 9) and by MFI for CD80 and CD86 (about 35% and 31% inhibition for CD80 and CD86, respectively (Fig. 7A). Similarly, Tregs abrogated IL6 secretion in the supernatant of FL B cells (median percentage of 26% inhibition, P = 0.0078, n = 8; Fig. 7B).

Tconv in FL samples could also be suppressed by Tregs (Fig. 1D). About a quarter (25.2% ± 16.1%, n = 12) of these Tconv were T_FH cells characterized by ICOS and CXCR5 coexpression (data not shown). We next searched for a putative suppression of Tconv-driven B-cell responses by Tregs. For this purpose, stimulated B cells were cocultured with Tconv, with or without Tregs at 1:1 ratio. Tconv induced a significant increase of both CD80 and CD86 expression on stimulated B cells, and of IL6 secretion (median increase of 67%, 121%, and 49% for CD80 MFI, CD86 MFI, and IL6, respectively). B-cell responses were inhibited after addition of Tregs (inhibition of 44%, 53%, and 39% for CD80 MFI, CD86 MFI, and IL6, respectively; Fig. 7C and D).

Of note, there was also a tendency for Tregs to suppress B-cell proliferation, although it was not statistically significant for all samples (Supplementary Fig. S3). These data suggest that intratumoral Tregs can directly or indirectly suppress FL B-cell responses upon activation.

Clarifying the impact of Tregs on FL pathogenesis is a critical issue for clinical implications. However, the lack of strictly a specific Treg marker may result in discrepancies between studies focusing on Tregs in human. FoxP3, which is considered as the most specific Treg marker, was reported to be transiently expressed in activated CD4⁺ and CD8⁺ T cells (38, 39), which, in contrast to Tregs, produce effector cytokines like IL2 and IFNγ (23, 39, 40). Using the CD25^high/CD127^low/neg combination, we could isolate T cells displaying Treg phenotypic markers and strong suppressive activity. Most CD25^highCD127^low/neg T cells were FoxP3^high and hardly secreted IL2 and IFNγ, further supporting their Tregs' nature. Thus, our results corroborate previous investigations (41, 42) suggesting that the CD25^highCD127^low/neg combination is useful for Tregs' identification.

In this study, we confirm using multiparameter flow cytometry that not only Tregs, but also ICOS⁺ Tregs, are overrepresented in FL tumors. We observed that ICOS⁺ Tregs expressed high levels of CXCR5, PD1, GITR, and CTLA-4, and thus fit the profile of T_FR that were described in mice and humans (32, 33) and in FL tissues as well (31). Of note, we demonstrate herein that FL B cells are able to promote generation of these intratumoral ICOS⁺/CD25^highFoxP3^high Tregs via ICOS/ICOSL interaction, which induced ICOSL downregulation on FL B cells. Furthermore, we provide evidence of a strong suppressive effect of intratumoral Tregs on malignant B cells.

Several studies have shown a favorable outcome for FL patients with high amounts of intratumoral Tregs, although this remains controversial (18, 43). These observations support the hypothesis that Tregs could control malignant FL cells, but no demonstration of such functional interaction was reported so far. In nontumoral human tonsils, B-cell responses were shown to be inhibited by Tregs (44, 45). We report herein that Tregs are able to suppress the activation process and cytokine secretion of FL B cells, either directly or indirectly.

Tregs are known to use various mechanisms to suppress immune responses, including cell-to-cell contact (45). Additionally, Tregs are known to kill B cells through Fas–FasL interaction or through the release of cytolytic molecules (46, 47). So, further study is needed to identify mechanisms involved in direct FL B cell inhibition. The indirect suppressive effect of Tregs upon FL B cells is in accordance with a report showing that T_FH cells support FL B-cell activity via IL4 and CD40L (31). Our data suggest that Tregs are able to inhibit IL4/CD40L-producing T cells, probably including T_FH, which could hamper FL growth.

Nonetheless, the favorable prognostic value of high numbers of Tregs in FL tumors remains debated (11–13). In mouse models, depletion of Tregs induced decreased lymphoma burden (48). Unfavorable outcome was reported for FL patients with high levels of peripheral blood Tregs (49). A possible explanation that may reconcile the latter studies with our results may reside in the polarization of Tregs into distinct functional subsets. Tregs expressing Tbet, IRF4, STAT3, and Bcl6 appear to control Th1-, Th2-, Th17-, and T_FH-mediated immune responses, respectively (32, 50–52). Therefore, it is possible that intratumoral ICOS⁺ Tregs include multiple subsets, with opposite effects on tumor growth. The balance between these antagonist Tregs subsets may vary according to complex cues from the microenvironment. In this extent, it is noteworthy that we also observed a significant Tregs' increase in tissue samples from cHL, a lymphoma subtype that is known to undergo, like FL, crucial influences from the microenvironment, including a suspected favorable effect of Treg accumulation. This raises the question as to whether similar signaling networks could regulate Tregs in FL and cHL. The dependence of Tregs' status on the microenvironment of particular lymphoma subtypes is also supported by the previous observation that impoverishment of the Treg pool in angio-immunoblastic T-cell lymphoma (AITL) resulted from a proinflammatory microenvironment involving Th17 cells (53, 54).

We observed an increase in the amount of Tregs, including ICOS⁺/CD25^highFoxp3^high Tregs, among CD4⁺ T cells in coculture with FL B cells. This result is consistent with recent reports of Treg differentiation induced by malignant B cells (22–24). Tregs' increase could be due to either Tconv conversion or Tregs expansion. Our observation of significant Ki67 expression in ICOS⁺ Tregs favors the view of Tregs' expansion. On the other hand, it is also possible that lymphoma cells could convert Tconv to Tregs, as previously reported (22, 23). Tregs enrichment in FL tumors probably result from the combination of diverse mechanisms, including conversion and expansion, as well as migration of Tregs induced by CCL17- and CCL22-secreting FL B cells (55).

Survival, expansion, and suppressive functions of human Tregs are known to be strongly dependent on ICOS costimulation in physiologic settings (25). Accordingly, we show that coculture of ICOSL⁺ FL B cells with T cells induced generation of ICOS⁺/CD25^highFoxP3^high Tregs. Using neutralizing antibodies, we show that ICOS/ICOSL interaction is critical for this induction. ICOSL expression on FL B cells was downregulated following interaction with ICOS⁺ T cells, which may explain why ICOSL was undetectable in vivo. It remains, however, difficult to decipher what could be the primum movens of the cascade of ICOS/ICOSL interactions in vivo. Whatever the triggering event could be in FL, such a mechanism is globally reminiscent to that described in other human cancers, in which ICOS⁺ Tregs' enrichment is triggered by ICOSL produced by pDCs or tumor cells (26–28). Thus, our findings provide a basis to target the ICOS/ICOSL pathway for the development of future immunotherapies in FL patients.

No potential conflicts of interest were disclosed.

Conception and design: K.-S. Le, S. Pastor, C. Caux, C. Ménétrier-Caux, L. Xerri, D. Olive

Development of methodology: K.-S. Le, M.-L. Thibult, S. Just-Landi, S. Pastor, F. Gondois-Rey, C. Caux, D. Olive

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.-L. Thibult, F. Broussais, R. Bouabdallah, R. Colisson, D. Leroux, L. Xerri, D. Olive

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-S. Le, F. Gondois-Rey, S. Granjeaud, R. Colisson, C. Ménétrier-Caux, L. Xerri, D. Olive

Writing, review, and/or revision of the manuscript: K.-S. Le, R. Bouabdallah, C. Caux, C. Ménétrier-Caux, L. Xerri, D. Olive

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.-S. Le, M.-L. Thibult, S. Just-Landi, R. Bouabdallah, R. Colisson, D. Olive

Study supervision: L. Xerri, D. Olive

Other (performed experiments): K.-S. Le

The authors thank Pr. Thierry Fest (CHU de Rennes, Rennes, France), Pr. Richard Kroczek (Molecular Immunology, Robert Koch-Institute, Berlin, Germany), Cytometry platform, Experimental Histo-Pathology (ICEP) platform, IBiSA Cancer Immunomonitoring platform, and Department of Biopathology for their help.

This work was supported by grants from Institute National du cancer (INCa), Agence Nationale de la Recherche (ANR-11-EMMA-0045), Canceropole PACA, the pharmaceutical company GlaxoSmithKline (GSK) and by fellowships (K.-S. Le) from Ministère de l'Enseignement Supérieur et de la Recherche and the Ligue Nationale contre le Cancer (LIGUE). D. Olive team was labeled “Equipe FRM DEQ 201 40329534.” D. Olive is a senior scholar of the Institut Universitaire de France.

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.