In breast carcinomas, patient survival seems to be negatively affected by the recruitment of regulatory T cells (Treg) within lymphoid aggregates by CCL22. However, the mechanisms underpinning this process, which may be of broader significance in solid tumors, have yet to be described. In this study, we determined how CCL22 production is controlled in tumor cells. In human breast carcinoma cell lines, CCL22 was secreted at low basal levels that were strongly increased in response to inflammatory signals [TNF-α, IFN-γ, and interleukin (IL)-1β], contrasting with CCL17. Primary breast tumors and CD45+ infiltrating immune cells appeared to cooperate in driving CCL22 secretion, as shown clearly in cocultures of breast tumor cell lines and peripheral blood mononuclear cells (PBMC) or their supernatants. We determined that monocyte-derived IL-1β and TNF-α are key players as monocyte depletion or neutralization of these cytokines attenuated secretion of CCL22. However, when purified monocytes were used, exogenous human IFN-γ was also required to generate this response suggesting a role for IFN-γ–producing cells within PBMCs. In this setting, we found that human IFN-γ could be replaced by the addition of (i) IL-2 or K562-activated natural killer (NK) cells or (ii) resting NK cells in the presence of anti-MHC class I antibody. Taken together, our results show a dialogue between NK and tumor cells leading to IFN-γ secretion, which in turn associates with monocyte-derived IL-1β and TNF-α to drive production of CCL22 by tumor cells and subsequent recruitment of Treg. As one validation of this conclusion in primary breast tumors, we showed that NK cells and macrophages tend to colocalize within tumors. In summary, our findings suggest that at early times during tumorigenesis, the detection of tumor cells by innate effectors (monocytes and NK cells) imposes a selection for CCL22 secretion that recruits Treg to evade this early antitumor immune response. Cancer Res; 71(19); 6143–52. ©2011 AACR.
Cancer immunosubversion is a process by which tumor cells escape destruction by the immune system through a variety of mechanisms including the production of immunosuppressive cytokines and the alteration of dendritic cell (DC) functions (1, 2).
Several studies have shown that immune cells are present and functional in solid tumors and may promote both humoral and cellular antitumor immune responses. As an example, high levels of CD8+ T cells within the tumors have been associated with a better clinical prognosis in colorectal cancer (3). However, in most of the cases these T cells are unable to counteract tumor progression. In cancer patients, increased levels of CD4+CD25highFOXP3+ regulatory T cells (Treg), a lymphocyte subset with immunosuppressive properties, are described in the peripheral blood, the primary tumor microenvironment, and in the draining lymph nodes, supporting a role for Treg in cancer-induced immunosuppression. However, their effect on tumor progression varies according to the tumor type in humans. Treg have a negative impact on survival in lung, pancreatic, gastric, liver, or ovarian carcinoma patients (4–7), whereas they may exert a beneficial role in B-cell lymphoma, head and neck, or colon carcinoma (8, 9) or have no impact in colon, prostate, renal, or anal squamous cell carcinoma (10, 11; for review, see ref. 12).
We recently obtained evidence, in breast carcinoma, that selectively activated Treg accumulation within lymphoid aggregates, but not in the tumor bed, has a negative impact on patients' survival (13). Elucidating the mechanisms involved in Treg trafficking and accumulation in the breast tumor environment is thereby critical for innovative therapeutic development to fight tumor-induced immunosuppression.
Experiments in mice using Treg from CCR4−/− or conditional CCR4 knockout in FOXP3+ Treg compartment have recently identified the critical role of CCR4 in Treg trafficking in secondary lymphoid organs or tissues (14, 15). Curiel and colleagues strongly suggested a role for CCR4/CCL22 axis in Treg recruitment in ovarian ascitis (4).
We recently showed the selective loss of membrane CCR4 on tumor-associated Treg (TA-Treg) consecutive to an active recruitment through a CC chemokine (CCL22), and that breast tumors lacking CCL22 are not colonized by Treg independently of their CCL17 expression status (13), strongly suggesting the importance of CCL22 in TA-Treg recruitment within breast tumors.
CCL22, produced by myeloid DCs (mDC), B cells, macrophages, keratinocytes, or epithelial cells (6, 16, 17) and CCL17 closely related to CCL22, produced by monocyte-derived DC (18) and keratinocytes (19), are 2 ligands for CCR4 (6, 20) preferentially expressed on Th2 lymphocytes (21) and Treg (for review, see ref. 12). In peripheral blood mononuclear cells (PBMC), CCL22 is upregulated by interleukin (IL)-4, whereas it is downregulated by IFN-γ treatment (22). In contrast, IFN-γ favored CCL22 secretion by keratinocytes (16, 23) and intestinal epithelial cells (24).
In this study, we showed that breast tumor cell recognition by NK cells leads to their activation and IFN-γ secretion, which in turn triggers CCL22 production by tumor cells through cooperation with monocyte-derived IL-1β and TNF-α.
Materials and Methods
Breast tumor cell lines culture
All tumor cell lines used in this study originated from American Type Culture Collection except CLB-SAV generated in the laboratory. Cell lines were cultured in RPMI 1640 (Invitrogen) completed with 10% FBS (Lonza), 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen; complete medium) at 37°C in a 5% CO2 incubator.
Primary breast tumors
Breast tumor tissues collected at the Centre Léon Bérard after patient informed consent were mechanically dilacerated to obtain “mechanic tumor disaggregation supernatants” and then subjected to enzymatic digestion as previously described (13).
Flow cytometry analyses (ADP Cyan; Beckman Coulter) were conducted to assess the percentage of NK cells (CD3−NKp46+) and macrophages (CD4+CD68+CD163+; all from Becton Dickinson except for CD163 from eBiosciences and CD68 from Dako Cytomation) within primary tumor cell suspension after gating on CD45+ cells, and data were analyzed with FlowJo Analysis Software (Tree Star).
Immune cells (CD45+) or NK cells (CD3−CD56+) and macrophages (CD4lowCD163+) were purified on single-cell suspension from breast primary tumor or ascitis, respectively, by cell sorting (FACS Aria; Becton Dickinson).
Breast tumor supernatants
Single-cell suspensions from primary or metastatic (ascitis, pleural effusion) breast tumors were incubated at a final concentration of 15 × 106 cells/mL in complete medium in petri dishes. Cultured tumor cell supernatants were collected after 48 hours, filtrated on 0.22 μm, and frozen.
Expression of CCL22 on paraffin-embedded sections of breast tumor or peritumoral tissue was analyzed with a goat anti-CCL22 antibody (Ab; Santa Cruz) as previously described (13). Routinely used CD163 (mIgG1; Menarini Diagnostics) staining was carried out according to the manufacturer. NKp46+ was detected as previously described (25) cells with a goat immunoglobulin G Ab (R&D Systems). Hematoxylin-counterstained sections were dehydrated and mounted. For negative control slides, primary antibodies were replaced by a nonimmune serum.
Cytokines and antibodies
Recombinant human granulocyte macrophage colony-stimulating factor (rhGM-CSF; specific activity: 2 × 106 U/mg, used at 100 ng/mL) and rhIL-4 (specific activity: 106 U/mg, used at 50 ng/mL) were from Schering Plough Research Institute. rhTNF-α (specific activity: 5 × 106 U/mg) at 10 ng/mL was provided by Cetus Corporation. rhIL-1β (109 IU/mg), rhIFN-γ (2 × 107 IU/mg), and IL-1RA were from Peprotech. rhCCL22, rhCCL17, rhCXCL12, and monoclonal Ab (mAb) against CXCL12 and CCL22 as well as isotype controls used for neutralization experiments were from R&D Systems.
Purification of cell subsets from peripheral blood
Total PBMCs were isolated from heparinized blood obtained from healthy volunteers by Ficoll Hypaque density gradient centrifugation (Dominique Dutscher). Purified mDCs and monocyte fractions were obtained using positive selection kits, whereas untouched NK cells were purified using negative selection kit (Miltenyi Biotech) and purity was confirmed by flow cytometry. For depletion experiments, different cell subsets [myeloid cells, (mDC), plasmacytoid DCs (pDC), monocytes, NK cells, and T cells] were specifically depleted from PBMCs using positive selection kits with magnetic beads. The absence of remaining positive cells in the depleted fraction was confirmed by flow cytometry.
Tumor cell lines were cultured at 2 × 105 cells/mL in complete medium in 48-well plates (Becton Dickinson) and incubated for 24 or 48 hours in medium condition or in the presence of rhIFN-γ (0.1–100 ng/mL depending on the studies). Coculture experiments were carried out by incubating 105 tumor cells with 106 PBMC for 24 or 48 hours in the presence of 100 ng/mL rhGM-CSF with or without rhIFN-γ. To characterize the cell subset responsible for CCL22 secretion, PBMC cell supernatants (PBMC-SN) and tumor cell supernatants (TUM-SN) were generated by 24-hour incubation of either 106 PBMC or 105 tumor cells in 48-well plates (500 μL).
CCL22, CCL17 and IL-1β, and TNF-α levels were quantified in cell supernatants using ELISA from R&D Systems and Bender MedSystems, respectively.
Migration in response to CCL22
CCR4 expression on the CCRF-CEM cell line was confirmed by flow cytometry (Supplementary Fig. S1). Migration assays were conducted using Transwell (6.5-mm diameter; CoStar) with 5 × 105 cells/well. After 2 hours of preincubation at 37°C, CCRF-CEM cells were placed in 3-μm pore size inserts (100 μL) and tested for their ability to migrate in response to rhCCL22 (1–50 ng/mL) or culture supernatants (50%) added in the lower well. After 1 hour and 30 minutes of incubation at 37°C, cells were collected in cold PBS–EDTA and resuspended after centrifugation in 100 μL. The number of migrated cells was analyzed by flow cytometry. In blockade experiments, anti-CXCL12 or -CCL22 mAb or their isotype controls were incubated for 30 minutes with culture supernatants before CCRF-CEM cells were added in the insert.
High levels of CCL22 but not CCL17 are detectable in primary breast tumors
Analyses of breast TUM-SN showed the production of high levels of CCL22 in TUM-SN from primary tumors (Fig. 1A) that decreased in metastatic ones. CCL22 was also detectable in supernatants of mechanical tumor disaggregation from 27 primary tumors (mean = 1.02 ng/mL; range, 0.13–6.9 ng/mL). Low levels of CCL17 were detected in these TUM-SN. Moreover, supernatant from nontumor tissues (healthy breast tissue and fibro-adenoma) did not produce significant levels of CCL22 or CCL17.
We have previously shown that expression of CCL22, but not that of CCL17 by tumor cells in breast tumors, correlates with TA-Treg infiltration (13).
Breast tumor cell lines produce functional CCL22 in response to Th1/innate immunity but not Th2 signals
Contrasting with primary tumors, spontaneous CCL22 secretion by breast tumor cell lines in vitro was low to undetectable (Fig. 2A), suggesting mechanisms of regulation and a role of the microenvironment in CCL22 expression by tumor cells. Indeed, addition of rhIFN-γ, a CCL22 inducer on keratinocytes (16, 23) and intestinal epithelial cells (24), induced strong CCL22 secretion on 5 of 7 tested cell lines (0. 28–1.1 ng/mL for IFN-γ; Fig. 2A). In contrast, in PBMCs, CCL22 production was downregulated by rhIFN-γ but upregulated by rhIL-4 (Fig. 2B).
When bulk primary breast tumor disaggregation was used, the secretion of CCL22 was lost upon depletion of CD45+ immune cells. This CCL22 production by CD45-negative primary tumor was restored (9-fold increase) either by addition of associated CD45+ infiltrate (3 × 104 CD45+ for 8 × 104 tumor cells) or rhIFN-γ (Fig. 2C), strongly suggesting the cooperation between tumor cells and immune cells for specific CCL22 secretion.
This observation was confirmed using breast tumor cell lines. The addition of allogeneic PBMCs to breast tumor cell lines strongly enhanced the production of CCL22 but not of CCL17. We observed a 12.8-, 18-, and 121-fold increase for CLB-SAV, MDA-MB453, and MCF-7 cell lines, respectively, in coculture condition when compared with tumor cells alone. This secretion was further enhanced by rhIFN-γ addition (17- and 5.5-fold increase in CCL22 production for CLB-SAV and MDA-MB453 or MCF-7, respectively; Fig. 2D).
To test the functionality of the CCL22 secreted in [PBMC/tumor cell] coculture supernatants, we used the CCRF-CEM T-cell line that expresses CCR4 and migrates in response to rhCCL22 in a dose-dependent manner (Supplementary Fig. S1A and B). The coculture supernatant favored the CCRF-CEM cell migration in a Transwell Assay (5-fold increase over background level) that was specifically blocked by preincubation of these supernatants with an anti-CCL22 neutralizing mAb but not with an anti-CXCL12 able to attract CXCR4+ CCRF-CEM (Supplementary Fig. S1C).
Soluble factors produced by activated PBMCs induce CCL22 secretion by tumor cells
Whereas PBMCs alone are devoid of IFN-γ secretion (Supplementary Fig. S3, bottom), coculture with tumor cells increased this production favoring CCL22. Furthermore, as previously mentioned, addition of rhIFN-γ strongly enhanced CCL22 production in [PBMC/tumor cell] coculture while decreasing that of PBMCs (Fig. 2D). We therefore wished to decipher the relative role of PBMCs and tumor cells in CCL22 secretion in the [PBMC/tumor cell] coculture in the presence of rhIFN-γ. We compared the impact of rhIFN-γ-activated PBMC-SN on CCL22 production by tumor cell lines with that of rhIFN-γ-activated TUM-SN on PBMCs. As shown in Fig. 3, CCL22 levels secreted by rhIFN-γ-treated tumor cells were strongly enhanced in the presence of [rhIFN-γ-PBMC-SN] (2.75-, 19.36-, and 16.46-fold, respectively, for MDA-MB453, MCF-7, and CLB-SAV, respectively). In contrast, [rhIFN-γ-TUM-SN addition did not affect the low CCL22 levels detected in rhIFN-γ-treated PBMC cultures. Together, those data strongly suggest that rhIFN-γ-treated PBMCs produce soluble factors capable of inducing CCL22 production by tumor cells.
Interestingly, the effects observed were specific for CCL22 as CXCL8 that was produced by rhIFN-γ-activated PBMCs was downregulated in the presence of tumor cells and not induced when tumor cells were cultured with PBMC-SN (data not shown).
Monocytes and IFN-γ are both required for the secretion of CCL22 by tumor cells
To determine the major cell fraction within PBMC responsible for the effects observed on tumor cells, specific depletions of myeloid cells (CD33+), monocytes (CD14+), mDC (BDCA1+BDCA3+), pDC (BDCA2+), NK cells (CD56+), or T cells (CD3+) were carried out using magnetic beads. Each depleted fraction was added on tumor cell lines in the presence of rhIFN-γ to assess the CCL22 production. As shown in Fig. 4A, whereas addition of PBMCs induced a strong CCL22 production (2.95 ± 0.1 ng/mL), we observed a drop in this secretion when monocyte (CD14−) or myeloid cell (CD33−)-depleted fractions were used (86% and 75% inhibition, respectively). In contrast, the depletion of NK cells or T cells did not decrease the CCL22 secretion. The increase observed with T-cell depletion likely results from increased monocyte percentage in the culture. The depletion of mDCs reduced the basal level of CCL22 produced by PBMCs alone as shown in Supplementary Fig. S2 (26) but did not affect the CCL22 production by tumor cells. These results suggest that monocytes are the main actors in CCL22 production by tumor cells within the coculture.
To confirm their role, purified monocytes were added to tumor cells in the presence or not in the presence of rhIFN-γ (Fig. 4B). Whereas purified monocytes were not able to mimic PBMC action on tumor cells, further addition of rhIFN-γ-induced CCL22 levels comparable with those obtained with PBMC. This effect was specific to monocytes as mDC, even in presence of IFN-γ, did not reconstitute PBMC effect. Of most importance, these results suggest that monocytes act in cooperation with other cell subsets (i.e., NK cells, NKT cells, or T cells) capable of IFN-γ secretion to increase CCL22 secretion by tumor cells.
Involvement of rhIL-1β and TNF-α in inducing CCL22 secretion by tumor cells
As shown above (Fig. 4A), depletion of monocytes strongly reduced the ability of tumor cells to produce CCL22 in coculture. Monocytes are strong producers of IL-1β and TNF-α, previously described to cooperate with IFN-γ in CCL22 production on epithelial cells or keratinocytes (24, 27, 28). As shown in Supplementary Fig. S3, whereas PBMC produced low IL-1β and TNF-α levels (6.5 ± 0.2 and 99 ± 1 pg/mL, respectively), the addition of rhIFN-γ increased their secretion (IL-1β, 133 ± 12 pg/mL; and TNF-α, 879 ± 53 pg/mL). These 2 cytokines are also detected in [tumor cell line/PBMC] coculture in the presence of rhIFN-γ. Moreover, the loss of CCL22 production in monocyte-depleted fraction was associated with the absence of IL-1β and TNF-α secretion in the coculture (data not shown).
Whereas CCL22 production by MDA-MB453 cells was mostly dependent on IFN-γ (Supplementary Fig. S4A and C), the culture of MCF-7 (Supplementary Fig. S4B and D) or CLB-SAV (data not shown) with a cross range of recombinant cytokines showed an important impact of low doses of IL-1β (100 pg/mL) or TNF-α (10 ng/mL) on CCL22 production with an additive effect of IFN-γ (Supplementary Fig. S4B and D).
To confirm a role for these 2 cytokines within rhIFN-γ-PBMC-SN, in CCL22 secretion by tumor cells, we tested the impact of IL-1 receptor antagonist (IL-1RA) or an anti-TNF-α blocking Ab, previously validated (Supplementary Fig. S5A), on tumor cell line cultures. As shown in Fig. 4C, treatment with either IL-1RA or anti-TNF-α mAb was able to block 40% of CCL22 secretion induced by rhIFN-γ-PBMC-SN. The simultaneous blockade of IL-1β and TNF-α decreased CCL22 secretion up to 80%, showing the role of IL-1β and TNF-α contained in rhIFN-γ-PBMC-SN on CCL22 secretion by tumor cells.
NK cells and monocytes cooperate to induce CCL22 production by tumor cells
To better understand the mechanisms involved in IFN-γ secretion within [PBMC/tumor cell] coculture (Supplementary Fig. S3) that act in synergy with IL-1β and TNF-α, we hypothesized that NK cells could be activated and secrete IFN-γ after interaction with tumor cells.
To mimic NK activation, NK cells were pretreated with IL-2 for 16 hours. We tested their impact on CCL22 production by tumor cells in the presence of purified monocytes or mDC-depleted PBMC fraction. Whereas activated NK cells, mDC-depleted fraction, or purified monocytes each alone (Fig. 5A and B) did not trigger CCL22 production by tumor cells, a combination of activated NK cells with either mDC-depleted fraction (Fig. 5A) or purified monocytes (Fig. 5B) induced CCL22 levels comparable with those obtained in the presence of PBMCs or exogenous rhIFN-γ (Fig. 5A). This suggests that IFN-γ released by activated NK cells cooperates with monocytes to promote CCL22 release by tumor cells.
Interaction of K562 tumor cell line with NK cells also favors their activation (29). The addition of K562 (1:1 K562:NK ratio) to resting NK cells in the presence of purified monocytes and tumor cells increased CCL22 secretion by tumor cells that was dependent on IFN-γ as anti-IFN-γR1 blocking mAb reversed this effect (Supplementary Fig. S6).
Because tumor cell lines upregulate MHC class I in response to IFN-γ or TNF-α (Supplementary Fig. S7), we neutralized MHC class I expression on tumor cells as an alternative approach to revert blockade of NK activation through MHC class I/killer inhibitory receptor (KIR) interactions. The preincubation of tumor cells with blocking anti-MHC class I mAb (W6/32) before the addition of resting NK cells and monocytes significantly increased the CCL22 production. This increase was strictly dependent on the presence of NK cells and monocytes (Fig. 6). In these experimental conditions, addition of IL-1RA, anti-IFN-γR1, and anti-TNF-α mAb used alone have all shown a moderate to strong effect depending on the cell line, but when combined they completely blocked anti-MHC class I impact, showing the involvement of IFN-γ, TNF-α, and IL-1β (Fig. 6).
NK cells and macrophages colocalize with tumor cells in situ
As showed in Fig. 7A and B, NK cells (CD3−NKp46+) as well as macrophages (CD163+CD68+) were detected within the primary tumor cell suspensions by flow cytometry [mean = 3.72% (0.15%–8.2%) for NKp46+ and mean = 11.7% (0.58%–37.1%) for CD163+]. As shown by IHC on paraffin-embedded tumor sections, NK cells (NKp46+; Fig. 7C and D) as well as macrophages (CD163+; Fig. 7E and F) are localized in the vicinity of tumor cells. Moreover, purified in situ activated ascite-derived macrophages are able to cooperate with NK cells to favor a strong CCL22 production by breast tumor cell line (Fig. 7G). Together, these data suggest the potential recognition of tumor cells by NK cells favoring in combination with macrophages, the initiation of CCL22 secretion by these tumor cells.
In this study, we showed that recognition of transformed mammary epithelial cells favors NK cell activation and subsequent IFN-γ secretion associated with the release of monocyte-derived IL-1β and TNF-α that triggers CCL22 production by tumor cells. This tumor cell–associated CCL22 secretion favors blood CCR4+ Treg recruitment leading to the development of a tolerogenic environment conducive to the tumor immunosubversion and development.
We previously reported a strong correlation between CCL22 expression by tumor cells and the presence of Treg within breast tumor environment. Breast tumors lacking CCL22 are not colonized by Treg independently of their CCL17 expression status (13). Moreover, Treg recruitment in tumor environment induces a loss of CCR4 expression, a phenomenon observed when Treg are cultured in vitro with CCL22 but not with CCL17 (30). Similar observations were made for CCR7 expression that was downregulated on T cells after interaction with CCL19 but not with CCL21 (31, 32). In agreement with our observation, CCR4 and CCL22 requirement for Treg recruitment was also reported in the mouse model of inflammatory bowel disease, in which the inability of CCR4−/− Treg to migrate within the colon tissue leads to disease exacerbation (15).
Our IHC analyses show, in peritumoral breast tissue samples, polarized apical CCL22 secretion by healthy luminal epithelial cells within lobular acini as described for other chemokines (CXCL8, GROβ, GROγ, GROα, ENA78, MIG, IP10, and RANTES) detected in the milk or the colostrum (33, 34). Moreover, nonhematopoietic cells such as keratinocytes and epithelial cells can secrete CCL22 (16, 24, 35). Polarized CCL22 secretion toward the lumen has also been described in colon epithelium (24). Moreover, the cyclic hormonal modulation may also affect Treg recruitment within the mammary gland via CCL22 secretion by epithelial cells. Indeed, treatment of women with progesterone favors in the endometrium a high CCL22 production by stromal cells and glandular epithelial cells at the end of the hormonal cycle (35). Taken together, these results suggest that CCL22 secretion within the breast tissue may be part of the mammary gland physiology controlling the local inflammation associated with tissue remodeling either at the end of the menstrual cycle or during breastfeeding.
In accordance with structural disorganization characteristic of primary breast tumor tissue, we observed that CCL22 secretion is no more polarized favoring its diffusion within the tumor environment that may favor recruitment of macrophages, NK cells, Th2 cells (6), and Treg (for review, see ref. 12) expressing CCR4. Moreover CCL22 production is strongly enhanced when compared with healthy tissue. This is consistent with the levels of CCL22 found either in primary breast tumor mechanical disaggregation SNs (median = 1.16 ng/mL; range, 0.23–8.8 ng/mL) or in 48-hour culture primary breast TUM-SN with more than 40-fold increase in CCL22 levels (median = 2.91 ng/mL; range, 0.53–12.4 ng/mL) in comparison with nontumor SN (median = 0.07 ng/mL; range, 0.03–0.23 ng/mL; P = 0.004). It is also important to notice that CCL22 content is 10-fold higher than that of CCL17 (median = 0.3 ng/mL; range, 0–4.4 ng/mL). Importantly, we show using either primary breast tumor or tumor cell lines the cooperation between tumor cells and immune infiltrate to induce high quantities of CCL22, whereas CCL17 secretion remains barely detectable. Interestingly, although healthy bronchial epithelial cells secrete CCL17 (36), their tumor counterpart in lung carcinoma pleural effusion produces CCL22 (37). Taken together, these results suggest the capacity of the tumor environment to modulate the chemokine arsenal of epithelial cells to favor the migration of specific cell subsets. CCL17, via the recruitment of Th2 CCR4+ cells, will favor a Th2 response as described in atopic dermatitis (38), whereas CCL22 is more specialized in the recruitment of Treg as observed in tumors (for review, see ref. 12).
IL-4 and IL-13, critically involved in the development of cutaneous pathologies like atopic dermatitis, have been largely shown to induce CCL22 secretion by cells of myeloid origin (monocytes and mDCs; refs. 22, 39) and to favor CCL17 production by fibroblasts (40). In contrast, we showed in this study that IL-4 reduces the CCL22 production in breast tumor epithelial cell lines as previously described for immortalized keratinocytes (16, 23), colon epithelial cells (24), and glioma cell lines (41).
In this study, we deciphered the mechanisms involved in the increased secretion of CCL22 within the tumor environment. We showed the existence of a dialogue between tumor cells and circulating immune cells leading to CCL22 production by tumor cells and to Treg recruitment. We reported that breast tumor cell lines produced CCL22 in response to rhIFN-γ, as previously described for keratinocytes (16, 23). This secretion is strongly enhanced in coculture with PBMCs but is lost after myeloid cell (CD33+) or monocyte (CD14+) depletion showing the major role of monocytes in this CCL22 secretion although they do not secrete CCL22 by themselves.
NK cells constitute a unique component of the innate immune system able, without specific sensitization, to recognize autologous cells undergoing various form of stress, such as malignant transformation (42). Target recognition occurs via the integration of negative and positive signals mediated by inhibitory (KIR) or activating (KAR) receptors expressed at the surface of NK cells. Breast tumor cells expressing ULBP or MICA/MICB markers that bind NKG2D on NK cells will stimulate their IFN-γ secretion (Supplementary Fig. S5; refs. 43, 44). However, expression of MHC class I (Supplementary Fig. S7), a KIR ligand, by breast tumor cells reduced this IFN-γ secretion. In coculture of breast tumor cell lines with purified monocytes, rhIFN-γ could be omitted upon addition of NK cells in conditions leading to their activation, that is, in the presence of IL-2, K562 NK target cell line, or anti-MHC class I Ab. All these culture conditions lead to IFN-γ secretion required for CCL22 production as shown by the use of blocking anti-IFN-γR Ab.
In this cell line, MICA (NKG2D-L) expression in breast tumors has been associated with a poor prognosis (43). This could result either from the production of soluble MICA that blocks the killing function of NK cells or the impact of IFN-γ secretion by NK cells on Treg recruitment through CCL22 secretion by tumor cells. In the Lewis Lung carcinoma mouse model, depletion of NK cells blocked CCL22 production in the tumor environment; however, NK cells were proposed as the major source of CCL22 (45). In contrast, we never detected, in our experimental set-up, CCL22 secretion by resting as well as activated NK cells.
The replacement of purified monocytes/macrophages by the combination of rhIL-1β and rhTNF-α in the culture of breast tumor cells with rhIFN-γ or NK cells strongly enhanced the CCL22 production. This observation is in agreement with previous publications on keratinocytes and colon epithelial cell lines (16, 24), whereas blockade of IL-1β and TNF-α abrogate this secretion in [PBMC/tumor cell] coculture. Interestingly, IFN-γ increased IL-1R1 and TNF receptor on tumor cell lines (data not shown), as previously described (46), suggesting a potential amplification loop of CCL22 production. Taken together, these results suggest the importance of inflammation in the high CCL22 levels in the breast tumor environment that will favor Treg recruitment leading to reduced specific antitumor immune response. This is in agreement with studies in colon tissue reporting the involvement of intestinal flora-mediated chronic inflammation in the increased recruitment of Treg (for review, see ref. 47). This suggests that inflammation in the mammary gland may participate in the tumor development. In favor of this, TNF-α secretion by leukocytes infiltrating tumors strongly contributes to mammary carcinogenesis in murine mammary models (48). Importantly, the in situ analyses on primary breast tumors allow us to show the presence of NK cells and macrophages in the vicinity of tumor cells.
Treg have been described to reduce NK cell cytotoxicity (for review, see ref. 49) suggesting that CCL22 production by tumor cells inducing Treg recruitment represents one of the mechanisms elaborated by tumors to avoid its destruction through NK cell cytotoxicity.
Taken together, our results allow us to propose a model in which mammary epithelial cell transformation processes favored activation of NK cells present in the breast tissue by reducing KIR and inducing KAR ligand expression and their subsequent IFN-γ secretion, leading to the production of TNF-α and IL-1β by resident monocytes/macrophages. Acting together, these 3 cytokines will favor CCL22 overproduction by tumor cells, allowing the recruitment of CCR4+ blood Treg that favor the development of a tolerogenic environment.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
J. Faget and M. Gobert are grant holders of the Ligue Nationale contre le Cancer. This work was financially supported in part by grants from “le comité départemental du Rhône de Ligue Contre le Cancer,” the ARC Association (ARC-5074), the Breast Cancer Research Foundation and Institut National du Cancer grant INCA ACI-63-04, ACI 2007-2009.
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.