NFAT5-Regulated Macrophage Polarization Supports the Proinflammatory Function of Macrophages and T Lymphocytes

Macrophages can react to multiple signals generated during development or upon disruption of tissue homeostasis to engage in processes such as organogenesis, antipathogen responses, tissue repair, or antitumor defense (1). Their ability to perform these functions efficiently and to coordinate with other immune and nonimmune cells requires an impressive plasticity to reversibly acquire pro- or anti-inflammatory functions in response to microenvironment cues (2–4). Macrophages can adopt different functional phenotypes that are generally classified with respect to two main types: those conditioned by proinflammatory stimuli such as IFN-γ, also referred as classically activated or M1, and those polarized by anti-inflammatory cytokines such as IL-4 and IL-13, also known as alternatively activated or M2 macrophages (4–6). The capacity of macrophages to function in pro- or anti-inflammatory modes plays a key role in both the triggering and resolution of immune responses, but can also constitute a potential vulnerability for the immune system as imbalances in macrophage function can lead to pathological immune reactivity or tolerance. In this regard, the tumor microenvironment provides an interesting scenario, where despite proinflammatory and potentially antitumor macrophages being found (7), anti-inflammatory macrophages are the dominant type and control tumor progression by performing key protumoral trophic and immunotolerant functions (8–12).

Specialization into distinct macrophage subtypes relies on different transcription regulators that induce specific gene expression programs (13, 14). The transcription factors STAT1 and IRF5 are central players in the induction of the proinflammatory polarization of macrophages (15, 16), and activation of STAT6 is a central mechanism that induces their alternative polarization (5, 13). Together with these master regulators, other transcription factors are known to facilitate macrophage polarization. In this regard, alternative polarization was also found to be regulated in different contexts by p50/NF-κB dimers (17, 18), IFN regulatory factor 4 (19), Krüppel-like factor 4 (20), the CREB/C/EBPβ axis (21), c-Myc (22), and the nuclear receptors peroxisome proliferator activated receptor (PPAR) γ and PPARδ (23–25). The diversity of transcription regulators capable of modulating the acquisition of pro- and anti-inflammatory profiles by macrophages likely underlies their ability to register multiple types of cues for switching their function accordingly.

Despite the knowledge accumulated in transcription regulators of macrophage polarization, our understanding of this process is still incomplete. For instance, it is intriguing that, apart from the core regulators of inflammatory polarization STAT1, IRF5, and p65/NF-κB (13), other pathways involved in this process could also conserve the ability to support alternative polarization. This was observed for the Notch pathway, which supports inflammatory gene expression and antitumor responses (26, 27), but can also control immunoregulatory or alternatively activated macrophages (28, 29). Mouse models deficient for the transcription factor NFAT5 have become instrumental for the identification of specific roles of this factor in different immune cells. In this regard, although its function in B and T lymphocytes has been mainly studied in hypertonicity-induced transcription responses (30–32), in innate immune cells such as macrophages NFAT5 has been shown to be required for TLR-induced responses to pathogens (33), a role that NFAT5 performs in isotonic conditions (33) but that can be enhanced under hypernatremia (34). As we had previously found that many of the TLR-induced NFAT5-regulated genes, such as inducible NO synthase (iNOS) or TNF-α, have a central role in inflammatory responses (33), we were interested in understanding the contribution of NFAT5 to the balance between classically or alternatively activated macrophage functions.

Mice heterozygous for Nfat5 were described previously (35, 36) and used for generating NFAT5-deficient and littermate control mice in a 129/sv background. NFAT5-floxed mice in a pure C57BL/6 background were described previously (36, 37) and used to generate different conditional knockouts for NFAT5. By crossing them with mice carrying the Cre recombinase under the control of the Vav1 gene promoter [Vav-Cre (38)] or the LysM promoter [LysM-Cre (33)], we generated mice that deleted NFAT5 in hematopoietic and myeloid cells respectively. In addition, the Mx1-Cre transgene [Mx-Cre (39)] was used to generate an inducible deletion of Nfat5 in adult NFAT5-floxed mice by poly(I:C) (catalog tlrl-pic; InvivoGen) injection. For this, Mx-Cre NFAT5-floxed and littermate Mx-Cre control mice were injected three times every other day with poly(I:C) (15 mg/kg) and bone marrow was extracted 12 d later to generate bone marrow–derived macrophages (BMDM). All mice were analyzed between 6 and 10 wk of age. Mice were bred and maintained in specific pathogen-free conditions, and animal handling and experiments were in accordance with protocols approved by the ethics committee of the Barcelona Biomedical Research Park/ Pompeu Fabra University Animal Care and Use Committee, and carried out in accordance with the Declaration of Helsinki and the European Communities Council Directive (86/609/EEC).

To obtain BMDM, mice were sacrificed and the femoral and tibial marrow flushed with DMEM supplemented with 2 mM glutamine, 50 μM β-mercaptoethanol, 1 mM sodium pyruvate, and penicillin/streptomycin (incomplete medium) (all from Thermo Fisher Scientific) using 25G syringes. Then 10 × 10⁶ BM cells were incubated in complete medium (incomplete plus 10% FBS) with 25% (v/v) of L929-conditioned medium (as a source of M-CSF) in 150 mm-diameter sterile Petri dishes (catalog 82.1184.500; Sarstedt) for 6–7 d. Differentiated macrophages were collected by gentle pipetting after incubating 10 min with ice cold 1× PBS, 5 mM EDTA. For Western blot, ELISA and mRNA analysis, six-well plates (catalog 140675; Nunc, Thermo Fisher Scientific) were used and macrophages were plated at 1 × 10⁶ cells per well. For CD4⁺ T lymphocyte cocultures, 12-well plates (Labclinics, catalog PLC30012) were used with 0.7 × 10⁶ cells per well. For cocultures with tumor cells, 4 × 10⁶ BMDMs were plated in 100 mm-diameter plates (catalog 172958; Thermo Fisher Scientific) and incubated with Lewis lung carcinoma (LLC) tumor cells. For macrophage polarization analysis, replated macrophages were treated with 100 U/ml of recombinant murine IFN-γ or 10 ng/ml of recombinant murine IL-4 (catalogs 12343536 and 12340043; ImmunoTools) for 24 h to induce M1 and M2 polarization, respectively. Cells were then activated with LPS at 0.3 ng/ml (from Escherichia coli 055:B5, catalog L2880; Sigma) for the indicated time points.

Total RNA from BMDMs (1 × 10⁶) was isolated using the High Pure RNA Isolation System (catalog 11828665001; Roche) and quantified in a NanoDrop (ND-1000) spectrophotometer. Then 100–600 ng of total RNA was retro-transcribed to cDNA using the First Strand cDNA synthesis system with random primers (catalog 04897030001; Roche). For real-time quantitative PCR, LightCycler 480 SYBR Green I Master Mix (catalog 04887352001; Roche), LightCycler 480 Multiwell Plates (catalog 4729749001; Roche), and the LightCycler 480 Real-Time PCR System (Roche) were used according to the manufacturer’s instructions. For samples with low cell numbers (10³–10⁵ cells), the RNeasy Microkit (catalog 74004; Qiagen) was used and the RNA was retro-transcribed using the SuperScript III First-Strand Synthesis system (catalog 18080-051; Thermo Fisher Scientific). In all cases, samples were normalized to L32 (L32 ribosomal protein gene) mRNA levels using the LightCycler Software, version 1.5. Primer sequences are listed in Supplemental Table I.

For Western blotting, BMDMs were lysed in Triton X-100 lysis buffer (1 × 10⁶ cells in 100 μl; 1% Triton X-100, 40 mM Hepes pH 7.4, 120 mM NaCl, 1 mM EDTA, 1 mM PMSF, 5 μg/ml leupeptin, 5 μg/ml aprotinin, and 1 μg/ml pepstatin A, 1 mM sodium fluoride, 1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, and 10 mM β-glycerophosphate). Protein concentration was quantified using the BCA assay (catalog 23227; Thermo Fisher Scientific) and lysates were boiled in reducing 1× Laemmli buffer. Next, 10–50 μg of total protein were subjected to SDS-PAGE and transferred to PROTRAN membranes (BA83; Schleicher & Schuell). Membranes were blocked with 5% dry milk in TBS and immunoblotted with the specific Abs indicated below in TBST 0.05%. Membranes were then washed with TBST, incubated with a HRP-conjugated anti-rabbit secondary Ab, washed with TBST, and developed with enhanced chemoluminescence detection (catalog RPN2106; ECL Western Blotting Detection Reagents, Amersham). The primary Abs were rabbit polyclonal anti-NFAT5 Ab (PAI-123; Affinity BioReagents), rabbit anti-iNOS (catalog sc-651; Santa Cruz Biotechnology), rabbit anti-arginase 1 (catalog 9819; Cell Signaling Technology), and rabbit anti–Fizz-1 (catalog ab39626; Abcam).

Cell-free supernatants from BMDM cultures were analyzed for IL-12 and IL-6 production using commercially available ELISA kits (R&D Systems), mouse IL-12/IL-23p40 DuoSet ELISA (catalog DY2398-05), and mouse IL-6 DuoSet ELISA (catalog DY406-05), following the manufacturer’s instructions.

Briefly, 2 × 10⁵ cells were blocked for 20 min in 100 μl of 1× PBS containing 10% FBS, 0.1% sodium azide (staining buffer), and 0.2 μg of an Ab against the Fc receptor CD16/32 (catalog 101302; BioLegend). Cells were then incubated with surface marker-specific Abs in the same solution (1 μg of Ab per 10⁶ cells) for 30 min in the dark at 4°C. After washing them twice with staining buffer, cells were analyzed with FACSCalibur or LSR II flow cytometers. Data analysis was carried out using FlowJo software (Tree Star). Abs for flow cytometry were from eBioscience: CD4 PE-Cy5.5 (RM4-5 clone), CD8 APC (53-6.7 clone), CD45.2 PerCP-Cy5.5 (104 clone), Ly6G PE-Cy7 (1A8 clone), CD11b FITC (M1/70 clone), Ly6C PE (HK1.4 clone), MHC-II (IA/IE) APC (M5/114.15.2 clone); from BD Pharmingen: CD44 FITC (IMT clone), and CD62L PE (MEL-14 clone).

Wild-type and NFAT5-deficient BMDMs were seeded in 12-well plates in triplicate (0.7 × 10⁶ cells per well), polarized to M1 and M2 phenotypes for 18 h, and then treated with 0.3 ng/ml of LPS for 12 h before adding CD4⁺ T cells. CD4⁺ T cells were isolated from the spleen and peripheral lymph nodes of wild-type C57BL/6 mice by negative selection using the MagniSort Mouse CD4 T cell Enrichment Kit (catalog 8804-6821-74; eBioscience) according to the manufacturer’s instructions. Upon isolation, cells were incubated with 1 μg of anti-CD3 per million cells (catalog 553058; BD Pharmingen) for 1 h at 4°C, washed with DMEM medium, and incubated with the BMDM cultures in a 1:1 ratio for 48 h. CD4⁺ cells were harvested from the coculture supernatants by gentle pipetting, resuspended in 1× PBS plus 10% of FBS, isolated by positive selection (Dynabeads CD4 Positive Isolation Kit, catalog 11331D; Thermo Fisher Scientific), and lysed in RNA lysis buffer (Roche) for total RNA extraction.

Escherichia coli (AmpR) saturated cultures were opsonized with mouse serum before being incubated for 30 min at 37°C with wild-type and NFAT5-deficient M1 or M2 polarized macrophages using a multiplicity of infection of 1 (calculated for bacterial cell cultures as: OD₆₀₀ 1.0 = 8 × 10⁸ cells per ml). As a negative control of phagocytosis, a parallel assay was maintained at 4°C. For the killing assay, infected macrophages were treated for 1 h with a high dose of gentamicin (50 μg/ml) to kill extracellular bacteria, then washed and maintained in a low dose of gentamicin (6 μg/ml) for 0, 3, and 6 h. To determine phagocytosis and killing abilities, BMDMs were lysed at the indicated time points with 1% deoxycholate in 1× PBS and the lysates were seeded in ampicillin-containing agar plates to quantify bacterial colonies.

Two different syngeneic tumor models were used. ID8 cells derived from a C57BL/6 ovarian epithelial carcinoma were used as a syngeneic serous tumor model (40). ID8 cells stably expressing luciferase (ID8-luc) were a kind gift of Dr. K.F. Roby (University of Kansas), Dr. J.R. Conejo-Garcia (The Wistar Institute), and Dr. L. Zitvogel and Dr. P. Roberti (Institut Gustave Roussy). Cells were grown in DMEM supplemented with 5% FBS and 1× insulin-transferrin-sodium selenite media supplement (100×, catalog I1884; Sigma). For in vivo tumor development assays, 5 × 10⁶ subconfluent ID8-Luc cells in 200 μl of 1 x PBS were injected i.p. in 6–8 wk old C57BL/6 female mice, either wild-type (LysM-Cre Nfat5 ^+/+) or LysM-Cre Nfat5^fl/fl. Tumor growth was assessed between days 3 and 35 postinoculation by i.p. injection of 200 μl of luciferin at 12.5 mg/ml (catalog LUCK-500; GoldBio) and bioluminescent image detection and analysis was carried out using the IVIS200 (PerkinElmer). Analysis of peritoneal macrophages and T cell subsets was carried out 1 d after luminescence measurement, at days 4, 7, 14, and 36. Additional experiments were carried out in wild-type C57BL/6 mice first inoculated with ID8-Luc cells, and 30 d later injected i.p. with 9 × 10⁶ wild-type or NFAT5-deficient BMDM. Tumor growth was assessed by bioluminescence measurement 4 and 10 d after BMDM injection. On the last day of the experiment mice were sacrificed and peritoneal cells were harvested by peritoneal lavage using ice-cold 1× PBS and analyzed by flow cytometry to determine the distribution of lymphocyte subsets. LLC cells derived from C57BL/6 mice were kindly provided by Dr. I. Melero (Center for Applied Medical Research, Pamplona, Spain). LLC cells were grown in complete medium and maintained at subconfluency by passing them with gentle pipetting. For solid tumor development, a 1:1 mixture of 1 × 10⁵ LLC cells and wild-type or NFAT5-deficient macrophages were s.c. injected in the right back flank of 6–8 wk old C57BL/6 female mice. Alternatively, LLC cells (2 × 10⁵ cells per mouse) were injected in LysM-Cre Nfat5^fl/fl or littermate control mice. Tumor growth was periodically measured using a caliper and the tumor volume was calculated using the formula L × W2 × 0.52, where L = maximal length and W = maximal width. Mice were sacrificed at day 12 or 15.

Peritoneal macrophages from mice bearing ID8-Luc tumors were isolated by peritoneal lavage and positive selection using magnetic beads (Dynabeads sheep anti-rat IgG, catalog 11035; Life Technologies) coated with rat anti-CD11b M1/70.15 mAb supernatant. For tumor-associated macrophage (TAM) analysis in LLC tumor samples, tumors were minced using a scalpel and digested with 0.5 mg/ml of collagenase A (catalog 10103578001; Roche) plus 0.01% of DNaseI (catalog D4263-5VL; Sigma) in complete DMEM medium without β-mercaptoethanol during 1 h at 37°C in rotation. Samples were then filtered through a 70 μm cell strainer to remove undigested fragments and the filter washed by adding 20 ml of DMEM. Filtered cells were then centrifuged for 8 min at 1200 rpm and pellets resuspended in 500 μl of staining buffer. Macrophages were then isolated by positive selection with rat anti-CD11b M1/70.15 mAb and sheep anti-rat IgG magnetic beads.

Sodium chloride, Trizma base, glycine, EDTA, sodium orthovanadate, β-glycerophosphate, PMSF, leupeptin, pepstatin A, aprotinin, SDS, sodium pyrophosphate, methanol, BSA, Triton X-100, Tween 20, and lysogeny broth were purchased from Sigma-Aldrich. Sodium fluoride was from Merck. HEPES was from Lonza. GM-CSF and M-CSF were from ImmunoTools.

Statistical analyses were carried out using the GraphPad Prism 5 software. Normality (Gaussian distribution) of samples was determined by a D’Agostino–Pearson normality test before determining statistical significance with an unpaired t test (for sets of samples with a Gaussian distribution) or Mann–Whitney U test (samples with a non-Gaussian distribution). A two-way ANOVA test was used when comparing various groups of samples, and a one-sample t test was used when samples were compared with a reference control sample (set as 100%). The statistical analyses used are indicated in each respective figure legend.

As a first approximation to explore new roles of NFAT5 in pro- and anti-inflammatory macrophage functions, we analyzed how the lack of NFAT5 affected macrophages conditioned with IFN-γ or IL-4, as these cytokines are prototypical inducers of responses at both ends of the functional spectrum of macrophages. We also analyzed the response of IFN-γ and IL-4–polarized macrophages to subsequent stimulation with LPS. LPS was used to boost the acquisition of proinflammatory features in IFN-γ–polarized macrophages, and in the case of IL-4–polarized ones, to induce a switch from an anti-inflammatory to a proinflammatory state (Fig. 1A). For simplicity, we have used the nomenclature M1 to refer to macrophages preconditioned with IFN-γ, and M2 for macrophages conditioned with IL-4 alone. First, we observed that IFN-γ and IL-4 alone did not change the basal expression of NFAT5 in mouse BMDM, but found that IFN-γ pretreatment enhanced the induction of NFAT5 by LPS both at the mRNA and protein levels, whereas IL-4 did not, or even reduced it mildly (Fig. 1B, 1C). Similarly, we found that human macrophages derived from blood monocytes expressed more NFAT5 when treated with the proinflammatory cytokine GM-CSF, compared with those treated with the anti-inflammatory polarizing factor M-CSF (Fig. 1D). These results showed that expression of NFAT5 was higher in proinflammatory or M1 macrophages than in alternative or M2 ones. We then analyzed the contribution of NFAT5 to gene expression in macrophages polarized with IFN-γ or IL-4, using different mouse models of NFAT5 deletion in the whole body (35, 36) or in myeloid and hematopoietic cells (Supplemental Fig. 1A–D). IFN-γ conditioning strongly enhanced the induction by LPS of main proinflammatory gene products iNOS, IL-1β, IL-6, and IL-12β compared with macrophages stimulated only with LPS or conditioned with IL-4 before LPS (Fig. 2, Supplemental Fig. 1E). Parallel analysis of NFAT5-deficient macrophages revealed reduced mRNA and protein expression of inflammatory gene products iNOS, IL-12p40, and IL-6 in response to LPS stimulation (Figs. 2, 3, Supplemental Fig. 1E).

We also found that lack of NFAT5 affected the expression of several genes induced by IL-4. Arginase 1, CD163, YM1 (Chi3l3), and Fizz-1 (Retnla) expression was decreased in IL-4–polarized NFAT5-deficient cells (Fig. 2, Supplemental Fig. 1E). Other gene products such as mannose receptor C type 1 (MRC1 or CD206) and Gas6 were NFAT5 independent in IL-4 polarized macrophages, although we also noticed that expression of MRC1 mRNA was more resistant to downregulation by LPS in NFAT5-deficient macrophages. Attenuated induction of the prototypical IL-4–inducible products arginase 1 and Fizz-1 in NFAT5-deficient macrophages was also observed at the protein level (Fig. 3D). Regulation of IFN-γ– and IL-4–responsive genes by NFAT5 was consistently observed in macrophages from two different mouse models of conditional deletion of Nfat5 (induced by Mx-Cre and Vav-Cre), as well as macrophages differentiated from NFAT5-deficient mice (Figs. 2, 3, Supplemental Fig. 1E).

Altogether, our results showed that in addition to regulating gene expression in macrophages activated by different TLR ligands (33), NFAT5 also modulated their response to the polarizing cytokines IFN-γ and IL-4. Our findings showing that NFAT5 could promote the expression of various IL-4–responsive genes in primary macrophages differed from those in a recent report using the cell line RAW264.7, in which NFAT5 repressed the expression of arginase 1, IL-10, and MRC1/CD206 induced by IL-4 (41). As experiments by Choi et al. only achieved partial downregulation of NFAT5 with small interfering RNA in the cell line RAW264.7, it is possible that they might have missed relevant differences that we have identified in this study using primary macrophages with a complete deletion of NFAT5. Altogether, this block of results indicates that, rather than promoting macrophage gene expression in a fixed proinflammatory direction, NFAT5 could modulate different outcomes depending on the stimulatory environment.

Our analysis of diverse genes characteristic of IFN-γ or IL-4–stimulated macrophages suggested that NFAT5-deficient cells could exhibit functional defects or biases in response to these cytokines. IFN-γ and IL-4 influence the intrinsic antimicrobial capacity of macrophages, for instance, IFN-γ enhances their bactericidal activity to allow a tighter control of pathogen load upon infection (42). We compared the phagocytosis and bacteria-killing capacity of wild-type and NFAT5-deficient macrophages primed with IFN-γ or IL-4. Macrophages pretreated with IFN-γ or IL-4 were comparably capable of phagocytosing live E. coli, and lack of NFAT5 only caused them a mild decrease in their phagocytic capacity (Fig. 4A). By contrast, NFAT5-deficient macrophages exhibited a marked delay in their capacity to kill intracellular bacteria, both under IFN-γ and IL-4 conditioning (Fig. 4B).

Macrophages conditioned by IFN-γ or IL-4 have the ability to promote T lymphocyte polarization toward Th1 or Th2 respectively, which in turn reinforces the feedback between macrophages and CD4 and CD8 T cells. NFAT5 enhanced the expression of the pro-Th1 cytokine IL-12 in IFN-γ–polarized macrophages stimulated with LPS, but also induced Fizz-1 and arginase 1 in response to IL-4 (Figs. 2, 3, Supplemental Fig. 1E). Although Fizz-1 and arginase 1 are induced by IL-4 and are characteristic markers of M2 macrophages, they have a pro-Th1 effect as they attenuate Th2 polarization in favor of Th1 (43, 44). These results led us to ask about the ability of NFAT5-deficient macrophages to influence T cell polarization. Activation of CD4 T lymphocytes was elicited by stimulation with a soluble anti-CD3 agonistic Ab in the presence of macrophages preconditioned with IFN-γ or IL-4 that were subsequently treated with LPS. We observed that LPS enhanced the capacity of IFN-γ– or IL-4–polarized macrophages to respectively skew Th1 or Th2 responses (Fig. 5). Despite some differences in Th polarization markers induced in CD4 cells by macrophages derived from the two types of NFAT5-conditional knockout mice, we could observe that NFAT5-deficient macrophages were less capable of promoting IL-2 and IFN-γ expression in T cells, but were better inducers of IL-4 (Fig. 5). This result was consistent with the positive role of NFAT5 in inducing the pro-Th1 mediators IL-12, arginase 1 and Fizz-1 (Figs. 2, 3, Supplemental Fig. 1E), and suggested that although NFAT5 could regulate diverse macrophage genes in response to IL-4 or IFN-γ conditioning, its net effect in macrophage-mediated T cell stimulation was to facilitate Th1 polarization at the expense of Th2. Therefore, NFAT5 function in macrophages could contribute to promoting type 1 proinflammatory responses and limit type 2 immune responses in different contexts.

Collectively, our diagnostic experiments with IFN-γ and IL-4 revealed that NFAT5 could regulate macrophage responses to pro- and anti-inflammatory polarizing stimuli, which could influence the specialization of T lymphocytes. These findings led us to explore how NFAT5 would influence macrophage function in more complex settings where macrophage activity in one or other direction may affect the outcome of the overall immune response. In this regard, the interaction between macrophages and cancer cells in the tumor microenvironment is shaped by a balance between pro- and anti-inflammatory inputs that ultimately determines tumor progression. Tumors can evade immune attack by instructing local immune cells to become tumor tolerant. Although tumors can harbor M1 inflammatory macrophages with antitumor capacity, the overall balance is shifted toward a dominant M2-like, tolerance-promoting environment that attenuates proinflammatory responses and favor tumor progression (7, 8).

We first assessed the response of peritoneal macrophages to a syngeneic tumor in vivo, using the model of the ovarian epithelial carcinoma cell line ID8-Luc (C57BL/6 background). These cells induce tumors in the peritoneal cavity that resemble those seen in advanced stages of ovarian carcinoma (40), and have been used for the analysis of tumor-infiltrating macrophages (45). Initial setting-up experiments showed that peritoneal inoculation of ID8-Luc cells elicited the accumulation of different TAM subsets, as well as an increase in the local proportion of effector-memory (CD62L^neg, CD44^hi) CD4 and CD8 T cells (Supplemental Fig. 2A, 2B). In this experimental setting, peritoneal CD11b⁺ macrophages of ID8-Luc–bearing mice expressed diverse genes indicative of both M1 and M2 activation profiles (Supplemental Fig. 2C). We also observed that ID8-Luc tumor burden progressed slowly and in some mice decreased gradually over time (Supplemental Fig. 2D). Next, we compared the tumor burden and macrophage activation parameters between myeloid-specific NFAT5-deficient (LysM-Cre Nfat5^fl/fl) and littermate control mice at days 7 and 35 after inoculating ID8-Luc cells. We chose day 7 as it corresponded to an early peak of macrophage activation soon after tumor inoculation (Supplemental Fig. 2C), and 5 wk as it could be informative of a longer-term interaction between the tumor and the immune response. We found that tumor burden at day 7 was moderately reduced in myeloid-specific NFAT5-deficient mice (Fig. 6A), whose peritoneal macrophages also showed an increased proportion of TAM-A cells (Ly6C^hi MHC class II^neg) and a mild decrease in IL-12β mRNA expression, although they did not present significant defects in the expression of other pro- and anti-inflammatory markers (Supplemental Fig. 3A, 3B). We also did not find differences in the proportions of naive and effector CD4 and CD8 peritoneal T cells between both mouse genotypes (data not shown). At later time points from 2 to 5 wk, we did not detect differences between both mouse genotypes in tumor progression rates (Fig. 6A). Also, comparison of the percentages of tumor-infiltrating macrophage subsets and gene expression patterns at 5 wk did not show differences between wild-type and LysM-Cre Nfat5^fl/fl mice, although we observed that macrophages at 5 wk differed from 1 wk in expressing lower levels of arginase 1 and CCL2, as well as having a lower TAM-A percentage (Supplemental Fig. 3A, 3B). We then tested a different syngeneic (C57BL/6) tumor model of LLC cells. LLC cells injected s.c. gave rise to solid tumors whose growth progression and size were comparable in wild-type and myeloid-specific NFAT5-deficient (LysM-Cre Nfat5^fl/fl) mice throughout the 14 d of examination (Fig. 6B) and showed no obvious differences between both mouse genotypes in macrophage polarization markers and TAM subsets (Supplemental Fig. 3C, 3D). Altogether, results with these tumor models suggested that lack of NFAT5 in endogenous myeloid cells in vivo could cause a modest impact in the dynamics of macrophage activation in one of the tumor types tested, but they also suggested that its depletion from myeloid cells did not affect overall tumor progression. Interpreting the precise contribution of NFAT5 to the function of endogenous TAMs in these models was difficult seeing that infiltrating macrophages in both types of tumors exhibited a considerable dispersion in subset distribution and magnitude of gene expression between individual mice. Also, LysM-Cre conditional knockout mice would not only lack NFAT5 in different macrophage subsets but also in other myeloid cells such as neutrophils and dendritic cells, whose pro- and anti-inflammatory activity at different stages of tumor development may contribute with variable outcomes to tumor progression. In this regard, it has been shown that macrophages at different intratumoral maturation stages can differ considerably in expression patterns of pro- and anti-inflammatory genes (12).

In view of these results we decided to use different approaches, consisting of the adoptive transfer of homogeneous populations of in vitro–differentiated BMDM into tumor-bearing mice, or coinoculation of BMDM with tumor cells (45, 46). By challenging the tumors with a relatively large number of exogenous, homogenously differentiated macrophages we expected to circumvent the potential variability contributed by the deletion of NFAT5 in different myeloid cell lineages in LysM-Cre conditional knockout mice, and the heterogeneity in the activity of these populations at different moments of tumor progression. Our results showed that NFAT5-deficient BMDM injected in the peritoneum of mice with ID8-Luc tumors were less effective than wild-type BMDM at reducing tumor burden (Fig. 7A) and also reduced the accumulation of peritoneal CD8 T effector cells (Fig. 7B). We also found that NFAT5-deficient BMDM were less effective than wild-type ones at restraining the growth of s.c. LLC carcinomas when tumor cells were coinjected simultaneously with BMDM (Fig. 7C). These results suggested that NFAT5 could confer antitumor advantage when a population of homogeneous macrophages was adoptively transferred in a sufficiently high proportion relative to tumor cells. By contrast, in experiments where tumor cells were inoculated alone in wild-type or myeloid-specific LysM-Cre Nfat5^fl/fl mice, the potential capacity of NFAT5 to modulate antitumor function in endogenous macrophages was possibly masked by the heterogeneity of myeloid cell subsets, whose contribution to tumor expansion and sensitivity to lacking NFAT5 could be heterogeneous.

Our results also suggested that tumor cells might be able to rapidly instruct recruited macrophages to suppress antitumor functions so that tumor cells could persist and expand. This possibility led us to analyze the response of NFAT5-deficient macrophages to tumor cells in direct coculture experiments. We found that ID8-luc and LLC cells did not induce inflammatory markers iNOS and IL-12β in unpolarized wild-type macrophages (data not shown), and so we asked whether tumor cells could influence the inflammatory phenotype of macrophages previously polarized as M1. These experiments showed that LLC cells downregulated iNOS and IL-12β and induced the expression of arginase 1 in BMDM previously activated as M1 with IFN-γ plus LPS (Fig. 7D). We also observed that NFAT5-deficient macrophages, which already expressed less iNOS and IL-12β than wild-type ones upon IFN-γ and LPS stimulation, suffered a more pronounced downregulation of these genes and enhanced arginase 1 induction by LLC cells (Fig. 7D). These results indicated that lack of NFAT5 exacerbated the tumor-induced switch of macrophage toward an anti-inflammatory phenotype, and were consistent with the gene expression profile regulated by NFAT5 in polarized macrophages, as well as with our finding that NFAT5-deficient BMDM were poorer stimulators of Th1 responses and effector CD8 T cells (Figs. 5, 7B).

Our detailed analysis of the role of NFAT5 in classical and alternative macrophage polarization indicated its prevalent proinflammatory role in different macrophage functions that range from the direct control of pathogen load and T cell polarization to anti-tumor activity in vivo. The overall effect of NFAT5 in facilitating proinflammatory outcomes was paralleled by its high expression levels in macrophages stimulated with GM-CSF or IFN-γ plus LPS, in comparison with its low expression in alternative macrophages induced with M-CSF- or IL-4.

Whereas the contribution of NFAT5 to macrophage functions associated with a proinflammatory profile was consistent with its capacity to enhance the expression of genes characteristic of classically or M1-polarized macrophages, we also identified NFAT5 as a regulator of a group of markers of alternatively polarized macrophages. To a different extent, lack of NFAT5 impaired the expression of markers such as Arg1, Cd163, Chi3l3 (YM1), or Retnla (Fizz-1). Intriguingly, arginase 1 and Fizz-1, which are induced in macrophages by Th2 cytokines, have been shown to function in a negative feedback loop as attenuators of Th2 cells to favor Th1 inflammatory responses (43, 44). Therefore, it can be proposed that the contribution of NFAT5 to the expression of particular alternative macrophage markers also aligns with its capacity to enhance pro-Th1 mediators like IL-12 in M1 macrophages, so that altogether NFAT5 would increase the effectiveness of macrophages to activate Th1 cells when switching from M2 to M1 functions in response to inflammatory signals. Nonetheless, although this interpretation agrees with our characterization of NFAT5 as a proinflammatory factor, its capacity to regulate diverse macrophage features under both M1 and M2 inducers suggests that it could tune different modes of response in macrophages placed in complex stimulatory environments, without necessarily always driving them toward a fixed proinflammatory outcome.

The gene expression profile regulated by NFAT5 in polarized macrophages suggested that it could enhance the capacity of macrophages to promote type 1 versus type 2 immune responses (47, 48). Direct coculture of CD4 T lymphocytes with polarized NFAT5-deficient macrophages showed that they were weaker inducers of Th1 and stronger inducers of Th2 responses than wild-type ones, and NFAT5-deficient macrophages inoculated at the tumor site in mice bearing ovarian carcinoma cell tumors were less effective at inhibiting tumor growth and supporting the local accumulation of effector CD8 T cells, altogether suggesting an impaired capacity to elicit type 1 immune responses. In this regard, CD8 T lymphocytes are central players in antitumor responses induced spontaneously or upon therapy (49–51), and studies in ovarian carcinoma patients have shown that the presence of dense infiltrates of activated CD8 T cells associates with a good prognosis to antitumor treatment (52, 53).

We have used in vivo models of tumor progression as one readout to test how NFAT5 could contribute to macrophage function in a complex in vivo microenvironment, where the activity spectrum of local macrophages would encompass both pro- and anti-inflammatory functions (10, 45). Although we could detect differences in some responses of tumor-infiltrating macrophages between wild-type and LysM-Cre Nfat5^fl/fl mice, our overall results suggested that NFAT5 played a limited role in how endogenous macrophages influenced tumor expansion. In contrast, we found that adoptively transferred NFAT5-deficient macrophages had impaired antitumor capacity, which could be attributed in part to their reduced ability to promote a proinflammatory environment and engage CD8 effector T cells. This interpretation is consistent with our finding that NFAT5 enhanced the capacity of macrophages to stimulate IFN-γ and IL-2 production in CD4 T lymphocytes and promoted macrophage expression of iNOS and IL-12, all of which can activate antitumor CD8 cytolytic responses (11, 51). These results, and our coculture experiments showing a greater resistance of NFAT5 wild-type macrophages to the tumor-induced M1 to M2 switch, suggested that NFAT5 could facilitate macrophage-mediated mechanisms that ultimately support immunosurveillance against tumors. Regarding the different results obtained with adoptively transferred and endogenous macrophages, these could be discussed as several mutually nonexclusive interpretations. Myeloid-specific LysM-Cre Nfat5^fl/fl mice would delete NFAT5 not only in different macrophage subsets, but also in other myeloid cells such as dendritic cells, neutrophils, and eosinophils, all populations that could exhibit heterogeneous NFAT5-dependent effects with regards to tumor expansion. By contrast, in the adoptive transfer experiments we inoculated a large number of a homogeneous type of macrophage, likely tilting the equilibrium toward an immune response dominated by the transferred BMDM and not by endogenous myeloid cells. This dominant role of adoptively transferred macrophages has been reported in other systems such as the correction of lung alveolar proteinosis by adoptive transfer of GM-CSF receptor-competent macrophages in Csf2rb-deficient mice (54). Apart from these points, when interpreting the role of NFAT5 in how macrophages influence tumors, we should also consider that persistent inflammation upon tissue damage can drive tumor promotion (55–57) through cytokines like IL-6 and TNF-α (58, 59). In light of these considerations, it is tempting to wonder whether NFAT5 might have some influence on the capacity of macrophages to promote tumor initiation in an environment of sustained inflammation. Experimental testing of this possibility would require extensive work in different models of inflammation-induced tumor progression.

The capacity of NFAT5 to regulate different gene sets in proinflammatory or alternative macrophages recalls dual roles described for the Notch/recombination signal binding protein for Ig κ J region (RBP-J) and inhibitor of κB kinase β (IKKβ) pathways. On the one hand, the Notch pathway is a major contributor to proinflammatory macrophage responses in part through its activation of IRF8 (26, 27), but on the other hand it can also regulate specific alternative polarization markers (28, 29). Regarding IKKβ, this kinase can play a dominant role in the induction of numerous proinflammatory genes through activation of p65/NF-κB, but at the same time it can also limit inflammatory responses by inhibiting STAT1, and macrophages lacking IKKβ indeed become refractory to the immunosuppressive pressure of tumors and exhibit exacerbated inflammatory responses to bacterial products (45, 60). In previous work we had found that NFAT5 acted downstream of Notch1 to promote thymocyte maturation (61), and we had also observed that long-term activation of TLRs induced the accumulation of NFAT5 mRNA and protein in macrophages in a p65/NF-κB– and IKKβ-dependent manner (33). These observations suggest the possibility that macrophages might use NFAT5 in concert with Notch and IKKβ-regulated pathways to tune their pro- and anti-inflammatory capabilities under different polarizing stimuli. It is intriguing that macrophages experiencing anti-inflammatory or M2-promoting conditions do not entirely shut down proinflammatory pathways, but maintain active a set of transcription factors and signaling pathways such as NFAT5, Notch/RBP-J and IKKβ capable of mobilizing rapid inflammatory responses. As macrophages exhibit a considerable degree of plasticity in reacting to a broad diversity of potential inputs from their microenvironment, having dually responsive transcription regulators could allow them to stay poised and maintain a state of readiness to rapidly react to danger signals and disruptors of tissue homeostasis.

We are grateful to Dr. Marc Schmidt-Supprian and Dr. Thomas Graf for providing the Mx1-Cre and the Vav1-Cre transgenic mouse models, respectively; Dr. Ignacio Melero for providing the LLC cells; and Dr. Katherine F. Roby, Dr. José R. Conejo-Garcia, Dr. Laurence Zitvogel, and Dr. Paula Roberti for making available the ovarian carcinoma ID8-Luc cell line, as well as for providing valuable advice with culture. We thank Dr. Miguel López-Botet, Dr. Ángel L. Corbí, and Hector Huerga for insightful comments with this work; Dr. Jordi Pou for advice with polarization of human macrophages; and Dr. Oscar Fornas and team for guidance and expertise with flow cytometry assays. We thank María García Belando for technical support with mouse genotyping and members of our group for stimulating and helpful discussions.

The authors have no financial conflicts of interest.