BALB/c and C57BL/6 mice (female, 6–8 weeks old) were purchased from Jackson Laboratories (Bar Harbor, Maine, USA). A TFEB^flox mouse line was generated using the ES cells from EUCOMM depository in Europe, which are the same cells used for generation of liver-specific TFEB knockout mice.24 LysM-Cre mice were purchased from Jackson Laboratories and crossed with TFEB^flox mice to generate MΦ-specific TFEB knockout (MΦ-TFEB^-/-) mice. MΦ-specific TFEB transgenic (MΦ-TFEBtg) mice were generated and characterized previously.25 Female FVB/N mice were bred with male heterozygous C3(1)/SV40Tag mice (a gift from Dr Jeffrey Green, National Cancer Institute). All mice were housed in the University of South Carolina or Washington University in St. Louis Animal Research Facilities.

The 4T1 mouse mammary tumor cell line was obtained from the American Type Culture Collection (ATCC). EO771 cells were maintained in culture as described previously.23 Peritoneal MΦs (pMΦs) were isolated from thioglycollate-elicited mice. Bone marrow cells from age-matched female wild-type (WT) C57Bl/6, MΦ-TFEBtg and MΦ-TFEB^-/- mice were used to generate bone marrow–derived MΦs (BMDMs). Bone marrow cells were seeded in 6-well or 12-well plates for 4 hours, and non-adherent cells were removed, then cultured in complete medium containing 30% L929 cell-conditioned medium as a source of M-CSF for 7 days. The cells were cultured in high glucose Dulbecco’s modified Eagle medium (DMEM; Invitrogen Life Technologies) with 10% fetal bovine serum (FBS) (Invitrogen) and penicillin/streptomycin at 37°C in a humidified 5% CO₂ atmosphere. MΦs were polarized according to established protocols.26

Tumor conditioned medium (TCM) was prepared by culturing 4T1 or EO771 cells following the procedures described previously.23 Briefly, tumor cells were seeded at 5×10⁶ cells per 75 cm2 dish and cultured to 90% confluence. Then, the medium was removed, and the cells were rinsed with serum-free DMEM twice. The medium was then replaced with fresh serum-free DMEM, and the supernatant was collected after 24 hours. The collected TCM was centrifuged to remove cells and filtered through a 0.45 µm filter (Corning).

BMDMs from MΦ-TFEBtg and WT mice were treated with either serum-free DMEM (control), or serum-free DMEM supplemented with 20% concentrated EO771 TCM (ECM) for 24 hours. Cells were then rinsed twice with phosphate-buffered saline (PBS) and lysed using 700 µL Qiazol. RNA was extracted using the miRNeasy kit (Qiagen). Samples were then further cleaned prior to analysis using the RNeasy MinElute Cleanup Kit (Qiagen) protocol. An Agilent 2100 Bioanalyzer was used to determine both the quality and the quantity of the RNA samples. All samples had an RNA integrity number of >8. RNA amplification and labeling was performed using the Agilent Low Input Quick Amp labeling kit according to the manufacturer’s instructions. RNA was again purified using the Qiagen RNeasy mini kit. Dye incorporation and cRNA yield were also assessed. After labeling, samples were hybridized to SurePrint G3 Mouse GE 8×60K Microarrays (Cat. # G4858A-028005) using a gene expression hybridization kit (Agilent) according to the manufacturer’s instructions. A high-resolution Agilent DNA microarray scanner (Cat. # G2565CA) was used to obtain the raw data and images were saved in TIFF format for subsequent microarray analysis. Data were extracted from these images and background corrected data were uploaded into GeneSpring GX software for analysis. After quantile normalization of the data, differentially expressed genes were identified by using the moderated t-test with Benjamini-Hochberg multiple testing correction. Cut-off values of 0.05 and 2.0 were used for p-value and fold-change, respectively. Unsupervised hierarchical cluster analysis was performed with GeneSpring GX. Gene ontology (GO) analysis was performed with GeneSpring. Ingenuity Pathway Analysis (Qiagen) was performed on these data in order to further investigate the interconnection between differentially expressed genes and their biology. A heatmap was generated using normalized intensities and colored to represent gene expression variations for each individual gene.

For the tumor growth experiments, female BALB/c or C57BL/6 mice 6–8 weeks old, 4T1 or EO771 cells (2×10⁵) in 20 µL PBS were injected subcutaneously into the mammary fat pads immediately caudal to the fourth pair of nipples. Tumor volume was determined by caliper measurement and using the equation volume=(length×width²)/2.

Tumors were weighed, cut into small fragments (<3 mm) and digested in 5 mL of dissociation solution (RPMI 1640 medium supplemented with 10% FBS, Collagenase type I (200 U/mL), and DNase I (100 µg/mL)) for 60 min at 37°C. Erythrocytes were lysed with red blood cell lysing buffer (Sigma, St. Louis, Missouri, USA). Cell suspensions were passed through 70 µm cell strainers, then washed and resuspended in staining buffer. TAMs (F4/80+) were isolated from tumor suspensions using EasySep PE or FITC Positive Selection Kits (Stem Cell Technologies).

Cell populations in the tumors were analyzed using flow cytometry as previously described.23 Briefly, cells were stained with anti-CD3 FITC, anti-CD4 APC or anti-CD8 APC, and anti-CD25 PE (Biolegend) in PBS containing 2% FBS for 30 min at 4°C. Samples were washed twice with staining buffer and analyzed with flow cytometry using a BD FACS Aria II flow cytometer and CXP software V.2.2. Data were collected for 20,000 live events per sample.

To identify TFEB binding sites in the PPARγ promoter, we searched for the E-box motifs on peroxisomeproliferator-activated receptor γ (PPARγ) promoter or intron regions and found four E-boxes in the upstream region of mouse TFEB transcription start site. The promoter region containing TFEB binding sites were cloned into pGL3 basic vector to generate three constructs: PPARγ-1158 (−797 to +361) contains all 4 E-boxes; PPARγ-891 (−530 to +361) contains E boxes 2–4 which are close to each other; and PPARγ-464 (−103 to +361) does not contain any E-boxes. RAW264.7 cells were seeded in triplicate in 12-well plates and allowed to settle for 12 hours, then transfected with one of the above pGL3-PPARγ promoter plasmids or empty pGL3 plasmids and pRL-CMV (internal control) as well as either lentiviral TFEB (PWPI-TFEB) or control lentiviral vector (PWPI). The luciferase of both firefly and Renilla signals were measured using Dual Luciferase Reporter assay kits (Promega, Madison, Wisconsin, USA). Luciferase activity was normalized to Renilla activity to control for transfection efficiency. The level of luciferase activity of the empty vector and in the absence of TFEB (PWPI) was defined as ‘1’. Fold activation was estimated according to this level of activity.

Total RNA was isolated and purified using Qiagen RNeasy Kits (Qiagen). RNA (2 µg) was then reverse-transcribed using iScript cDNA Synthesis Kit (Bio-Rad). Quantitative real-time PCR (qPCR) was conducted on a CFX96 system (Bio-Rad) using iQ SYBR Green Supermix (Bio-Rad). All primers used for qPCR analysis were synthesized by Integrated DNA Technologies. All assays were conducted following the manufacturer’s instructions. The relative amount of target mRNA was determined using the comparative threshold (Ct) method by normalizing target mRNA Ct values to those of 18S RNA. PCR thermal cycling conditions were 3 min at 95°C, and 40 cycles of 15 s at 95°C and 58 s at 60°C. Samples were run in triplicate. The primer sequences are listed in online supplementary table S1.

Whole cell lysate was prepared using RIPA buffer (Pierce) supplemented with protease inhibitor cocktail and phosphatase inhibitor cocktail (Sigma). The protein concentrations were determined using the BCA protein assay kit (Pierce, Rockford, Illinois, USA). Samples were diluted in 2×Laemmli buffer (Bio-Rad) and boiled for 10 min. Proteins (20 µg) were separated in 10% SDS-polyacrylamide gel electrophoresis precast gels (Bio-Rad) and transferred onto nitrocellulose membranes (Bio-Rad). Non-specific binding sites on the membranes were blocked with 5% non-fat milk in phosphate buffered saline with tween 20 (PBST). Membranes were first probed with TFEB (1:2000; Bethyl Laboratories), PPARγ (1:1000), NLRP3, p-p65, p65, Lamp1, Hif1α, MIF, cytosolic phospholipases A2 (cPLA2), inducible nitric oxide synthase (iNOS), arginase 1 (Arg1), or β-actin (1:1000; Sigma) antibodies, followed by goat anti-rabbit or anti-mouse secondary antibody conjugated with horseradish peroxidase (Millipore). Protein detection was conducted using Pierce ECL Substrate (Pierce).

The breast cancer patient survival data were obtained from The Cancer Genome Atlas (TCGA) database and Kaplan-Meier plotter database (www.kmplot.com).27 Based on the best expression cut-off value (FPKM) of each gene, patients were classified into two groups, association between survival rate and gene expression was examined, or the HR was calculated. Survival curves were estimated with the Kaplan-Meier method and compared using the log-rank test.

For breast tumor tissues from patients, deidentified formalin-fixed paraffin embedded tissues were collected from mastectomy surgery with ethical approval by Nanjing Drum Tower Hospital in 2015. Sections were cut (4 µm thick), transferred to a warm water bath, and placed on a glass slide. The following immunofluorescence staining test was examined by the corresponding anti-human antibodies: TFEB (1:300, Invitrogen) and CD68 (1:300, Abcam).

Data were presented as mean±SD or SEM as indicated. Statistical significance was calculated using the Student’s t-test (two group comparison) or one-way analysis of variance (ANOVA) followed by post hoc Dunnett’s test (multigroup comparison) using the GraphPad Prism V.6.0. For the TCGA database analysis, p values were calculated for the comparison of the 5-year survival rates between the two groups. For the Kaplan-Meier plotter database, univariate Cox proportional hazard modeling of the genes was used to determine their independent impact on patients’ survival, and to estimate the corresponding HR, setting low expression as the reference group; HRs and 95% CI, as well as the p values, were reported. P<0.05 was considered significant.

TAMs are often considered to adopt an M2 phenotype, which is typically induced by IL-4 or IL-13. However, the TME is a complex milieu; in addition to IL-4 and IL-13, many other factors, including hypoxia, apoptotic cancer cells and their debris, growth factors, and proinflammatory or anti-inflammatory cytokines secreted from cancer cells and stromal cells, act on TAMs. Therefore, the phenotype of TAMs is much more complex than that of typical M2 MΦs. Breast cancer EO771 cell conditioned medium (ECM) was used to treat BMDMs to mimic TAMs, and a microarray was performed to compare the gene expression profiles between those TAM-mimic cells and control MΦs. The results are summarized in a heatmap; we also plot a heatmap to compare the expression of the same set of genes in control MΦs, M1 MΦs (induced by LPS+IFNγ), and M2 MΦs (induced by IL-4) using the data we obtained from a previous microarray,28 in order to visualize the gene expression differences between TAM-mimic cells and M1 or M2 MΦs (figure 1A). First, we confirmed that like in IL-4 induced M2 MΦs, some typical M2 markers, including Arg1, IL-10, and TGFβ1, are increased in ECM-treated MΦs. However, unlike in M2 MΦs, the expression of many M1 markers, such as IL-1β, IL-6, tumor necrosis factor α (TNFα), iNOS and NLRP3, also are significantly increased in ECM-treated MΦs. Additionally, more resembling M1 than M2 MΦs, ECM-treated MΦs have elevated expression of HIF-1α, macrophage migration inhibitory factor (MIF), cPLA2, and cyclo-oxygenase-2 (COX2). COX2 catalyzes the production of prostaglandin E2 (PGE2), a key tumor-promoting and immune-suppressive metabolite in the TME. Also, unlike M2 MΦs but more like M1 MΦs, ECM-treated MΦs have reduced expression of TFEB and PPARγ. PPARγ controls several inflammatory response and cellular metabolism pathways. Using western blot analysis, we confirmed that TFEB and PPARγ are increased in M2 MΦs, but reduced in ECM-treated MΦs; while NLRP3 is reduced in M2 MΦs, but increased in ECM-treated MΦs (figure 1B).

We previously showed that TFEB overexpression in MΦs attenuated orthotopic breast tumor growth, while injection of MΦs with lentiviral knockdown of TFEB accelerated breast tumor growth.23 To further examine the function of MΦ TFEB in breast tumors, we aimed to test if TFEB ablation in MΦs promotes breast tumor growth via conferring TAMs with a more tumor-promoting phenotype. We therefore generated MΦ-specific TFEB knockout (MΦ-TFEB^-/-) mice (figure 1C), whose BMDMs express TFEB at only ~25% level of WT control (figure 1D,E). We generated bone marrow chimeric mice by transplanting bone marrow cells from MΦ-TFEB^-/- mice or Cre-littermates to lethally irradiated WT C57Bl/6 mice using our standard protocol.23 We implanted EO771 cells into mammary fat pads of the mice 4 weeks after the bone marrow transplantation. Tumor growth was monitored for 35 days, showing TFEB ablation in MΦs promoted tumor growth (figure 1F). At sacrifice, the weight of tumors from mice transplanted with bone marrow cells from MΦ-TFEB^-/- mice was significantly increased compared with that of control mice (figure 1G). Single cell suspensions were obtained from tumors, and flow cytometry analysis showed that tumors in MΦ-TFEB^-/- bone marrow recipient mice contained similar percentages of total F4/80⁺ MΦs or CD3⁺ T cells, but more CD206⁺ MΦs and CD3⁺/CD4⁺ T lymphocytes (figure 1H and online supplementary figure S1). MΦs were isolated from the tumors, and gene expression was analyzed using qPCR. The result shows that TAMs from the tumors in MΦ-TFEB^-/- bone marrow recipient mice had lower expression levels of TFEB (~50%) and PPARγ (~70%), but higher levels of Arg1 (~3.7-fold), IL-10 (~2.2-fold), and COX2 (~2.5-fold) (figure 1I), suggesting TFEB ablation exaggerated the tumor-promoting phenotype of TAMs in breast tumors.

With the eventual goal of developing strategies to increase TFEB expression or activate TFEB in TAMs to halt breast tumor development, we next examined how TFEB overexpression may ameliorate the aberrant gene expression profile in TAMs. We have generated MΦ-TFEBtg mice in our previous studies.21 23 BMDMs from WT or MΦ-TFEBtg mice were treated with ECM and a microarray analysis was performed. A heatmap shows that 148 genes were downregulated, and 140 genes were upregulated in ECM-treated TFEBtg MΦs compared with ECM-treated WT MΦs (online supplementary figure S2A). Most significantly affected pathways relevant to MΦ functioning in inflammation and cancer were determined by Ingenuity IPA canonical pathway analyses. It was found that among the most significantly suppressed pathways in TFEBtg cells includes HIF-1α signaling, IL-10 signaling, IL-6 signaling, and VEGF signaling pathways (online supplementary figure S2B). The differences between WT and TFEBtg MΦs in the expression levels of the genes shown in figure 1A were also examined. It was shown that TFEB overexpression significantly ameliorates the tumor-promoting gene expression profile in TAM-like MΦs (figure 2A). The microarray results were confirmed by qPCR (figure 2B). These results suggest that TFEB overexpression may alter the TAM-like MΦ phenotype by (1) suppressing typical M2 MΦ markers such as Arg1, IL-10, MMP9, and VEGFA; these molecules are known to cause immunosuppression and tumor angiogenesis; (2) increasing the expression of genes involved in autophagy and lysosome biogenesis, such as LC3, Lamp1, P62, and ATG5; (3) enhancing PPARγ and inhibiting the NLRP3/IL-1β/IL-6 inflammatory axis; and (4) suppressing the PGE2 production axis HIF1α/MIF/cPLA2/COX2. We examined in our microarray data set if TFEB overexpression affected the expression of genes related antigen presenting function of MΦs when they are treated with TCM, such as β2-microglobulin, cathepsin E, and Tap1. We found no such genes were differentially expressed in WT or TFEBtg MΦs, suggesting that TFEB expression levels may not affect MΦs antigen presentation.

Western blot was performed to confirm some of the gene expression results using WT, TFEBtg and TFEB^-/- BMDMs. The results show that, in ECM-treated BMDMs, TFEB overexpression increased the protein levels of PPARγ and Lamp1 but decreased the phosphorylation of p65 and the protein levels of NLRP3, HIF-1α, MIF, cPLA2, and Arg1, while TFEB knockout had the opposite effects on these proteins (figure 2C). We further examined the impact of TFEB overexpression on the production of three direct MΦ effector molecules IL-1β, IL-6 and PGE2. We treated WT and TFEBtg BMDMs with ECM for 24 hours, then the cells were rinsed and cultured in fresh, serum-free DMEM for a further 16 hours before the medium was collected for ELISA measurement of IL-1β and IL-6, and colorimetric measurement of PGE2. The data show that TFEB overexpression significantly suppressed the elevated levels of IL-1β, IL-6 and PGE2 in ECM-treated BMDMs (figure 2D).

TFEB binds to a short consensus sequence CANNTG called the E-box on target gene promoters or intron regions to control gene transcription.29 It has been reported that TFEB transcriptionally regulates PPARγ expression in human adipocytes.30 To test if TFEB directly regulates PPARγ transcription in mouse MΦs, we performed a promoter assay. As shown in figure 3A, TFEB overexpression increased luciferase activity in the cells transfected with E-box containing plasmids PPARγ-1158 and PPARγ-891, suggesting that the first and at least one of 2–4 E-boxes are directly involved in TFEB binding.

To provide further evidence that PPARγ is directly involved in mediating the effects of TFEB on TAMs, the PPARγ antagonist GW9662 was used to inhibit its activity in BMDMs. WT and TFEBtg BMDMs were treated with ECM with or without 10 µM GW9662 for 24 hours (figure 3B,C and online supplementary figure S3). Both mRNA and protein levels of TFEB in BMDMs were not altered by PPARγ inhibition, and PPARγ inhibition did not change Beclin 1 protein levels in BMDMs. Interestingly, while PPARγ mRNA expression was not changed by GW9662, its protein levels were remarkably reduced by GW9662 in ECM-treated TFEBtg BMDMs to a level comparable with ECM-treated WT BMDMs (both almost undetectable). The mechanism by which GW9662 reduces PPARγ protein in BMDMs is unknown. Importantly, PPARγ inhibition increased p65 phosphorylation and the expression of NLRP3, IL-1β, HIF-1α, and cPLA2 in both ECM-treated WT and TFEBtg BMDMs. The differences between ECM-treated WT and TFEBtg BMDMs in p65 phosphorylation and expression of NLRP3 (mRNA and protein), IL-1β (mRNA), HIF-1α (protein), and cPLA2 (protein) were diminished, but the differences in HIF-1α and cPLA2 mRNA levels were not decreased. These results suggest that PPARγ partially mediates the suppressive effects of TFEB on NF-κB activation and NLRP3 and HIF-1α upregulation in ECM-treated BMDMs. Taken together, these data demonstrate that TFEB directly regulates PPARγ transcription and partially through this mechanism, TFEB modulates NLRP3/IL-1β and HIF1α/cPLA2 pathways in MΦs.

TFEB is a key regulator of autophagy and lysosome biogenesis,20 and autophagy and associated lysosomal activation are known to counteract NLRP3 inflammasome activity15 and mediate HIF-1α degradation.16 PI3K is required for autophagy and its inhibitor LY294002 can inhibit autophagic sequestration, a required first step of autophagy.31 Therefore, we aimed to examine if autophagy inhibition by LY294002 could negate the impact of TFEB overexpression on MΦs. We treated WT and TFEBtg BMDMs with ECM with or without LY294002 and measured the expression of some of the key molecules mentioned previously. LY294002 could effectively block autophagy by abolishing Beclin 1 protein, but it did not change the expression of TFEB and PPARγ, nor did it affect p65 phosphorylation in either ECM-treated WT or TFEBtg BMDMs (figure 4A–C and online supplementary figure S4). Importantly, autophagy inhibition significantly enhanced NLRP3, IL-1β, HIF-1α, and cPLA2 expression in both ECM-treated WT and TFEBtg BMDMs, and diminished the differences in HIF-1α and cPLA2 protein levels in those cells (figure 4A–C and online supplementary figure S4). These results suggest that autophagy partially mediates the beneficial effects of TFEB on TAMs.

Our previous study showed that trehalose, a natural disaccharide, activates MΦ TFEB and enhances MΦ autophagy and phagocytosis of apoptotic cells.21 We here first confirmed that similar to HPβCD, also a TFEB activator,32 and chloroquine, an autophagic flux inhibitor that activates TFEB,33 trehalose increased TFEB nuclear translocation and total cellular TFEB in ECM-treated MΦs (online supplementary figure S5A,B). To test whether trehalose can modulate TAM function in breast tumors through activating TFEB, we treated BMDMs with ECM with or without trehalose. qPCR showed that while trehalose did not affect the expression of TFEB in ECM-treated MΦs, activation of TFEB by trehalose replicated the gene expression profile observed in TFEBtg MΦs, including increased expression of PPARγ and Lamp1, and decreased expression of NLRP3, IL-1β, HIF-1α, cPLA2 and COX2 (figure 5A). The results were further confirmed by western blot analysis (figure 5B and online supplementary figure S5C). Furthermore, trehalose treatment reduced the secretion of IL-1β, IL-6, and PGE2 from ECM-treated MΦs (figure 5C). We then examined if the activation of TFEB by trehalose is dependent on autophagy. We treated BMDMs from MΦ-specific Atg5 knockout mice25 and control WT littermates with trehalose, and we found that (1) Atg5 depletion is very efficient in the MΦs from MΦ-specific Atg5 knockout mice, (2) trehalose significantly increased Atg5 expression in WT MΦs but not in Atg5-deficient MΦs, and (3) trehalose equally increased the expression of TFEB targeted gene PPARγ in WT and Atg5-deficient MΦs (online supplementary figure S5D). These results confirmed that trehalose activates TFEB in MΦs independent of autophagy since Atg5 is required for autophagy.

We next tested if trehalose treatment could inhibit breast tumor growth. We treated female Balb/c mice with trehalose for 25 days at 2 g/kg, three times per week, starting 1 day after 4T1 cell inoculation, and monitored the growth of the tumors. It is shown that trehalose treatment significantly delayed tumor growth (figure 5D). To verify if the benefit of trehalose in indeed due to the modulation of the immune microenvironment, we performed qPCR to measure the expression of Arg1 and IFNγ to reflect the abundance of M2-like TAMs and cytotoxic T cells, respectively, in the tumors. We found that the tumors from trehalose treated mice displayed significantly lower expression of Arg1 but higher expression of IFNγ (online supplementary figure S6A). The experiment was repeated using C57Bl/6 mice inoculated with EO771 cells; a sucrose control group also was included. The results show that trehalose, but not sucrose, significantly reduced breast tumor growth (figure 5E). In addition, we used the C3(1)/SV40Tag spontaneous breast tumor mouse model to confirm the effects of trehalose. Female C3(1)/SV40Tag mice were treated with trehalose (2 g/kg, three times a week) or vehicle from 6 to 17 weeks of age. The results demonstrated a trend in the reduction of mammary gland tumor size related to trehalose administration but not in the number of tumors (online supplementary figure S6B).

We used the TCGA database to analyze the association between the expression of TFEB or its regulated genes and breast cancer survival. We found 822 patients with low TFEB expression and 253 patients with high TFEB expression. The data indicate that 5-year survival is significantly higher in the patients with higher TFEB expression levels compared with those with lower TFEB expression levels (87% vs 80%, p=0.0036) (figure 6A). Similarly, higher expression levels of SOCS3 and PPARγ are correlated with a higher survival rate, while higher expression levels of HIF-1α are associated with a lower survival rate (figure 6A). The result was confirmed using another online analysis platform. TFEB, SOCS3, and PPARγ are shown to be positive prognostic factors, while HIF-1α is a negative prognostic factor (figure 6B). Although the expression of these genes is not only from TAMs, the data suggest that activating TFEB and thus increasing the expression of SOCS3 and PPARγ while reducing the expression of HIF-1α in breast tumors may be beneficial for patients with breast cancer. To further examine if suppression of TFEB expression in TAMs is true in human breast tumors, we stained TFEB using human breast tumor specimens. Breast tumors from five triple negative breast cancer patients were stained for CD68, TFEB and 4′,6-diamidino-2-phenylindole (DAPI). The data show that TFEB staining intensity is significantly lower in the MΦs in the tumor nest than that in the tumor periphery (online supplementary figure S7).