In established tumors, tumor-associated macrophages (TAM) orchestrate nonresolving cancer-related inflammation and produce mediators favoring tumor growth, metastasis, and angiogenesis. However, the factors conferring inflammatory and protumor properties on human macrophages remain largely unknown. Most solid tumors have high lactate content. We therefore analyzed the impact of lactate on human monocyte differentiation. We report that prolonged lactic acidosis induces the differentiation of monocytes into macrophages with a phenotype including protumor and inflammatory characteristics. These cells produce tumor growth factors, inflammatory cytokines, and chemokines as well as low amounts of IL10. These effects of lactate require its metabolism and are associated with hypoxia-inducible factor-1α stabilization. The expression of some lactate-induced genes is dependent on autocrine M-CSF consumption. Finally, TAMs with protumor and inflammatory characteristics (VEGFhigh CXCL8+ IL1β+) are found in solid ovarian tumors. These results show that tumor-derived lactate links the protumor features of TAMs with their inflammatory properties. Treatments that reduce tumor glycolysis or tumor-associated acidosis may help combat cancer.
Macrophages are involved in many physiologic processes, including the maintenance of tissue homeostasis, antimicrobial defense, and the control of immune responses (1). Macrophages detect diverse signals (e.g., microbes and cytokines) and sense changes in nutrient and oxygen availability. The phenotypes and functions of these cells continually adapt to tissue requirements (1–3), resulting in cellular diversity that complicates classification of macrophages (4). Classification of macrophages into M1 and M2 cells, corresponding to extremes of a continuum of phenotypes, is still widely used (4, 5). M1 are tumoricidal cells that produce inflammatory and Th1-promoting cytokines and low concentrations of IL10. In contrast, M2 and M2-like cells have immunoregulatory and trophic properties and are involved in maintaining tissue homeostasis and orchestrating wound healing (2). Such cells produce high concentrations of IL10 and mediators involved in the recruitment of regulatory, Th2, and naïve T cells, but produce only little Th1-promoting and inflammatory cytokines (6). Nevertheless, single-cell analysis has identified human macrophages with features of both M1 and M2 (7, 8), particularly in established tumors (9–13). Most of the known macrophage polarization factors tend to antagonize their own effects rather than acting in an additive or synergistic manner. The polarization factors capable of generating human macrophages with mixed M1/M2 profiles therefore remain to be identified.
Human and mouse macrophages differ in terms of differentiation and polarization factors and the expression of M1 and M2 markers (14–16). These differences make it impossible to extrapolate results obtained with mouse macrophages to human cells.
In most advanced solid tumors, tumor-associated macrophages (TAM) constitute the largest infiltrating leukocyte population (17). Extensive TAM infiltration has been correlated with a poor prognosis in various cancers (2, 6). TAMs affect almost all aspects of tumor biology, as they promote angiogenesis, tumor progression, invasion, and metastasis. This contribution to the various phases of tumor development highlights their functional diversity. Early in tumor development, macrophages have an antitumor M1 phenotype, due to the production of GM-CSF and IFNγ by activated leukocytes (6). Subsequent tumor growth then leads to the generation of signals perceived by monocytes/macrophages as indicating a need for tissue repair, resulting in the generation of TAMs with properties resembling those of M2 macrophages. Nonresolving chronic inflammation is a hallmark of carcinogenesis that contributes to the exhaustion and inefficiency of T cells. TAMs orchestrate tumor-associated inflammation (6, 18, 19), suggesting that they may display both M2-like and M1-like features. However, the signals present in the tumor environment and involved in the acquisition of a mixed M1/M2 profile by TAMs remain unclear.
Most cancer cells sustain high rates of proliferation via aerobic glycolysis (Warburg effect; ref. 20) and divert glucose-derived pyruvate away from the mitochondria toward lactate production. Lactate concentrations in normal tissues at rest are within the 1.8 to 2 mmol/L range, whereas lactate concentrations in solid tumors range from 6 to 15 mmol/L, reaching 40 mmol/L in extreme cases (21). The export of lactate via lactate–proton symporters is accompanied by protons (H+), the latter being responsible for the low extracellular pH in tumor tissues. Lactic acidosis is common in most solid tumors, and high concentrations of lactate are associated with aggressiveness and resistance to treatment (22). Lactate is not solely a by-product of glycolysis, but is also a fueling and signaling molecule, rescuing cancer cells from glucose deprivation (23). In an acidic extracellular milieu necessary for its cellular internalization, lactate may act as an immunoregulatory mediator. Lactate dampens cytotoxic T-cell and natural killer–cell functions, favors regulatory T cell (Treg) survival (24–27), and triggers the expression of M2 genes in murine macrophages (28–30).
The impact of chronic exposure to lactate on the differentiation of human monocytes remains to be determined. As TAMs are mostly derived from monocytes and lactic acidosis is a feature of tumors (22, 31), monocytes probably differentiate into TAMs in conditions of lactic acidosis. We report that lactate drives the differentiation of human monocytes into macrophages with an inflammatory and trophic phenotype similar to that of TAMs in ovarian tumors.
Materials and Methods
Monocyte isolation and cell generation
Peripheral blood mononuclear cells were obtained from healthy human volunteers (Blood Collection Center, Angers, France; agreement ANG-2017-01) by standard density-gradient centrifugation on Lymphocyte Separation medium (Eurobio). CD14+ monocytes were then isolated by positive magnetic cell sorting (Miltenyi Biotec); purity was routinely >98%. Monocytes (1 × 106 cells/mL) were cultured for 5 days in complete medium (CM) consisting in RPMI 1640 (Lonza) supplemented with 10% FCS (Eurobio), 2 mmol/L l-glutamine, 1 mmol/L sodium pyruvate, 0.1 mmol/L nonessential amino acids, 10 mmol/L Hepes, penicillin (100 U/mL), and streptomycin (100 μg/mL, all from Lonza) in the presence of M-CSF (50 ng/mL, ORF Genetics) to generate M-MΦ, or GM-CSF (50 ng/mL, R&D Systems) to generate GM-MΦ. Prototypic M2 cells (M+IL4-MΦ) were generated by culturing monocytes in CM containing M-CSF (50 ng/mL) for 5 days, with the addition of IL4 (50 ng/mL, R&D Systems) at day 3. Monocyte-derived dendritic cells (Mo-DC) were generated by culturing monocytes for 5 days with GM-CSF (50 ng/mL) and IL4 (50 ng/mL) in CM.
Lactic acidosis, acidosis, and lactosis
Monocytes (1 × 106 cells/mL) were differentiated under lactic acidosis by culture for 5 days in CM with 50 ng/mL GM-CSF and 10 mmol/L lactate (Sigma-Aldrich) with a resulting pH of 6.5. In some experiments, 2 to 40 mmol/L lactate was used. In others, monocytes were cultured for 5 days in CM with GM-CSF plus 10 mmol/L lactate in the presence of 3 μmol/L GW2580, 125 μmol/L α-cyano-4-hydroxycinnamate (CHC), 1 μmol/L acriflavine, 5 mmol/L oxamic acid (all from Sigma-Aldrich), 15 μmol/L GSK2837808A (Tocris), neutralizing anti–M-CSF (4 μg/mL) or isotype control mAbs (4 μg/mL, both from R&D Systems), or 1.25 μmol/L AZD3965 (Cayman Chemical). All inhibitors were used at the highest nontoxic concentration (Supplementary Table S1). Monocytes were differentiated under acidosis by culture for 5 days in CM containing GM-CSF plus 12.35 mmol/L HCl to reach a final pH of 6.5. Monocytes were differentiated under lactosis by culture for 5 days in CM containing GM-CSF plus 10 mmol/L sodium lactate (Sigma-Aldrich; pH = 7.3).
Cells were counted by flow cytometry using 123count eBeads (Thermo Fisher Scientific) after staining with 7-AAD or LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Molecular Probes) to discriminate dead cells from living cells at the end of the 5 to 7 days of culture. Macrophages (0.5 × 106 living cells/mL) were stimulated with 200 ng/mL LPS (from Escherichia coli serotype O111:B4; Sigma-Aldrich) in fresh CM (to avoid interference in cytokine quantification due to acidic pH). Cytokines were quantified in the 24-hour cell culture supernatants by fluorescent-bead–based multiplex immunoassay (R&D System), except IL12p70 (Diaclone), endothelin-1, and IL20 (both from R&D Systems) which were quantified using conventional ELISA; IL20 was quantified in 5-fold concentrated supernatants using Amicon Ultra centrifugal filters (Merck Millipore). Soluble CD115 (sCD115) and M-CSF were quantified by ELISA (R&D Systems) in the supernatants of monocytes cultured without or with 10 mmol/L lactate; lactate did not interfere with CD115 and M-CSF quantification (not shown).
Flow cytometry analysis
In order to discriminate dead cells from living cells, all samples were stained with LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Molecular Probes) before flow cytometry analysis using mAbs to CD1a, CD14, CD163, or CD115. Intracellular cytokine and pERK1/2 expression was evaluated on day-5 macrophages stimulated for 24 hours with LPS (in the presence of monensin for the last 6 hours; BD Bioscience) or on dissociated ovarian cancer cells stimulated or not for 6 hours with 50 ng/mL LPS (in the presence of monensin). Briefly, cells were fixed (4% PFA in PBS) and permeabilized (0.1% saponin in PBS containing 0.1% BSA) before incubation with mAbs to TNFα and IL6. The expression of IL1, CXCL8, and VEGF was also evaluated in activated ovarian cancer TAM. After mechanical and enzymatic dissociation, cells were stimulated, or not, for 6 hours with 50 ng/mL LPS in the presence of monensin. After membrane permeabilization, cells were incubated with anti-IL1β, anti-CXCL8, anti-VEGF, or anti-CD14 and related control mAbs (all from BioLegend). Data were acquired using a FACSCanto II flow cytometer (BD Biosciences) and analyzed with the FlowJo software (Tree Star). Results are expressed as relative of fluorescence intensity (RFI) defined as the ratio of specific fluorescence [mean fluorescence intensity (MFI) values with specific mAb] over nonspecific fluorescence (MFI values with isotype control mAbs) or as a percentage of labeled cells. Antibodies are listed in Supplementary Table S2.
Analysis of mRNA
Total RNA was extracted and reverse-transcribed as previously described (32), and the amounts of mRNA for the proteins indicated were evaluated by qPCR. Relative quantification was performed by the 2−ΔΔCT method, using RPS18, EF1A, TBP, RPL13A, and PPIA as references (33). This method of calculation is based on the conversion of linear Cq values onto a logarithmic scale, with PCR efficiency as an exponential function; the geometric mean value of mRNA amounts for selected housekeeping genes is used as a normalization factor, making it possible to eliminate intersample variation.
Lactate and glucose quantification
Glucose and lactate concentrations in cell culture supernatants were determined using enzymatic assays (Abcam) according to the manufacturer's instructions. Results are expressed in mmol/L or as a rate (mmol/L/24 h) of consumption or production (concentration values divided by the number of days of culture), with positive and negative values corresponding to consumption and production, respectively.
Analysis of lactate metabolism
In order to monitor extracellular lactate metabolism, monocytes were differentiated for 5 days with 50 ng/mL GM-CSF plus 10 mmol/L 13C-lactate sodium (Sigma-Aldrich), pH = 6.5. 13C-containing metabolites present in day-5 cell culture supernatants and cell extracts were analyzed by nuclear magnetic resonance (NMR) spectroscopy and liquid chromatography coupled to high-resolution mass spectrometry (LC-MS), respectively.
Freeze-dried supernatants were lyophilized and dissolved in 500 μL maleate-EDTA buffer (12.5 mmol/L EDTA, 250 mmol/L maleate, pH = 7) and 50 μL heavy water (D2O; Sigma-Aldrich). After centrifugation (7,500 × g, 3 min, 15°C), samples were transferred into a 5-mm tube (Z107373; Bruker Biospin) before analysis with an NMR spectrometer Advance 700 Mz (Bruker Biospin). NMR analyzes were performed at 298 K (25°C) without tube spinning, using a proton-decoupled carbon pulse program (zgig) with 90° pulses (10 μs, 50 W). Acquisition parameters: decoupling sequence waltz16, 0.9 seconds acquisition time, 65 k size of FID, and a relaxation delay (D1) of 1.2 seconds. Note that 12,500 scans were done, representing about 10-hour analysis per sample. 13C-quantitation was performed by peak integration with TopSpin 3.5 pl 7 using maleate (125 μmol per sample) as an internal standard.
Frozen cell pellets were resuspended in cold methanol (MeOH) and vortexed for 1 minute. Samples were extracted with MeOH:acetonitrile:H2O (2:2:1; v:v), and the mixture was then vortexed for 30 seconds and centrifuged (8,000 × g, 10 minutes, 4°C). The resulting supernatant was evaporated to dryness using a miVac duo concentrator (Genevac Ltd.). The dry extract was reconstituted with 200 μL of a solution of 2% aqueous MeOH (initial conditions of the chromatographic elution gradient) and subjected to a second centrifugation (8,000 × g, 10 minutes, 4°C) prior to mass spectrometry analysis. The analytical system was based on a reverse phase ultrahigh-performance liquid chromatography (Dionex UltiMate 3000; Thermo Scientific) coupled to a high-resolution mass spectrometer (LC-MS; Q Exactive; Thermo Scientific). LC-MS analysis was performed according to a recently validated method (33). Results are expressed as a percentage of redistribution of 13C into metabolites present in the day-5 cell culture supernatants (RMN) or isotopic enrichment in each molecule taken individually in day-5 cell lysates (LC-MS).
Measurements of oxygen consumption rate
The XF96 extracellular flux analyzer (Seahorse analyzer; Agilent Technologies) was used to determine the bioenergetic profile of macrophages. Day-5 cells were seeded (50 × 103 cells/well) in XF96 plates and allowed to recover for 24 hours. Cells were then incubated in bicarbonate-free DMEM (Sigma-Aldrich) supplemented with 11 mmol/L glucose, 2 mmol/L l-glutamine, and 1 mmol/L sodium pyruvate in a CO2-free incubator for 1 hour. The oxygen consumption rate (OCR) was recorded to assess the mitochondrial respiratory activity and glycolytic activity. After measurements under basal conditions, cells were treated sequentially with 2 μg/mL oligomycin and 3 μmol/L carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP; Sigma-Aldrich). Nonmitochondrial respiration (OCR value after addition of 1 μg/mL antimycin A) was subtracted from all OCR measurements. ATP-linked respiration was estimated from the difference between the basal and the oligomycin-inhibited respiration, proton-leak respiration is obtained by subtracting nonmitochondrial respiration from the oligomycin rate, and the maximal respiratory capacity was the rate in the presence of the uncoupler FCCP. Three independent replicates of each measurement were generated, and results were normalized according to cell concentrations.
In order to evaluate the impact of tumor cell–derived lactate on the phenotype of macrophages, monocytes were cocultured with the primary tumor cells CASC170, CKC0829, and CKT0151 isolated from patients with primary ovarian cancer peritoneal carcinosis (Biological Resource Center, University Hospital of Rennes, Rennes, France; agreement BB-0033-00056) and selected on their capacity to produce lactate (Supplementary Fig. S1); cells were negative for Mycoplasma. Monocytes were cocultured with the tumor cells using 12-well Transwell units with 0.4 μm diameter pore size polycarbonate filters (Corning). Monocytes (1 × 106 cells/mL) were seeded in the lower compartment containing 50 ng/mL GM-CSF, in the absence or presence of CHC, and 2.5 × 104 tumor cells were added in the upper part of the Transwell. After 5 days, macrophages were analyzed by flow cytometry and by qPCR. In some experiments, macrophages were washed and counted (0.5 × 106 living cells/mL in fresh CM) before stimulation with 200 ng/mL LPS for cytokine quantification.
Statistical analyses were performed with Prism 7 (GraphPad Software Inc.). Nonparametric Wilcoxon matched-pairs tests were used to compare means. For the comparison of antigen expression time courses, two-way ANOVA was performed, with Bonferroni post hoc correction for multiple comparisons. Values of P < 0.05 were considered statistically significant.
Lactic acidosis leads to generation of CD14high CD163high macrophages in humans
GM-CSF is overproduced at tumor sites and enables human monocytes to survive and differentiate into macrophages (15). We assessed the impact of lactic acidosis on monocyte differentiation by culturing human monocytes with GM-CSF, with or without lactate, and analyzing their phenotype on day 5. We used 10 mmol/L lactate for these experiments, as, consistent with the findings of a previous study (34), this was the highest nontoxic concentration for human monocytes/macrophages (Fig. 1A) and is the concentration found at tumor sites (27).
Under conditions of lactic acidosis (10 mmol/L lactate, pH = 6.5), monocytes differentiated into macrophages (GM+LA-MΦ), as demonstrated by the expression of the macrophage markers CD14 and CD163, but not of the Mo-DC marker CD1a (Fig. 1B and C). CD14 and CD163 were more strongly expressed on GM+LA-MΦ than on GM-MΦ, as reflected in both protein (Fig. 1C) and mRNA content (Fig. 1D). Expression of CD163 was upregulated in a time-dependent manner during differentiation (Fig. 1E). As previously reported (35), M-MΦ also had a CD14high CD163high phenotype (Fig. 1B–E).
The effects of lactate depend on its entry into cells
Lactic acidosis results from extracellular accumulation of lactate and protons, leading to a lowering of extracellular pH (27). We therefore evaluated whether the effects of lactate on monocytes/macrophages were due to an adaptation to acidosis or to lactosis. The differentiation of monocytes under conditions of acidosis (12.35 mmol/L HCl, pH = 6.5) or lactosis (10 mmol/L sodium lactate, pH = 7.3) had no effect on the expression of CD163 and CD14 on macrophages on day 5 (Fig. 2A, left plots).
Lactate import into cells is mediated by monocarboxylic acid transporters (MCT) and requires a low extracellular pH (28). As both lactosis and a low pH were required for the conversion of GM-MΦ into CD163high CD14high cells, we hypothesized that this process was dependent on lactate influx into cells. Consistent with this hypothesis, the MCT inhibitor CHC (used at 125 μmol/L; ref. 28) and the specific MCT1 inhibitor AZD3965 (used at 1.25 μmol/L; ref. 36) prevented lactate from upregulating the expression of CD14 and CD163 (Fig. 2A), suggesting that the impact of lactate on macrophage polarization resulted from its uptake into cells. MCT inhibitors blocked lactate influx and efflux (Supplementary Fig. S2A). We then formally demonstrated lactate entry and metabolism by macrophages, with 3H-labeled and 13C-labeled lactate. Results showed that the internalization of 3H-labeled lactate by differentiating GM+LA-MΦ was reduced in the presence of CHC (Supplementary Fig. S2B). NMR spectroscopy and LC-MS revealed the presence of 13C-metabolites in GM+LA-MΦ and culture supernatants, with malate, alanine, and glutamate being the most abundant of these metabolites (Fig. 2B). We also quantified glucose and lactate in the supernatants of GM-MΦ and GM+LA-MΦ during differentiation (Supplementary Fig. S2C). As expected (37), GM-MΦ consumed glucose and released lactate. GM+LA-MΦ were less glycolytic and consumed less glucose than GM-MΦ (Supplementary Fig. S2C). For GM+LA-MΦ, extracellular lactate corresponded to nonconsumed exogenous lactate and endogenous lactate released by glycolysis. Consistent with the observations reported above, extracellular lactate quantification at day 3 showed a consumption of exogenous lactate by GM+LA-MΦ (Supplementary Fig. S2C). Finally, mitochondrial respiration analyses showed less electron chain activity in GM+LA-MΦ than in GM-MΦ (37), although the mitochondria remained functional in both (Supplementary Fig. S2D).
Our data therefore show that (i) the increase in CD14 and CD163 expression induced by lactate depends on lactate uptake into cells and (ii) the lactate taken up is metabolized by human macrophages.
GM+LA-MΦ have a non–Th1-promoting inflammatory phenotype
Damage-associated molecules signaling via TLR4 accumulate in tumors (38). We therefore analyzed the cytokine profile of GM+LA-MΦ in response to the prototypic TLR4 agonist LPS.
As expected (35, 39), LPS-stimulated GM-MΦ had an IL10low IL12p70high M1-type phenotype and produced large amounts of inflammatory cytokines (Fig. 3A). Stimulated GM+LA-MΦ produced much more IL1β, IL6, TNFα, and CXCL8 than stimulated GM-MΦ, but only small amounts of IL10 (Fig. 3A). The upregulation of IL6 and TNFα production by GM+LA-MΦ was homogeneous in the cells of the CD14high CD163high population; we therefore excluded the possibility that the IL6high TNFαhigh phenotype resulted from coexistence of different cell subpopulations (Fig. 3B). GM+LA-MΦ, with and without stimulation, upregulated transcripts encoding IL1β, TNFα, and IL6, and several chemokines, including CCL2, CCL13, and CXCL5 (involved in myeloid cell recruitment), CXCL8 (involved in neutrophil recruitment), and CXCL2 and CXCL3 (which favor angiogenesis; Fig. 3C). Most of these chemokines are present at tumor sites and are associated with a poor prognosis (40).
Unlike GM-MΦ, GM+LA-Mϕ produced low amounts of the Th1-inducing factor IL12p70 and did not produce the Th1-attracting chemokine CXCL10 (Fig. 3A). Consistent with this finding, DCs generated in the presence of lactate are also unable to produce IL12 (24). GM+LA-MΦ also produced larger amounts of CXCL1 and CCL18 (involved in the recruitment of immunosuppressive myeloid cells) and of CCL24 (involved in naïve and resting T-cell recruitment) than GM-MΦ (Fig. 3C).
The capacity of lactate to induce IL1βhigh IL12low macrophages was dependent on its entry into cells, as this induction was prevented by MCT inhibitors (Fig. 2A) and not mimicked by acidosis. Indeed, consistent with previous studies (41, 42), monocyte differentiation under acidosis conditions increased the capacity of the resulting macrophages to secrete IL12p70 and IL1β, albeit to a lesser extent than in the presence of lactate (Fig. 2A). Finally, lactosis did not affect the secretion of IL1β and IL12p70 by macrophages (Fig. 2A).
Collectively, these results show that lactate induces proinflammatory macrophages (i.e., non–pro-Th1), a phenotype compatible with chronic inflammation and an ineffective adaptive immune response, as observed in established tumors (6).
Lactic acid–induced macrophages secrete tumor-promoting factors
TAMs are a source of growth factors (6). We therefore evaluated the capacity of GM+LA-MΦ to produce protumor factors. GM+LA-Mϕ produced larger amounts of transcripts than GM-MΦ, for (i) the EGFR ligands HB-EGF, TGFα, and epiregulin; (ii) the epithelial growth factors HGF and IL20; (iii) the epithelial–mesenchymal transition factor OSM; and (iv) the endothelial and mesenchymal cell proliferation and migration factors VEGF-A and PDGF-BB (Fig. 4A). Accordingly, stimulated GM+LA-MΦ produced larger amounts of HB-EGF, TGFα, IL20, OSM, VEGF-A, and ET-1 than GM-MΦ (Fig. 4B). Most of these factors are expressed by TAMs in established tumors and are associated with tumor growth and metastasis (43).
This protumor signature induced by lactate was different from that of prototypic M2 and M2-like cells (M-MΦ). GM+LA-MΦ produced larger amounts of HB-EGF, TGFα, and ET-1, and smaller amounts of VEGF-A than M2 and M2-like cells (Fig. 4B). Moreover, the amounts of most of the M2 transcripts induced by IL4 (44) were not upregulated by lactate (Supplementary Fig. S3).
Thus, lactate promotes the generation of macrophages with a phenotype combining inflammatory and tumor-trophic features.
Lactic acid enhances autocrine M-CSF consumption by differentiating myeloid cells
GM+LA-Mϕ displayed upregulation of several mRNAs, including those encoded by SERPINB2, FOLR2, SEPP1, and CD163, which are defined as M-CSF–induced genes (Fig. 5A and B; ref. 39). Consistent with the findings of studies showing that GM-CSF, by downregulating CD115 expression, limits M-CSF consumption and, consequently, the expression of M-CSF–induced genes by monocytes (35, 45–47), we observed an accumulation of autocrine M-CSF in the culture supernatants of GM-MΦ (Fig. 5C).
Based on these observations, we hypothesized that lactate might promote the consumption of autocrine M-CSF by differentiating MΦ. This hypothesis is supported by various lines of evidence. First, unlike GM-MΦ, GM+LA-MΦ displayed (i) an early increase in CD115 mRNA amounts (Supplementary Fig. S4A), (ii) an increase in membrane CD115 expression at 6 and 24 hours (followed by a return to amounts similar to those in GM-MΦ at 48 hours; Fig. 5D, left plot), and (iii) a decrease in sCD115 release (Fig. 5D, right plot), whereas total CD115 expression remained unchanged (Supplementary Fig. S4B). Second, M-CSF mRNA amounts increased in GM+LA-MΦ (Supplementary Fig. S4A), whereas M-CSF remained undetectable in the culture supernatants of these cells (Fig. 5C). Third, the CD115 tyrosine kinase inhibitor GW2580 (48), and a neutralizing anti-M-CSF (49), but not a control mAb, prevented the induction of CD163high CD14high macrophage production by lactate (Fig. 5E). GW2580 also decreased the expression of the M-CSF–dependent genes SERPINB2, FOLR2, SEPP1, and CD163 (Fig. 5B) and the secretion of VEGF-A and TGFα by GM+LA-MΦ (Fig. 5F). Fourth, flow cytometry revealed that the amount of ERK1/2 phosphorylation, which is induced by M-CSF and lactate (50), was higher in GM+LA-MΦ than in GM-MΦ (Supplementary Fig. S4C).
Although M-MΦ are not inflammatory cells (39), we observed that the acquisition of an inflammatory signature was dependent on autocrine consumption of M-CSF during the differentiation process (as illustrated by the decrease in IL1β and TNFα production by GM+LA-MΦ in the presence of GW2580; Fig. 5F). However, consistent with our observations, macrophages generated by incubation with GM-CSF, M-CSF, and IL6 (GM+M-MΦ) had an IL1βhigh, VEGF-Ahigh, and CD163high phenotype similar to that of GM+LA-MΦ (Supplementary Fig. S4D). This result is consistent with IL6 unlocking M-CSF consumption by myeloid cells (35, 45). Finally, the phenotype of GM+LA-MΦ cannot be entirely due to the consumption of M-CSF. First, GW2580 did not modulate the secretion of OSM, HB-EGF, IL12p70, or CXCL10 by GM+LA-MΦ (Supplementary Fig. S4E). Second, GM+M-MΦ produced smaller amounts of HB-EGF than GM+LA-MΦ (Supplementary Fig. S4D). Third, in the absence of GM-CSF, we detected no effect of lactate when monocytes were differentiated in the presence of M-CSF (Supplementary Fig. S3).
In conclusion, lactate induces the autocrine consumption of M-CSF by monocytes exposed to GM-CSF, and this mechanism may explain the acquisition of certain inflammatory and trophic features.
Acquisition of GM+LA-MΦ phenotype is dependent on lactate metabolism
Lactate inhibits cytoplasmic prolyl hydroxylases, which degrade the transcription factor hypoxia-inducible factor-1α (HIF-1α; ref. 51). Accordingly, lactate increases the expression of M2 genes by murine macrophages through the stabilization of HIF-1α (28). Consistent with this result, we observed that the HIF-1α inhibitor acriflavin (52) decreased the lactate-induced synthesis of M-CSF mRNA and, as previously reported in mice (53), of CD115 mRNA (Fig. 6A), supporting a role for HIF1-α in autocrine M-CSF consumption by human GM+LA-MΦ. Acriflavin also prevented upregulation of CD14 and CD163 (Fig. 6B) and of M-CSF–dependent marker expression (Fig. 6C), and production of IL1β, TNFα, HB-EGF, and OSM as induced by lactate (Fig. 6D).
Stabilization of HIF-1α induced by lactate requires lactate oxidation to pyruvate (54–56). Consistent with these studies, the LDH inhibitors oxamic acid and GSK2837808A prevented lactate from increasing CD163 and CD14 expression and IL1β and HB-EGF production by GM+LA-MΦ (Fig. 6E).
Thus, the acquisition of the GM+LA-MΦ phenotype by monocytes is dependent on lactate metabolism and HIF-1 stabilization.
Ovarian TAMs have a proinflammatory trophic phenotype
We evaluated the effect of tumor-derived lactate on monocyte differentiation, by culturing monocytes with lactate-producing primary ovarian tumor cells (Supplementary Fig. S1A). Macrophages generated in the presence of these tumor cells had high amounts of CD163, CD14, and IL1β (Fig. 7A and Supplementary Fig. S1B) and of SEPP1, FOLR2, STAB1, EREG, and OSM (Fig. 7B). These effects were prevented by CHC (Fig. 7A and B; Supplementary Fig. S1B). CKC829 and CKT151 cell lines (which produce IL6 and M-CSF in addition to lactate) induced a more pronounced expression of the M2-like marker CD163 on macrophages than CASC170 cells, reinforcing the involvement of M-CSF and IL6 in the acquisition of M2-like features by human macrophages generated under lactic acidosis (Supplementary Fig. S1A).
We then investigated whether the TAMs in solid ovarian tumors had a phenotype similar to that of GM+LA-MΦ. We monitored the amounts of (i) VEGF, as VEGFhigh TAMs are located in lactate-rich zones (29, 57), and (ii) the inflammatory cytokines CXCL8 and IL1β in LPS-stimulated cell suspensions (obtained after tumor dissociation). We detected three subsets of CD14+ cells: CXCL8− IL1β− VEGF+, CXCL8+ IL1β− VEGF+, and CXCL8+ IL1β+ VEGF+ (Fig. 7C). VEGF concentrations were highest in TAMs producing CXCL8 and IL1β (Fig. 7D), as observed in GM+LA-MΦ (Fig. 5F).
Collectively, these results demonstrate the existence of human TAMs with protumor and inflammatory features in ovarian tumors and highlight the role of tumor-derived lactate in their generation (Supplementary Fig. S5).
Most of the known macrophage-polarizing factors give rise to either trophic (M2) or inflammatory (M1) cells (5). Nevertheless, macrophages displaying both M1- and M2-like features have been reported in various situations, including cancer (12, 13). We show here that prolonged exposure to lactate, which is produced in large amounts in tumors (21, 27), confers on human macrophages a phenotype combining inflammatory and protumor features. The acquisition of this phenotype required the internalization and metabolism of lactate and was dependent on HIF-1 stabilization and, for some factors, on M-CSF consumption. These results identify tumor-derived lactate as a missing link reconciling the M2-like features of human TAMs with their inflammatory properties.
We report that GM+LA-MΦ produce numerous growth factors promoting tumor cell proliferation, migration, and invasion and that this profile more closely resembles that of macrophages generated with M-CSF (5, 39) than that of prototypic M2 cells generated with IL4 or IL13 (44). Accordingly, lactate has been reported to upregulate the expression of M2-type genes in mouse macrophages, independently of IL4 (28). Furthermore, M-CSF is overproduced in numerous solid tumors (58–60), whereas IL4 and IL13 are detected in a limited number of cancers (4, 61). In addition to having trophic properties, TAMs contribute to persistent inflammation (6) and also promote immune suppression (5, 62, 63). We show that GM+LA-MΦ have an inflammatory phenotype, as illustrated by their secretion of inflammatory cytokines and chemokines. The cytokine profile of GM+LA-MΦ suggests that they have the ability to subvert antitumor immune responses. They produce small amounts of CXCL10 and IL12, which are involved in the recruitment of CD8+ T cells and the generation of Th1 cells (40), respectively. Moreover, they produce high amounts of cytokines that favor the local accumulation of immunosuppressive myeloid cells (such as CXCL1) and of naïve T cells (CCL18 and CCL24) that ultimately differentiate into Tregs (64, 65). They also produce M-CSF that promotes monocyte recruitment and differentiation into regulatory M2-like cells (35).
We show that GM+LA-MΦ acquire most of their characteristics through the autocrine consumption of M-CSF, which is regulated in human myeloid cells. For example, M-CSF accumulates in the cell culture supernatants of DCs and M1-like macrophages (35, 45, 66), and its consumption is induced by IL6 in differentiating cells (35, 45, 67). Our results show that lactate also triggers M-CSF consumption by increasing the expressing of M-CSF and CD115, decreasing sCD115 release, and increasing CD115 downstream signaling. GM-CSF induces IL10low IL1βhigh TNFαhigh cells, whereas M-CSF induces anti-inflammatory IL10high IL1βlow TNFαlow cells (35, 39, 68). As M-MΦ are anti-inflammatory cells, we hypothesized that lactate would attenuate the inflammatory phenotype of GM-MΦ. However, GM+LA-MΦ actually had a more pronounced inflammatory phenotype than GM-MΦ. This observation can be explained by the induction of M-CSF consumption by IL6, increasing the capacity of macrophages to produce proinflammatory cytokines. Consistent with this hypothesis, in situations of chronic inflammation, M-CSF exacerbates rather than attenuates inflammation (15, 69, 70). Nevertheless, the phenotype of GM+LA-MΦ differs from that of GM+M-MΦ, confirming that the acquisition of this phenotype results from the integration of different signals.
The impact of lactate on macrophage differentiation is dependent on its internalization and metabolism, and requires HIF-1 stabilization. In murine macrophages, the expression of CD115 (53) and the induction of the M2 genes Arg1 and VEGF by lactate (28) depend on HIF-1 stabilization. As HIF-1 induces expression of the IL1β and IL6 genes (71), its stabilization can also directly induce an inflammatory phenotype in GM+LA-MΦ.
Accordingly, the phenotype of GM+LA-MΦ was not induced by lactosis or acidosis. Lactate decreased the amount of IL12 production, but, as in other studies (41), we found that acidosis increased IL12 secretion by human myeloid cells. Acidosis also increased IL1β secretion by macrophages, but to a lesser extent than lactate.
Our results, together with published results for studies in mice (28–30), highlight the role of tumor-derived lactate in shaping the phenotype of human and mouse TAMs. We show that tumor-derived lactate induces the production of inflammatory protumor macrophages in the presence of oxygen. VEGFhigh TAMs have been reported to be localized in ischemic zones, characterized by hypoxia and high concentrations of lactate (29, 57). Hypoxia acts in synergy with lactate to induce proangiogenic protumor TAMs (28, 29), but hypoxia alone does not skew murine TAMs toward a tumor-promoting phenotype (30, 72). In addition, the ERK1/2 pathway, which is involved in the CD115-induced signaling, can also be activated by lactate and hypoxia (50), highlighting the additive/synergistic impact of different differentiation signals in ischemic areas. These studies and our results confirm that lactate is a tumor-derived metabolite playing a role in the induction of TAMs. Finally, we report that TAMs in solid ovarian tumors are functionally heterogeneous, with some expressing an inflammatory and protumor phenotype (IL1β+ CXCL8+ VEGF+); these TAMs, with a phenotype resembling that of GM+LA-MΦ, are probably located in or near ischemic zones in ovarian cancers.
The cross-talk between lactate, M-CSF, and GM-CSF may affect cancer treatment. In the early stages of tumor development, GM-CSF participates in antitumor immune surveillance by inducing antitumor macrophages (73). By contrast, in advanced tumors, the amounts of GM-CSF and M-CSF increase, and high amounts of GM-CSF or M-CSF are associated with poor prognosis and metastasis (15). GM-CSF then favors the recruitment and accumulation of immunosuppressive myeloid cells and its neutralization reduces tumor growth (74–77). Similarly, M-CSF promotes monocyte recruitment and differentiation into M2-like cells (58, 78). The ability of lactate to induce, in the presence of GM-CSF, the consumption of M-CSF by monocytes/macrophages suggests that the lactate secreted by growing tumors in vivo switches monocyte differentiation from M1 cells to inflammatory and trophic TAMs. This helps explain the antagonistic roles of GM-CSF at different stages of tumor development (Supplementary Fig. S5). Finally, as M-CSF and lactate act together to shape the phenotype of human TAMs in established tumors, combination treatments aiming to decrease tumor cell glycolysis and to neutralize M-CSF/CD115 could work synergistically to re-educate TAMs and direct them against tumors.
In conclusion, we demonstrate that lactate confers a protumor and inflammatory phenotype on human macrophages, underlining its role in inducing persistent and unresolved cancer-related inflammation and sustaining tumor growth, without initiating an effective antitumor immune response. These observations highlight the potential therapeutic value of drugs that decrease tumor glycolysis as a means of reprogramming TAMs to boost the efficacy of immunotherapies.
Disclosure of Potential Conflicts of Interest
M. Campone is consultant for Pierre Fabre Oncology, Sanofi, Novartis, and Servier; reports receiving other commercial research support from Pfizer, AstraZeneca, Sanofi, Pierre Fabre Oncology, Takeda, AbbVie, Servier, Sandoz, and Accord; and reports receiving speakers bureau honoraria from Novartis. No potential conflicts of interest were disclosed by the other authors.
Conception and design: L. Paolini, C. Adam, L. Preisser, V. Procaccio, Y. Delneste, P. Jeannin
Development of methodology: L. Paolini, C. Adam, L. Preisser, S. Blanchard, P. Pignon, G. Lenaers, V. Procaccio
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Paolini, C. Adam, L. Preisser, S. Blanchard, P. Pignon, L.-M. Chevalier, R. Wernert, P. Raro, N. Ifrah, V. Lavoué, P. Descamps, A. Morel, V. Catros, G. Tcherkez, C. Bocca, J. Kouassi Nzoughet
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Paolini, C. Adam, L. Preisser, S. Blanchard, P. Pignon, V. Seegers, P. Descamps, G. Tcherkez, C. Bocca, J. Kouassi Nzoughet, V. Procaccio
Writing, review, and/or revision of the manuscript: L. Paolini, C. Adam, C. Beauvillain, M. Campone, N. Ifrah, P. Descamps, G. Tcherkez, J. Kouassi Nzoughet, V. Procaccio, Y. Delneste, P. Jeannin
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Paolini, C. Adam, V. Verrielle, G. Lenaers
Study supervision: Y. Delneste, P. Jeannin
The authors acknowledge Muriel Gervois (University Hospital of Rennes, Rennes, France), Catherine Guillet, and Jérome Cayon (PACeM facility; University of Angers) for expert technical assistance. This work was realized in the context of (i) the LabEX IGO program (National Research Agency; ANR-11-LABX-0016-01) and (ii) the SIRIC ILIAD program (National Cancer Institute, Ministry of Health and the Institute for Health and Medical Research; INCa-DGOS-Inserm_12558).
This work was supported by institutional grants from the University of Angers and Inserm and by grants from the Ligue contre le Cancer (comité du Maine et Loire, Angers, France) and the regional council of Pays de la Loire (Canceropole Grand-Ouest; projects PlasTICO and Concerto). L. Paolini is supported by a grant from Angers Loire Métropole, and C. Adam is a recipient of a grant from the French Ministry of Higher Education and Research.
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.