Differential Regulation of Macrophage Glucose Metabolism by Macrophage Colony-stimulating Factor and Granulocyte-Macrophage Colony-stimulating Factor: Implications for ¹⁸F FDG PET Imaging of Vessel Wall Inflammation

Despite the significant differences in the immunologic and metabolic profiles of macrophages stimulated with macrophage colony-stimulating factor (M-CSF) versus granulocyte-M-CSF, their phenotypes are not distinguishable with fluorine 18 fluorodeoxyglucose PET because of their similar level of glucose uptake.

Abstract

Purpose

To determine the divergence of immunometabolic phenotypes of macrophages stimulated with macrophage colony-stimulating factor (M-CSF) and granulocyte-M-CSF (GM-CSF) and its implications for fluorine 18 (¹⁸F) fluorodeoxyglucose (FDG) imaging of atherosclerosis.

Materials and Methods

This study was approved by the animal care committee. Uptake of 2-deoxyglucose and various indexes of oxidative and glycolytic metabolism were evaluated in nonactivated murine peritoneal macrophages (MΦ₀) and macrophages stimulated with M-CSF (MΦ_M-CSF) or GM-CSF (MΦ_GM-CSF). Intracellular glucose flux was measured by using stable isotope tracing of glycolytic and tricyclic acid intermediary metabolites. ¹⁸F-FDG uptake was evaluated in murine atherosclerotic aortas after stimulation with M-CSF or GM-CSF by using quantitative autoradiography.

Results

Despite inducing distinct activation states, GM-CSF and M-CSF stimulated progressive but similar levels of increased 2-deoxyglucose uptake in macrophages that reached up to sixfold compared with MΦ₀. The expression of glucose transporters, oxidative metabolism, and mitochondrial biogenesis were induced to similar levels in MΦ_M-CSF and MΦ_GM-CSF. Unexpectedly, there was a 1.7-fold increase in extracellular acidification rate, a 1.4-fold increase in lactate production, and overexpression of several critical glycolytic enzymes in MΦ_M-CSF compared with MΦ_GM-CSF with associated increased glucose flux through glycolytic pathway. Quantitative autoradiography demonstrated a 1.6-fold induction of ¹⁸F-FDG uptake in murine atherosclerotic plaques by both M-CSF and GM-CSF.

Conclusion

The proinflammatory and inflammation-resolving activation states of macrophages induced by GM-CSF and M-CSF in either cell culture or atherosclerotic plaques may not be distinguishable by the assessment of glucose uptake.

^© RSNA, 2016

Online supplemental material is available for this article.

Introduction

Macrophages are the frontline defense of innate immunity and play key roles in the pathogenesis of many vascular diseases, including atherosclerosis (1) and aortic aneurysm (2). Dichotomization of macrophages into classic (M1) and alternatively (M2) polarized is a broadly used framework to represent the extreme pro- and antiinflammatory states of the macrophage activation spectrum. M1 polarized macrophages promote a type 1 T helper lymphocyte immune response (Th1) and are critical in antimicrobial and antitumor defense. M2 polarized macrophages are associated with a type 2 T helper lymphocyte (Th2) response and contribute to tissue repair and resolution of inflammation (3). The balance between M1 and M2 polarized macrophages is critical in determining the fate (ie, resolution vs progression) of inflammation (1,4).

Fluorine 18 (¹⁸F) fluorodeoxyglucose (FDG) positron emission tomography (PET) was extensively investigated to detect the inflammatory state and macrophage burden of atherosclerotic plaques (5–7). However, its potential for the identification of vulnerable plaques remains unclear (8). Complicating the interpretation of ¹⁸F-FDG uptake are the divergent metabolic responses associated with different macrophage activation state, which may have contributed to the inconsistent results from clinical ¹⁸F-FDG PET studies (8–14). For example, glucose uptake and glycolysis are induced by bacterial-derived products, such as lipopolysaccharide (13), modified low-density lipoprotein (15,16), or hypoxia (17). However, the link between a proinflammatory macrophage phenotype and enhanced glucose uptake may be valid for only a number of stimuli and does not represent a universal phenomenon (9,14). Therefore, further investigations into the role of microenvironmental factors, such as various immunoregulatory cytokines and growth factors, (modified) lipoproteins, and hypoxia, in the regulation of macrophage metabolism are required to improve our understanding of the biologic basis of ¹⁸F-FDG uptake in atherosclerosis.

Macrophage colony-stimulating factor (M-CSF) and granulocyte-M-CSF (GM-CSF) are key growth factors involved in the development of the myeloid lineage. Whereas both M-CSF and GM-CSF promote survival and differentiation of macrophages, they can induce differing effects on macrophage phenotypes (18). M-CSF is present in many tissues in the steady-state condition and is implicated in the maintenance of the tissue macrophage homeostasis and the resolution of inflammation (18). However, GM-CSF expression is mostly limited to inflammatory conditions (18). The cytokine profiles of M-CSF and GM-CSF stimulated macrophages resemble M2 and M1 polarized macrophages, respectively (18). The balance between these antagonistic M2-like effects of M-CSF and M1-like effects of GM-CSF may be fundamental in determining whether inflammatory processes progress or resolve (18,19).

M-CSF and GM-CSF are abundant in atherosclerotic plaques and are produced by endothelial and vascular smooth muscle cells in response to modified low-density lipoprotein (20). Local production of M-CSF promotes the survival of foam cells and contributes to development of the early-stage, but not late-stage, atherosclerotic lesions (20,21). However, GM-CSF renders plaque macrophages susceptible to apoptosis and promotes the progression of advanced atherosclerotic plaques (22). GM-CSF also contributes to the development of aortic aneurysm (23), dissection, and intramural hematoma (24).

The purpose of this study was to determine the divergence of immunometabolic phenotypes of macrophages stimulated by M-CSF and GM-CSF and its implications for ¹⁸F-FDG imaging of atherosclerosis. We hypothesized that M-CSF and GM-CSF induce distinct metabolic profiles in macrophages and atherosclerotic lesions, which can be distinguished by the measurement of ¹⁸F-FDG uptake. To determine the specificity of ¹⁸F-FDG imaging as a biomarker for phenotypic changes of macrophages in atherosclerosis, we investigated the metabolic changes in response to M-CSF and GM-CSF inducing inflammation-resolving (M2-like) and proinflammatory (M1-like) activation states, respectively.

Materials and Methods

Experiments were performed according to regulations of the animal care and use committee. Detailed methods are available in the online supplement.

Study Design

Peritoneal macrophages from C57BL/6 J mice were activated for 2 days with M-CSF (MΦ_M-CSF) or GM-CSF (MΦ_GM-CSF). Nonactivated macrophages were considered as MΦ₀. Macrophages were incubated with ³H-2-deoxyglucose for 30 minutes. DNA-normalized uptake values are presented as relative to MΦ₀. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured at baseline and after administration of oligomycin, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, and rotenone+antimycin. Quantitative real-time polymerase chain reaction experiments were performed by using a high-throughput nanofluidic platform, according to the manufacturer’s protocols. The complete list of primers is provided in Table E1–E3 (online). Quantitative polymerase chain reaction amplification of a mitochondrial-DNA-encoded (cytochrome-c oxidase subunit-I) and a nuclear-DNA-encoded (NADH [ubiquinone] flavoprotein-1) gene was performed to determine the ratio of mitochondrial DNA to nuclear DNA. Macrophages were incubated in a culture media supplemented with D-glucose-6-¹³C for 2 hours. Metabolites were identified with combined liquid chromatography–mass spectrometry/gas chromatography–mass spectrometry analyses. ¹⁸F-FDG uptake was assessed in atherosclerotic aortas from low-density lipoprotein-receptor-deficient mice. After 24 hours of incubation with M-CSF or GM-CSF, aortas were incubated with ¹⁸F-FDG for 30 minutes and exposed to a phosphor screen.

Statistical Analysis

Data are expressed as mean ± standard error of the mean. Statistical analyses were performed with statistical software (SigmaPlot-12.5, Systat Software, Chicago, Ill; and Stata-12, StataCorp, College Station, Tex). P value less than .05 indicated statistical significance.

Results

Induction of Distinct Macrophage Phenotypes by M-CSF and GM-CSF

Cultured MΦ₀ displayed a mixed population of spherical and elongated cells (Fig 1, left image). Both M-CSF and GM-CSF induced morphologic changes over the course of the 2-day incubation. MΦ_M-CSF demonstrated an enlarged and flattened morphologic structure with long cytoplasmic extensions and a considerable number of cytoplasmic vacuoles (Fig 1, middle image). MΦ_GM-CSF were predominantly composed of both flat and elongated cells and a few round cells (Fig 1, right image).

Figure 1:

Images show morphologic structure of murine macrophages. Inverted light microscopy images of murine macrophages after 2-day incubation with vehicle (MΦ₀), M-CSF (MΦ_M-CSF), or GM-CSF (MΦ_GM-CSF) demonstrated induction of distinct morphologic structures. Numerous cytoplasmic vacuoles were present in MΦ_M-CSF (digitally magnified inset in the middle panel), but not in MΦ₀ and MΦ_GM-CSF.

Heat map analysis (Fig 2, A) revealed significant overexpression of a number of M1 markers (interleukin 1 β, interleukin 6, and arginase type II) in MΦ_GM-CSF and of M2 markers (folate receptor 2, insulin-like growth factor 1, and transferrin receptor) in MΦ_M-CSF, consistent with their reported M1-like and M2-like phenotypes, respectively. However, MΦ_M-CSF also showed increased expression of selected M1 markers (interleukin 12b, interleukin 17 receptor A, chemokine [C-C motif] ligand 2, and chemokine [C-C motif] ligand 7). Similarly, a number of M2 markers (chemokine [C-C motif] ligand 9, macrophage galactose N-acetyl-galactosamine-specific lectin 2, chitinase-like 3, and chemokine [C-C motif] ligand 17) were upregulated in MΦ_GM-CSF. These findings confirm that gene expression patterns in MΦ_GM-CSF and MΦ_M-CSF are distinct from established patterns of conventional M1 and M2 polarized macrophages (18). Principal component analysis score plots clearly depict the separate clustering of MΦ₀, MΦ_M-CSF, and MΦ_GM-CSF on the basis of M1 (Fig 2, B; Table E4 [online]) and M2 (Fig 2, C; Table E5 [online]) markers.

Figure 2:

Distinct inflammatory profiles of MΦ_M-CSF and MΦ_GM-CSF. A, Heat map representation of mRNA expression (relative to MΦ₀ ) of M1 and M2 polarization markers. Transcripts, which are differentially upregulated in MΦ_M-CSF or MΦ_GM-CSF, are marked in a blue and a red font, respectively. Principal component analysis score plots of the first two components of, B, M1 transcripts (explaining 41% and 31% of the variance) and, C, M2 transcripts (explaining 47% and 32% of the variance) confirm distinct clustering of MΦ₀, MΦ_M-CSF and MΦ_GM-CSF. The colored dots in, B, and, C, represent individual biologic replicates. D, Flow cytometry for selected polarization markers confirms the overexpression of transferrin receptor (TFRC ) and cluster of differentiation (CD)11c in MΦ_M-CSF and MΦ_GM-CSF, respectively. The expression of CD206 was induced to the same level by both M-CSF and GM-CSF. * P = .06.

We analyzed the expression of three cell surface activation markers, CD206, CD11c, and transferrin receptor by individual F4/80⁺ macrophages by using flow cytometry (Fig E1 [online]). The expression of transferrin receptor and CD11c was homogenously elevated by MΦ_M-CSF and MΦ_GM-CSF, respectively (Fig 2, D). The cell surface expression of CD206 was also homogeneously induced, but to a similar level, in both M-CSF and GM-CSF (Fig 2, D). Together, these data confirm that both M-CSF and GM-CSF induce unique macrophage phenotypes distinct from those induced by conventional activators of M1 and M2 polarization.

Enhanced ³H-2-Deoxyglucose Uptake by MΦ_M-CSF and MΦ_GM-CSF

Both M-CSF and GM-CSF caused a progressive increase in ³H-2-deoxyglucose uptake by macrophages, which after 48 hours was increased over sixfold relative to MΦ₀ (Fig 3, A). We next evaluated the expression of the major isoforms of glucose transporters expressed by macrophages (ie, glucose transporter-1 [Glut1], glucose transporter-6 [Glut6], and glucose transporter-3 [Glut3]) (25) (Fig 3, B–D). Consistent with the enhanced ³H-2-deoxyglucose uptake, we observed an over threefold increase in the expression of Glut1 and Glut6 by MΦ_M-CSF and MΦ_GM-CSF, compared with MΦ₀. Glut3, which represented a less abundant glucose transporter in macrophages (25), was induced by 2.3-fold in MΦ_GM-CSF, but not in MΦ_M-CSF.

Figure 3:

Stimulation of glucose uptake and induction of glucose transporters by M-CSF and GM-CSF. A, ³H-2-deoxyglucose uptake was substantially increased in MΦ_M-CSF and MΦ_GM-CSF over 48 hours. Consistently, a 2-day incubation with M-CSF or GM-CSF showed statistically significant increase in expression of the major macrophage glucose transporters, B, glucose transporter-1 (Glut1), C, glucose transporter-3 (Glut3 ), and, D, glucose transporter-6 (Glut6 ). * P < .05; *** P < .001 MΦ_M-CSF versus MΦ₀; ^## P < .01; ^### P < .001, MΦ_GM-CSF versus MΦ₀.

Mitochondrial Biogenesis and Enhanced Oxidative Metabolism in MΦ_M-CSF and MΦ_GM-CSF

We quantified OCR, as a marker of oxidative metabolism, at the basal state and in response to mitochondrial inhibitors (Fig 4 A, Fig E2 [online]). MΦ_M-CSF and MΦ_GM-CSF demonstrated 2.3-fold (P = .007) and 2.4-fold (P = .005) increase in basal OCR compared with MΦ₀. Oligomycin-sensitive OCR (representing the fraction of basal OCR attributable to adenosine triphosphate production) and oligomycin-insensitive OCR (representing proton leak across the inner mitochondrial membrane) were increased in both MΦ_M-CSF and MΦ_GM-CSF compared with MΦ₀ (Fig E2 [online]). The maximal respiratory capacity achieved in response to carbonyl cyanide-p-trifluoromethoxyphenylhydrazone was over 2.5-fold higher in MΦ_M-CSF and MΦ_GM-CSF compared with MΦ₀ (Fig 4, A, Fig E2 [online]). The rotenone and antimycin-insensitive fraction of OCR, an indicator of nonmitochondrial oxygen utilization, was also higher in MΦ_M-CSF and MΦ_GM-CSF compared with MΦ₀ (Fig 4, A, Fig E2 [online]).

Figure 4:

Graphs show divergence of oxidative and glycolytic metabolism in MΦ_M-CSF and MΦ_GM-CSF. A, Real-time measurements of OCR at basal state and after sequential incubation with inhibitors of specific mitochondrial components demonstrated a marked induction in cellular respiration by both M-CSF and GM-CSF. B, Consistently, there is a significant induction of mitochondrial biogenesis in both MΦ_M-CSF and MΦ_GM-CSF. MΦ_M-CSF demonstrates a significantly higher, C, basal ECAR and, D, ECAR-to-OCR ratio compared with MΦ₀ and MΦ_GM-CSF. However, ECAR increases to a comparable level in MΦ_M-CSF and MΦ_GM-CSF after incubation with an adenosine triphosphate synthase inhibitor, oligomycin. mtDNA = mitochondrial DNA, nDNA = nuclear DNA, NS = not significant.

Consistent with their increased OCR, MΦ_M-CSF and MΦ_GM-CSF had 1.8-fold (P < .001) and 1.5-fold (P = .002), respectively, higher mitochondrial DNA copy numbers per cell compared with MΦ₀, which indicated a remarkable mitochondrial biogenesis in these cells (Fig 4, B).

Divergence of Glycolytic Rate in MΦ_M-CSF and MΦ_GM-CSF

ECAR, a marker of glycolytic activity, was measured in basal state and in response to oligomycin (Fig 4, C). MΦ_M-CSF showed 4.7-fold (P < .001) and MΦ_GM-CSF showed a 2.8-fold (P = .046) higher basal ECAR compared with MΦ₀ (Fig 4, C, Fig E2 [online]). The maximal glycolytic rate achieved in response to inhibition of adenosine triphosphate synthesis by oligomycin was approximately threefold lower in MΦ₀ compared with MΦ_M-CSF and MΦ_GM-CSF (P < .001) (Fig 4, C). Despite the 1.7-fold lower basal ECAR of MΦ_GM-CSF compared with MΦ_M-CSF (P = .033), the maximal glycolytic rate in response to oligomycin was similar in both macrophage populations, consistent with a larger glycolytic reserve in MΦ_GM-CSF (Fig 4, C). Comparison of the basal ECAR-to-OCR ratio (Fig 4, D) indicated a significantly higher reliance of MΦ_M-CSF on glycolytic metabolism, for energy production or carbon metabolism, than either MΦ₀ (P = .023) or MΦ_GM-CSF (P = .035). Consistently, lactate production increased by 4.9-fold in MΦ_M-CSF and only 3.6-fold in MΦ_GM-CSF compared with MΦ₀ (P < .001) (Fig E3 [online]), which indicated a 40% higher basal lactate production in MΦ_M-CSF compared with MΦ_GM-CSF (P = .015).

Enhanced Expression of Glycolytic and Tricarboxylic Acid Cycle Enzymes in MΦ_M-CSF and MΦ_GM-CSF

Consistent with the OCR and ECAR data, there was a near-global induction by both M-CSF and GM-CSF of genes encoding both glycolytic (Fig 5a) and tricarboxylic acid cycle (Fig 5b) enzymes, albeit with some notable exceptions, including enolase 3, enolase 2, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 4, isocitrate dehydrogenase 1, and isocitrate dehydrogenase 3b. Enolase 2 was the only glycolytic transcript that was significantly downregulated by both M-CSF and GM-CSF. A few glycolytic (hexokinase 1, hexokinase 3, phosphofructokinase [muscle], 6-phosphofructo-2-kinase/fructose-2, 6-biphosphatase 2, and phosphoglycerate mutase 5) and tricarboxylic acid cycle (isocitrate dehydrogenase 2, isocitrate dehydrogenase 3g, oxoglutarate dehydrogenase, and succinate-Coenzyme A ligase β) enzymes were only induced in MΦ_M-CSF, but not in MΦ_GM-CSF.

Figure 5a:

Induction of glycolytic and tricarboxylic acid cycle enzymes by M-CSF and GM-CSF. Heat map representation of relative messenger ribonucleic acid expression (fold-change compared with MΦ₀ ) of genes encoding (a) glycolytic and (b) tricarboxylic acid cycle enzymes. Transcripts that are significantly upregulated in MΦ_M-CSF are blue and those overexpressed in both MΦ_M-CSF and MΦ_GM-CSF are green. (c) Transcripts that are differentially overexpressed in MΦ_M-CSF and MΦ_GM-CSF are in blue and red, respectively, in a metabolic pathway model. * P = .06.

Figure 5b:

Induction of glycolytic and tricarboxylic acid cycle enzymes by M-CSF and GM-CSF. Heat map representation of relative messenger ribonucleic acid expression (fold-change compared with MΦ₀ ) of genes encoding (a) glycolytic and (b) tricarboxylic acid cycle enzymes. Transcripts that are significantly upregulated in MΦ_M-CSF are blue and those overexpressed in both MΦ_M-CSF and MΦ_GM-CSF are green. (c) Transcripts that are differentially overexpressed in MΦ_M-CSF and MΦ_GM-CSF are in blue and red, respectively, in a metabolic pathway model. * P = .06.

Figure 5c:

Induction of glycolytic and tricarboxylic acid cycle enzymes by M-CSF and GM-CSF. Heat map representation of relative messenger ribonucleic acid expression (fold-change compared with MΦ₀ ) of genes encoding (a) glycolytic and (b) tricarboxylic acid cycle enzymes. Transcripts that are significantly upregulated in MΦ_M-CSF are blue and those overexpressed in both MΦ_M-CSF and MΦ_GM-CSF are green. (c) Transcripts that are differentially overexpressed in MΦ_M-CSF and MΦ_GM-CSF are in blue and red, respectively, in a metabolic pathway model. * P = .06.

Consistent with the higher glycolytic metabolism in MΦ_M-CSF, the expression of multiple glycolytic enzymes (ie, hexokinase 1, hexokinase 3, glucose phosphate isomerase 1, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2, 6-phosphofructo-2-kinase/fructose-2, 6-biphosphatase 3, phosphofructokinase [muscle], glyceraldehyde-3-phosphate dehydrogenase, phosphoglycerate kinase 1, triosephosphate isomerase 1, phosphoglycerate mutase 5, and pyruvate kinase [muscle]) was increased in MΦ_M-CSF compared with MΦ_GM-CSF (Fig 5c). Interestingly, for all irreversible reactions in the glycolytic pathway, at least one of the catalyzing enzymes was differentially upregulated by M-CSF. Aldolase C was the only glycolytic enzyme, which was differentially overexpressed by MΦ_GM-CSF. Conversely and consistent with the similar respiratory profile of MΦ_M-CSF and MΦ_GM-CSF, the majority of tricarboxylic acid cycle enzymes were induced to similar levels, including enzymes catalyzing the irreversible reactions of pyruvate conversion to acetyl-CoA (pyruvate dehydrogenase α, pyruvate dehydrogenase β, and dihydrolipoamide S-acetyltransferase) and the condensation reaction of acetyl-CoA and oxaloacetate by citrate synthase. However, a few other transcripts (isocitrate dehydrogenase 2, isocitrate dehydrogenase 3α, oxoglutarate dehydrogenase, dihydrolipoamide S-succinyltransferase, and succinate-coenzyme A ligase β) were expressed at significantly lower levels in MΦ_GM-CSF compared with MΦ_M-CSF.

Enhanced Glycolytic and Oxidative Flux of Glucose in MΦ_M-CSF and MΦ_GM-CSF

To confirm that enhanced ECAR and OCR as well as the overexpression of metabolic enzymes are associated with increased glycolytic and tricarboxylic acid cycle flux of glucose, we performed ¹³C isotope tracing metabolomics experiments. As shown in Figure E4 (online), the intracellular level of multiple metabolites were increased in MΦ_M-CSF (glucose/fructose 6-phosphate, fructose 1,6-bisphosphate, triose-phosphate, 2/3-phosphoglycerate, lactate [P = .07], acetyl-coenzyme A, citrate/isocitrate, succinate, and malate) and in MΦ_GM-CSF (glucose/fructose 6-phosphate, lactate [P = .07], citrate/isocitrate, succinate, and malate) compared with MΦ₀. Consistent with their higher ECAR, the level of several key glycolytic intermediates and the percentage of ¹³C-isotope enrichment (fructose 1,6-bisphosphate, triose-phosphate, and 2/3-phosphoglycerate) were increased in MΦ_M-CSF compared with MΦ_GM-CSF.

Enhanced Atherosclerotic Plaques ¹⁸F-FDG Uptake by M-CSF and GM-CSF

Autoradiography of aorta from high-fat-diet-fed low-density lipoprotein receptor-deficient mice demonstrated foci of increased ¹⁸F-FDG uptake, corresponding to atherosclerotic plaques, along the length of aortas, most prominently involving the ascending aorta and aortic arch (Fig 6a, b). Quantitative analysis of the autoradiography images (Fig 6c) demonstrated a 1.7- and 1.6-fold increase in average ¹⁸F-FDG uptake throughout the length of aorta after 24 hours stimulation with M-CSF and GM-CSF, respectively (P = .066). Similarly, the maximal plaque uptake in aortic arch and ascending aorta was more than 1.6-fold increased by both M-CSF (P = .027) and GM-CSF (P = .013). A similar pattern was observed in plaques within the distal segments of the aorta, though the difference between the groups was not statistically significant.

Figure 6a:

Induction of plaque ¹⁸F-FDG uptake by M-CSF and GM-CSF. Representative ex vivo ¹⁸F-FDG (a) autoradiography and (b) light images of atherosclerotic aortas of low-density lipoprotein receptor-deficient mice fed with high-fat diet for 3 months. (c) Box plot of the autoradiography images demonstrates an increase in the average ¹⁸F-FDG uptake along the aorta and maximal plaque uptake in the aortic arch and ascending aorta after stimulation with M-CSF and GM-CSF.

Figure 6b:

Induction of plaque ¹⁸F-FDG uptake by M-CSF and GM-CSF. Representative ex vivo ¹⁸F-FDG (a) autoradiography and (b) light images of atherosclerotic aortas of low-density lipoprotein receptor-deficient mice fed with high-fat diet for 3 months. (c) Box plot of the autoradiography images demonstrates an increase in the average ¹⁸F-FDG uptake along the aorta and maximal plaque uptake in the aortic arch and ascending aorta after stimulation with M-CSF and GM-CSF.

Figure 6c:

Induction of plaque ¹⁸F-FDG uptake by M-CSF and GM-CSF. Representative ex vivo ¹⁸F-FDG (a) autoradiography and (b) light images of atherosclerotic aortas of low-density lipoprotein receptor-deficient mice fed with high-fat diet for 3 months. (c) Box plot of the autoradiography images demonstrates an increase in the average ¹⁸F-FDG uptake along the aorta and maximal plaque uptake in the aortic arch and ascending aorta after stimulation with M-CSF and GM-CSF.

Discussion

Our data demonstrate distinct metabolic and inflammatory profiles for MΦ_M-CSF and MΦ_GM-CSF. However, both macrophage phenotypes showed identical levels of 2-deoxyglucose uptake. These findings and previous studies (9,14) suggest a limited ability of ¹⁸F-FDG PET to accurately identify phenotypic and thus functional changes in macrophages.

The basal constitutive expression of M-CSF in various tissues is suggested to compromise the capacity of macrophages to produce proinflammatory mediators during noninflammatory states (26). Consistently, our analysis revealed a lower level of interleukin 1 β, interleukin 6, and arginase type II expression in MΦ_M-CSF. During inflammatory processes, levels of GM-CSF increase and ultimately prevail over the homeostatic function of M-CSF and accentuate the expression of several proinflammatory cytokines, such as tumor necrosis factor–α, interleukin-12, and interleukin-23 (26). However, an increase in tissue levels of M-CSF at inflammatory sites, through either increased expression or exogenous administration, ameliorates the inflammatory response of macrophages and supports the resolution of the inflammation by favoring the reacquisition of homeostatic phenotype of macrophages, which decreases their expression of proinflammatory mediators (26).

Monocytes and macrophages use diverse metabolic pathways to adapt to various microenvironments, which range from well-oxygenated tissues and blood to inflamed and often hypoxic tissues (27). Macrophages have the capacity to effectively switch to glycolysis after exposure to hypoxia (14,17), bacterial products (9,13), or in response to certain other proinflammatory stimuli (eg, modified lipoproteins) (12,15). However, macrophages can efficiently use oxidative metabolism after exposure to the anti-inflammatory cytokine interleukin-4 to support their bioenergetic and metabolic needs (9,28). Beyond the potential survival advantage, this metabolic plasticity is involved in the regulation of macrophage responses to microenvironmental stimuli (13,14,17,28). This association between metabolic and phenotypic changes provides potential opportunities to develop imaging strategies aimed at interrogating the functional state of macrophages.

Although the general assumption is that enhanced ¹⁸F-FDG uptake reflects proinflammatory activation of macrophages, ex vivo data addressing the effect of proinflammatory cytokines on glucose uptake were inconsistent. For example, glucose uptake remains unaffected by proinflammatory activation of macrophages induced by interferon-γ in combination with tumor necrosis factor–α and/or interleukin-1β (9,14). However, hypoxia (14) and proinflammatory activation by lipopolysaccharide (9,13) or modified low-density lipoprotein (15,16) enhance glucose uptake. Our data indicate that both M-CSF and GM-CSF markedly increase 2-deoxyglucose uptake, which is associated with an upregulation of glucose transporters.

We noted a markedly increased oxidative metabolism and mitochondrial biogenesis in both MΦ_M-CSF and MΦ_GM-CSF compared with MΦ₀. Mitochondrial biogenesis and enhanced respiratory capacity were previously reported in conventional M2 polarized macrophages induced by interleukin-4 (9,28). Here, we demonstrate that these metabolic changes may occur in both M1-like (MΦ_GM-CSF) and M2-like (MΦ_M-CSF) macrophages. This finding is particularly interesting because M1-like MΦ_GM-CSF demonstrated mitochondrial biogenesis and a robust increase in OCR, whereas the conventional proinflammatory phenotypes induced by interferon-γ plus tumor necrosis factor–α or lipopolysaccharide suppress OCR (9).

Unexpectedly, we noted a higher glycolytic activity and upregulation of key glycolytic enzymes in MΦ_M-CSF compared with MΦ_GM-CSF. While broadly linked to the proinflammatory macrophage phenotype (12,13,17), our results indicate that glycolysis may increase in response to M-CSF despite its established M2-skewing effects. It is noteworthy that unlike the enhanced glycolysis observed in conventional proinflammatory M1 or hypoxic macrophages, which is associated with suppressed oxidative metabolism (9,12–14,17), M-CSF induces a robust increase in both glycolysis and mitochondrial metabolism, which depicts another level of complexity in macrophage metabolic diversity (9,12–17,28). The functional consequences of such metabolic complexity in the regulation of various aspects of macrophage function (eg, cytokine production, phagocytosis, and cell migration) remain to be determined.

Macrophages are highly heterogeneous cells with major functional and developmental differences (29). Two recent studies (10,11) in human monocyte-derived macrophages and murine bone marrow-derived macrophages reported enhanced glycolytic rate in response to GM-CSF compared with M-CSF. Although the reason for this conflicting data with our results is not clear, we speculate that the different sources of macrophages and differences in the ex vivo culture condition of macrophages, which is far from their complex in vivo microenvironment, may be key factors. Indeed, in the two studies referenced above (10,11), longer incubation periods with M-CSF and GM-CSF (7 days vs 2 days in our study) were used to achieve both the initial differentiation of monocytes and bone marrow progenitors into macrophages and subsequently the polarization of these cells. Of note, long-term GM-CSF incubation results in the differentiation of bone marrow progenitors into a heterogeneous myeloid population, which includes macrophages that express various degrees of macrophage and dendritic cell differentiation markers (10). We attempted to address this limitation by using mature macrophages, which are sufficiently polarized by only a short-term, 48-hour stimulation with M-CSF and GM-CSF. Another approach that has been taken to address the above limitations was the evaluation of ¹⁸F-FDG uptake in aortic organ culture as an independent experimental model, in which the microenvironment may resemble more to the in vivo condition compared with cell culture. Interestingly, this model confirmed the induction of a similar level of ¹⁸F-FDG uptake by both M-CSF and GM-CSF in atherosclerotic lesions.

Our data demonstrate that despite the differences in the immunologic and metabolic profiles of macrophages stimulated with M-CSF versus GM-CSF, their phenotypes are not distinguishable by ¹⁸F-FDG PET because of their similar level of glucose uptake. In agreement with the previous reports (9,14), enhanced glucose uptake and glycolysis appear to depend on specific stimuli and do not appear to represent a unique discriminating feature of proinflammatory M1 or M1-like macrophage phenotypes. Therefore, enhanced ¹⁸F-FDG uptake may best be viewed as an indicator of a higher macrophage burden of atherosclerotic plaques rather than changes in their specific (proinflammatory) activation state, as previously suggested (9,30). These findings suggest a more prudent approach in the interpretation of ¹⁸F-FDG PET imaging, which relies on a nearly ubiquitous metabolic process. Therefore, the development and validation of novel imaging tracers may provide alternative and more reliable approaches in the detection of vulnerable plaques through more specific targeting of processes that are upregulated in high-risk lesions.

Advances in Knowledge

• ■ Macrophage colony-stimulating factor (M-CSF) and granulocyte-M-CSF (GM-CSF), two major myeloid growth factors with known homeostatic versus inflammatory functions, induce unique metabolic profiles in macrophages that are distinct from those induced by conventional proinflammatory and inflammation-resolving polarizing stimuli.

• ■ Despite their inflammation-resolving phenotype, M-CSF–stimulated macrophages are more glycolytic than GM-CSF–stimulated macrophages, which indicates that enhanced glycolysis may not be an intrinsic feature of the proinflammatory macrophage phenotype.

• ■ Despite the induction of distinct inflammatory and metabolic profiles, M-CSF and GM-CSF induce comparable levels of glucose uptake in cultured macrophages and atherosclerotic plaques, prohibiting their discrimination on the basis of fluorine 18 (¹⁸F) fluorodeoxyglucose (FDG) uptake.

Implication for Patient Care

• ■ The association between the proinflammatory macrophage phenotype and enhanced glycolysis is specific to only certain microenvironmental stimuli (eg, immunoregulatory growth factors and cytokines, pathogen-derived factors, and tissue oxygenation level) rather than representing a universal and distinguishing feature of proinflammatory macrophages, which calls for a more prudent approach in the interpretation of ¹⁸F-FDG PET of vascular inflammation.