Absence of hexose‐6‐phosphate dehydrogenase results in reduced overall glucose consumption but does not prevent 11β‐hydroxysteroid dehydrogenase‐1‐dependent glucocorticoid activation

Hexose‐6‐phosphate dehydrogenase (H6PD) is thought to be the major source of NADPH within the endoplasmic reticulum (ER), determining 11β‐hydroxysteroid dehydrogenase 1 (11β‐HSD1) reaction direction to convert inert 11‐oxo‐ to potent 11β‐hydroxyglucocorticoids. Here, we tested the hypothesis whether H6pd knock‐out (KO) in primary murine bone marrow‐derived macrophages results in a switch from 11β‐HSD1 oxoreduction to dehydrogenation, thereby inactivating glucocorticoids (GC) and affecting macrophage phenotypic activation as well as causing a more aggressive M1 macrophage phenotype. H6pdKO did not lead to major disturbances of macrophage activation state, although a slightly more pronounced M1 phenotype was observed with enhanced proinflammatory cytokine release, an effect explained by the decreased 11β‐HSD1‐dependent GC activation. Unexpectedly, ablation of H6pd did not switch 11β‐HSD1 reaction direction. A moderately decreased 11β‐HSD1 oxoreduction activity by 40–50% was observed in H6pdKO M1 macrophages but dehydrogenation activity was undetectable, providing strong evidence for the existence of an alternative source of NADPH in the ER. H6pdKO M1 activated macrophages showed decreased phagocytic activity, most likely a result of the reduced 11β‐HSD1‐dependent GC activation. Other general macrophage functions reported to be influenced by GC, such as nitrite production and cholesterol efflux, were altered negligibly or not at all. Importantly, assessment of energy metabolism using an extracellular flux analyzer and lactate measurements revealed reduced overall glucose consumption in H6pdKO M1 activated macrophages, an effect that was GC independent. The GC‐independent influence of H6PD on energy metabolism and the characterization of the alternative source of NADPH in the ER warrant further investigations.

Hexose-6-phosphate dehydrogenase (H6PD) is thought to be the major source of NADPH within the endoplasmic reticulum (ER), determining 11bhydroxysteroid dehydrogenase 1 (11b-HSD1) reaction direction to convert inert 11-oxo-to potent 11b-hydroxyglucocorticoids. Here, we tested the hypothesis whether H6pd knock-out (KO) in primary murine bone marrowderived macrophages results in a switch from 11b-HSD1 oxoreduction to dehydrogenation, thereby inactivating glucocorticoids (GC) and affecting macrophage phenotypic activation as well as causing a more aggressive M1 macrophage phenotype. H6pd KO did not lead to major disturbances of macrophage activation state, although a slightly more pronounced M1 phenotype was observed with enhanced proinflammatory cytokine release, an effect explained by the decreased 11b-HSD1-dependent GC activation. Unexpectedly, ablation of H6pd did not switch 11b-HSD1 reaction direction. A moderately decreased 11b-HSD1 oxoreduction activity by 40-50% was observed in H6pd KO M1 macrophages but dehydrogenation activity was undetectable, providing strong evidence for the existence of an alternative source of NADPH in the ER. H6pd KO M1 activated macrophages showed decreased phagocytic activity, most likely a result of the reduced 11b-HSD1dependent GC activation. Other general macrophage functions reported to be influenced by GC, such as nitrite production and cholesterol efflux, were altered negligibly or not at all. Importantly, assessment of energy metabolism using an extracellular flux analyzer and lactate measurements revealed reduced overall glucose consumption in H6pd KO M1 activated macrophages, an effect that was GC independent. The GC-independent influence of H6PD on energy metabolism and the characterization of the alternative source of NADPH in the ER warrant further investigations.

Introduction
Glucocorticoids (GC) essentially regulate major physiological functions, with potent modulatory effects on the immune system [1]. Besides the adrenal GC synthesis, the tissue-and cell-specific interconversion of active 11b-hydroxy GC [cortisol in human, corticosterone (CORT) in rodents] and inactive 11-oxo GC [cortisone in human, 11-dehydrocorticosterone  in rodents] by 11b-HSD enzymes plays an important role in mediating GC effects [2]. Two isozymes are known, 11b-HSD2 converts 11b-hydroxy to 11-oxo GC and 11b-HSD1 catalyzes in vivo the reverse reaction. 11b-HSD1 is a bidirectional enzyme facing the endoplasmic reticulum (ER) lumen, where it interacts with hexose-6-phosphate dehydrogenase (H6PD) [3][4][5][6]. H6PD utilizes glucose-6-phosphate (G6P) to produce NADPH in the ER [7,8]. As the ER membrane is not permeable for pyridine nucleotides, the intraluminal regeneration of NADPH by H6PD determines 11b-HSD1 to predominantly catalyze the oxoreduction of 11-oxo GC [4,9]. Upon disruption of the cell membrane in homogenized cells or tissue, the intraluminal NADPH is rapidly metabolized and in the absence of G6P this results in a loss of oxoreduction activity. A switch from oxoreduction to dehydrogenation activity was observed when comparing 11b-HSD1 activity in liver microsomes of wild-type (WT) and H6pd knock-out (KO) mice, and the urinary excretion revealed a dramatic shift in the 11-oxo/11bhydroxy GC ratio [10]. H6PD is currently the only well characterized enzyme within the ER that generates NADPH and it is thought to be responsible for maintaining a high luminal NADPH/NADP + ratio. Therefore, we anticipated that in cells from H6pd KO mice 11b-HSD1 would catalyze predominantly the dehydrogenation reaction, thereby inactivating GC.
Glucocorticoids affect cells of the immune system and impact gene expression, enzymatic activity, and cellular differentiation and migration [11]. GC are, for example, crucial for T-cell selection by regulating apoptosis. In macrophages, they play a pivotal but not yet fully elucidated role in the phenotypic activation [12], and they suppress the expression and secretion of proinflammatory cytokines such as interleukin-1b (IL-1b), interleukin-6 (IL-6), interferon-c (IFNc), and tumor necrosis factor-a (TNF-a), thereby preventing an overshooting immune response [13][14][15][16]. Blood monocytes, derived from bone marrow progenitor cells, are constantly patrolling the blood vessels, fulfilling various functions in both innate and adaptive immunity. If activated by tissue damage or infection, they enter the affected tissue and undergo phenotypic activation. Depending on the type of activator, the classically activated M1 and the alternatively activated M2 macrophages are distinguished. IFNc, TNF-a, or lipopolysaccharide (LPS) induce a M1 phenotype producing proinflammatory cytokines and bactericidal mediators, whereas interleukin-4 (IL-4) or interleukin-13 (IL-13) result in an M2 phenotype contributing to the downregulation of inflammation and elimination of tissue debris [17][18][19][20][21].
The expression of 11b-HSD1 in patrolling monocytes is absent or very low, but strongly induced upon activation toward a M1 phenotype [22]. It has been shown that 11b-HSD1 is regulated dynamically during an ongoing inflammation, which is crucial for a controlled and rapid resolution of inflammation since GC restrain the M1 activated macrophages and promote a M2 phenotype [23,24]. Additionally, an increased 11b-HSD1 oxoreduction activity was hypothesized to cause a re-routing of the H6PD substrate G6P toward the ER, causing a change in overall glucose metabolism that has been shown to alter macrophage polarization [25,26].
In the present study, we investigated whether in bone marrow-derived macrophages (BMDM) from H6pd KO mice 11b-HSD1 would predominantly function as a dehydrogenase, thereby inactivating GC and resulting in enhanced expression of proinflammatory cytokines. Furthermore, we tested whether the decreased 11b-HSD1-dependent GC activation due to absence of H6PD activity would affect the activation of macrophage phenotypes and their functions. We found that ablation of H6pd did not cause a loss of GC activation by 11b-HSD1 but only a reduction by 40-50%, which translated into a decreased suppression by 11-DHC of proinflammatory cytokine expression. General macrophage functions were only moderately influenced and the H6pd KO only led to slightly exaggerated M1 activation state with a marginally increased proinflammatory cytokine release. However, the KO of H6pd did cause an overall increased glucose consumption and lactate secretion, most likely by a GC-independent mechanism.

Decreased glucocorticoid activation in H6pd KO M1 macrophages
Based on earlier observation of an almost complete loss of 11b-HSD1 oxoreduction activity in liver microsomes isolated from H6pd KO mice [10], we expected a loss of the oxoreduction activity in BMDM from H6pd KO mice and that the enzyme would predominantly exhibit oxidation activity. To test this assumption, WT and H6pd KO M1 macrophages were incubated with 100 nM 11-DHC for 24 h, followed by determination of the amount of CORT released into the medium by ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/ MS). Surprisingly, compared to WT, H6pd KO macrophages still retained 50-60% of the oxoreduction activity (Fig. 1A). Treatment of the macrophages with the 11b-HSD1 inhibitor BNW16 [27] fully abolished the oxoreduction activity in both WT and H6pd KO macrophages, confirming that this activity was indeed 11b-HSD1-specific.
Diminished suppressive effect of 11-DHC on proinflammatory cytokine expression in H6pd KO M1 macrophages To assess the functional impact of the altered 11b-HSD1 activity, macrophages were activated toward an M1 phenotype for 24 h followed by 24 h incubation with increasing concentrations of 11-DHC, which requires oxoreduction by 11b-HSD1 to CORT in order to exert a suppressive effect on the expression of proinflammatory cytokines. Hence, the decrease in proinflammatory cytokine mRNA expression can serve as a functional readout of the 11b-HSD1 oxoreduction activity. Unlike initially anticipated, only moderate effects were observed in the H6pd KO macrophages. While treatment with 11-DHC suppressed the expression of all four cytokines measured, only Mcp-1 and Il-1b mRNA expression in WT macrophages tended to show a more pronounced decrease at 100 and 1000 nM 11-DHC compared with H6pd KO macrophages, in line with higher 11b-HSD1 oxoreduction activity in WT macrophages (Fig. 1B,C). Preincubation of cells for 24 h with 100 nM 11-DHC followed by activation into M1 phenotype with LPS+IFNc did not result in significant changes of Il1-b, Mcp1, Il-6, and Tnfa mRNA expression when comparing either pretreated WT and KO cells or when comparing pretreated and untreated cells at 1, 2, 4, 8, or 24 h after stimulation (data not shown).

Impact of H6pd KO on macrophage phenotypes
To examine whether the observed difference in 11b-HSD1 activity and the reduced ability of the H6pd KO macrophages to decrease proinflammatory markers has an impact on macrophage phenotypic activation, the mRNA expression levels of phenotypespecific marker genes and other genes of interest were qualitatively assessed in M0, M1, and M2-polarized macrophages from 8-to 12-week-old mice by real-time PCR. Besides H6pd, which is clearly expressed in WT but not H6pd KO macrophages, the phenotype markers indicated a trend toward a more pronounced M1 phenotype in H6pd KO macrophages ( Fig. 2A). The overall phenotypic activation seemed to be maintained; the M1 macrophages highly expressed all M1-specific markers, for example, iNos, Mcp1, Il1-b, Tnfa, and Il-6, whereas the M2 macrophages expressed M2-specific markers such as Cd206, Ym1, and Fizz1. This shows that the absence of H6pd had only a minor impact on macrophage phenotypic activation. The glucocorticoid receptor (GR) was highly expressed in all phenotypes, whereas 11b-hsd1 mRNA was upregulated in the M1 activation state but hardly expressed in M0 and M2 macrophages. Therefore, further experiments focused on M1 activated macrophages. A time-course covering the initial burst phase was performed in WT and H6pd KO M1 macrophages and their M1 marker expression levels were compared at different time-points ( Fig. 2B-E). Although Tnfa and Il-1b mRNA expression levels showed a trend toward increased expression in H6pd KO compared to WT macrophages, they were not significantly altered at any time point. Similar results were obtained when comparing non-elicited peritoneal macrophages isolated from WT and H6pd KO mice activated toward an M1 phenotype, where only Tnfa showed a trend toward an increase in H6pd KO compared to WT (data not shown).
To assess the biological relevance of the observed trends in some M1 markers, secreted protein levels were measured. Therefore, M0 cells were incubated with LPS and IFNc for 4 h to induce the M1 activation state, followed by analysis of cell supernatants by ELISA. This revealed an approximately 1.4-fold higher secretion of monocyte chemoattractant protein-1 (MCP-1), IL-6 and TNFa proteins in H6pd KO compared to WT macrophages (Fig. 2F).

Impact of H6pd KO on general properties of M1 macrophages
Next, the impact of H6pd KO on typical M1 macrophage functions, such as phagocytosis, nitrite production, and cholesterol efflux, was examined. Nitrite production, mainly mediated by the inducible nitric oxidase (iNOS), was measured using the Griess assay. No nitrite production could be detected in supernatants of M0 and M2 cells, which is in agreement with the background levels of iNos mRNA expression in these phenotypes; however, it was well detectable in M1 macrophages, with a small but statistically significant decrease in H6pd KO cells (Fig. 3A). The  (Fig. 3B), revealed a significantly reduced phagocytic index in H6pd KO M1 macrophages (Fig. 3C). Cholesterol efflux, another important function of macrophages, was measured in all phenotypes using [ 3 H]-cholesterol and was calculated as the percentage of radioactivity in the medium from total radioactivity in medium plus cells. No difference could be detected between WT and H6pd KO macrophages (Fig. 3D). Thus, H6pd KO decreased phagocytic activity and had a modest effect on nitrite production, but did not affect cholesterol efflux.

Decreased glycolytic capacity in H6pd-deficient M1 macrophages
To investigate whether H6pd deficiency affects metabolic parameters that affect macrophage function, a glycolysis stress test was performed using a Seahorse XFp analyzer. Thereby, compounds targeting key elements of cellular metabolism are sequentially added at indicated time points. In the first injection, glucose was added as an energy source. Next, oxidative phosphorylation in the mitochondria was shut down by adding oligomycin and forcing the cells to cover their energy demand exclusively by glycolysis. Finally, 2deoxyglucose was added to inhibit glycolysis by blocking glucose hexokinase and verifying glycolysis as a major source of the previously measured basal extracellular acidification rate (ECAR). During this procedure, the ECAR and oxygen consumption rate (OCR) are measured in a transient microchamber.
The ECAR measurements showed a decreased baseline glycolytic activity in H6pd KO compared to WT M1 macrophages (Fig. 4A-C). In both WT and H6pd KO macrophages, the glycolytic reserve, indicated by the increase in ECAR upon addition of oligomycin, was low, pointing toward a negligible role of mitochondrial energy production. The OCR was also lower in H6pd KO compared to WT M1 macrophages under glucose deprivation (Fig. 4D-F). This difference was less pronounced upon addition of glucose, whereby the H6pd KO macrophages seemed to have lost the responsiveness to glucose addition. The nonresponsiveness in the OCR measurements is characteristic of M1 activated macrophages and represents a state of inhibited mitochondrial oxidative respiration [28]. To confirm the decreased acidification rates of the H6pd KO macrophages measured in the Seahorse analyzer by another approach, lactate levels were determined using a colorimetric assay kit. M0 macrophages of WT and H6pd KO were activated toward an M1 phenotype while supernatant samples were taken at 4, 24, and 48 h. The amount of lactate produced in H6pd KO macrophages showed a weak trend toward a decrease after 24 h, which became significant after 48 h (Fig. 4G).

Discussion
In this study, we showed that the loss of H6pd in macrophages unexpectedly did not abolish 11b-HSD1dependent conversion of 11-DHC to CORT but resulted in a rather moderate decrease of oxoreduction activity of only 40-50%. Dehydrogenation activity could neither be detected in intact WT nor in H6pd KO macrophages. The fact that there is no switch from 11-oxoreduction to dehydrogenation in H6pd KO macrophages as well as the moderate decrease in oxoreduction activity is surprising and might be explained by a compensatory adaptation involving a yet unknown NADPH generating system in the ER. Results from HEK-293 cells transfected with H6pd compared to mock transfected cells and data from HT-1080 connective tissue cells where H6pd was knocked down using siRNA showed a switch from 11b-HSD1 oxoreduction to dehydrogenation activity when H6pd was absent or knocked down [3,4,29]. These cell types might not express the alternative NADPH generating system present in the ER of macrophages. Also, studies with liver microsomes from H6pd KO mice showed an almost complete loss of oxoreduction but high dehydrogenation activity [10]. Notably, NADPH is not freely passing the ER membrane and during preparation of microsomes, the NADPH in the luminal compartment is used up because of the absence of the substrate for the luminal NADPH regenerating system (i.e. G6P for H6PD or the substrate of the yet unknown luminal regenerating system).
Currently, H6PD represents the only well characterized enzyme generating NADPH in the ER and based on the observation of an almost complete loss of 11b-HSD1 oxoreduction activity in liver microsomes from H6pd KO mice it was suggested that H6PD is the major source of NADPH within the ER [10]. Our observation of a rather moderate effect of H6pd KO on 11b-HSD1 oxoreduction activity in intact macrophages indicates the existence of an alternative yet less efficient source for NADPH in the ER of these cells. The distinct effect of H6PD deficiency in macrophage, HT-1080 and HEK-293 cells suggests a cell-type specific expression of such an alternative NADPH regenerating system in the ER. Earlier work provided evidence for the existence of a pentose phosphate pathway in the ER [30], whereby H6PD was shown to catalyze the first two steps of this pathway, converting G6P to 6-phosphogluconate [31]. There is evidence from work on liver microsomes for the existence of a 6-phosphogluconate dehydrogenase in the ER that produces NADPH ( [32,33], own observations); however, the corresponding gene remains to be identified. Nevertheless, as 6-phosphogluconate is not freely permeable across the ER membrane, this enzyme unlikely accounts for the observed remaining 11b-HSD1-dependent oxoreduction activity. The unknown source of NADPH in the ER needs to be uncovered, and H6pd KO macrophages constitute an interesting model for studying such mechanisms.
Interestingly, the absence of H6pd did not grossly interfere with macrophage activation into M1 and M2 phenotypes. Due to the moderate effect of H6pd KO on 11b-HSD1 oxoreduction activity, the M1 phenotype was barely affected, with only weak trends toward an increased expression of some M1 markers and an approximately 1.4-fold increased release of proinflammatory cytokines, most likely because of lower production of active GC. Considering the great inducibility of macrophage cytokine expression and secretion, these minor changes are unlikely to be of biological relevance. A more strongly enhanced release of proinflammatory cytokines was reported in 11b-HSD1-deficient mice [23,34]. Whereas M1 macrophages from H6pd KO mice showed comparable 11b-HSD1 expression to WT cells, the corresponding oxoreduction activity was reduced. This is in agreement with the above-mentioned regulation of 11b-HSD1 activity via cosubstrate availability. Thus, regarding GC activation the alternative NADPH source could not entirely compensate for the loss of H6pd, resulting in a weak trend toward a more pronounced M1 phenotype.
Other general macrophage functions, including cholesterol efflux and nitrite production, were not changed or were statistically significant but not considered biologically relevant, respectively. In the H6pd KO macrophages, the moderate effect on 11b-HSD1 was not sufficient to affect cyclodextrin-stimulated cholesterol efflux. This may seem surprising because in macrophages from 11b-hsd1-deficient mice an increase in ApoAI-stimulated cholesterol efflux was reported [35]. A possible explanation may be that overall cholesterol efflux was measured using cyclodextrin as cholesterol acceptor, which is not pathway specific, whereas ApoAI exclusively stimulates ABCA1-dependent efflux.
Furthermore, it was shown that GC promote phagocytic activity of macrophages and that the higher phagocytic activity of macrophages upon treatment with 11-DHC is dependent on 11b-HSD1 [23]. 11b-HSD1 was found to promote the rapid clearance of apoptotic leukocytes during the resolution of inflammation, whereby phagocytic clearance was delayed in macrophages from 11b-hsd1 KO mice. Whether the observed decrease in phagocytic activity of H6pd KO macrophages, most likely a consequence of the lower GC activation due to the limited supply of NADPH as 11b-HSD1 substrate, results in delayed phagocytic clearance and resolution of inflammation in vivo remains to be investigated.
Besides its role in GC activation, H6PD may affect other metabolic functions through catalyzing the first two steps of the pentose phosphate pathway in the ER [36]. The present analysis of the metabolism of M1 macrophages showed that they cover their energy demand almost exclusively through aerobic glycolysis. This is in agreement with the generally accepted Warburg-like metabolism found in dendritic cells and in macrophages upon LPS stimulation [37]. Noteworthy, the H6pd KO macrophages showed a markedly decreased overall glucose consumption, further consolidated by a decreased lactate concentration in cell supernatants. Such a decrease was also reported in cancer cells upon knock-down of H6pd and attributed to a H6PD-dependent glucose utilization pathway [38]. Additional support for the contribution of such a pathway comes from the increased OCR under glucose starvation observed in WT macrophages in the present study. The effect of using oxygen despite being primed for aerobic glycolysis has also been reported for T cells under glucose-starved conditions [39]. Since the H6pd KO macrophages show less oxygen consumption under these conditions, they seem to be less affected by the limited availability of glucose due to their reduced overall glucose consumption.
In conclusion, we showed that H6pd KO does not lead to major disturbances of macrophage phenotypic activation and that it results in only a moderate decrease in 11b-HSD1-mediated GC activation in M1 macrophages and not to a switch of this enzyme to primarily catalyze the oxidation reaction. H6pd KO M1-polarized macrophages showed only a slight trend toward a more aggressive M1 phenotype but displayed a significantly reduced phagocytic activity, in agreement with decreased GC activation. Importantly, H6pd KO M1 macrophages showed decreased glucose utilization and lactate production rates, most likely by a GC-independent mechanism. The results indicate a modulatory role of H6PD in 11b-HSD1-dependent GC activation as well as a GC-independent role in regulating energy metabolism.

Generation of bone marrow-derived macrophages and activation to M1 and M2 phenotype
Bone marrow-derived macrophages were obtained from femurs of female C57BL/6 WT and H6pd KO mice at the age between 10 and 12 weeks as described previously [40]. The H6pd KO mouse line was kindly provided by Gareth Lavery, University of Birmingham, UK [10]. Mice were housed in climate-controlled facility under standard conditions on a 12 h light/12 h dark cycle with free access to standard chow and drinking water. The mice were sacrificed by CO 2 , followed by isolation of BMDM. All procedures were approved by the Cantonal Veterinary Office in Basel, Switzerland. After cervical dislocation both femurs were dissected and the epi- supplemented with 10% FBS (Connectorate, Dietikon, Switzerland), 100 UÁmL À1 penicillin, 100 lgÁmL À1 streptomycin, followed by distribution into eight 6-cm bacterial dishes containing 5 mL of the same medium supplemented with 10 ngÁmL À1 macrophage colony-stimulating factor (BioLegend, San Diego, CA, USA). Cells were then incubated at 5% CO 2 and 37°C in order to induce differentiation into BMDM. After 3 days, 5 mL of new medium was added. Cells were harvested at day 7 by scraping. Phenotypic activation of BMDM was induced by supplementing the differentiation medium with 100 ngÁmL À1 LPS and 50 ngÁmL À1 IFNc for M1 phenotype, and 10 ngÁmL À1 IL-4 (cytokines were from Sigma-Aldrich) for M2 phenotype. The incubation time was 4 h in case of cytokine secretion measurements and 24 h for functional assays.  (Table 1) were designed using the IDT SciTool for real-time PCR and synthesized by MicrosynthÒ (Balgach, Switzerland).

Griess assay
Free nitrite was measured in culture supernatants using Griess reagent (Sigma-Aldrich). BMDM were seeded at a density of 0.5 9 10 6 cells/well in a non-tissue culture treated 12-well plate (Starlab, Switzerland) and activated toward M1 or M2 phenotypes or left in their unactivated state in phenol red-free medium for 48 h. A quantity of 100 lL of supernatant was transferred in triplicates onto a 96-well plate. Then, 100 lL of Griess reagent was added, the plate incubated for 5 min at 37°C, followed by measuring the absorbance at 540 nm. The results were normalized to total protein content, determined by using the Pierce BCA protein assay according to the manufacturer's protocol (Thermo Fisher Scientific, Waltham, MA, USA).

Determination of phagocytic index
The phagocytic index was determined as described previously [42]. Briefly, polystyrene latex beads with 3 lm mean particle size (Sigma-Aldrich) were opsonized with anti-goat IgG (Sigma-Aldrich) and labeled with donkey anti-goat IgG ALEXA-594 (Life Technologies, Zug, Switzerland). BMDM were seeded at a density of 0.5 9 10 5 cells/well on a 96-well plate and left in M1 differentiation medium for 24 h. A quantity of 100 lL of opsonized and labeled beadsolution was added to each well and the plate was

Measurement of cholesterol efflux
Cholesterol efflux was measured as described previously [43]. Briefly, BMDM were seeded at a density of 0.4 9 10 6 cells/well of a 12-well plate and either activated toward an M1 or M2 phenotype or left in the M0 state. Cells were incubated in medium containing 0.5 lCiÁmL À1 [ 3 H]-cholesterol (PerkinElmer, Boston, MA, USA) for 48 h, the medium was removed and cells were washed twice with prewarmed PBS. Serum-free medium was added for 18 h to allow equilibration of the labeled cholesterol between the cellular compartments. Cells were washed and incubated with new serum-free medium in the absence or presence of 1 mgÁmL À1 cyclodextrin (Sigma-Aldrich) as a cholesterol efflux acceptor for 4 h. The [ 3 H]-cholesterol efflux was then quantified by separate liquid scintillation counting of supernatant and cell lysate and reported as the ratio of radiolabeled cholesterol in the medium compared to that of medium and cells.

ELISA
Bone marrow-derived macrophages were seeded at a density of 0.5 9 10 6 cells/well in a 12-well plate and allowed to settle for 24 h. Then, BMDM were activated toward an M1 phenotype for 4 h, followed by collecting supernatants, and measuring TNFa, IL-1b, and MCP-1 levels using an ELISA Ready-SET-Go kit according to the manufacturer's protocol (eBioscience, San Diego, CA, USA). The values were normalized to total protein.

Seahorse assay
Bone marrow-derived macrophages were seeded at a density of 0.5 9 10 5 cell/well in a XFp plate (Agilent Technologies, Basel, Switzerland) and activated by addition of M1 medium for 24 h. Then, ECAR and OCR were measured as suggested by the manufacturer's protocol for the glycolysis stress test (Agilent Technologies). Briefly, the activated cells were equilibrated in XF base medium (Agilent Technologies) supplemented with 2 mM L-glutamine (Sigma-Aldrich) and placed in a CO 2 -free incubator at 37°C for 1 h prior to the assay. Then, glucose, oligomycin and 2deoxyglucose (contained in the kit) were sequentially injected in that order during continuous oxygen and H + measurements. Data were normalized to cell protein content.

Lactate assay
Lactate was determined in supernatants of M1 macrophages cultured in a 12-well plate at a density of 0.5 9 10 6 cells/well. Supernatants were collected by snapfreezing 50 lL of sample at each time-point. The samples were measured using the L-lactate assay (Abcam, Cambridge, UK) according to the manufacturer's protocol.
Results were normalized to total protein.

Statistical analysis
Statistical analyses were conducted using GRAPHPAD PRISM 5.0 software. To analyze the difference between WT and KO, first, the normal distribution of the data was determined using the Shapiro-Wilk normality test. For nonparametric distributed data, the Kruskal-Wallis test followed by the Dunn's posttest was applied. Parametric distributed data were analyzed using an unpaired t-test or one-way ANOVA. Values represent mean AE SD.