Adherence Affects Monocyte Innate Immune Function and Metabolic Reprogramming after Lipopolysaccharide Stimulation In Vitro

Adherence is an important first step of monocyte extravasation and differentiation to macrophages. When monocytes exit the blood vessel, adherence to the capillary endothelium and, subsequently, to a variety of extracellular matrix proteins occurs (1, 2). The importance of monocyte adherence for phenotype and function was already shown three decades ago. Several articles described substantial changes in gene expression of certain cytokines and growth factors upon adherence (3–7), as well as alterations in surface marker expression (8, 9). However, in more recent years, the influence of adherence on monocyte function has received less attention. The vast majority of recent studies involving monocyte function has been performed under adherent conditions, but this might not be optimal to study monocyte involvement in systemic diseases or syndromes. One clear example of such a systemic syndrome is sepsis, a complex condition characterized by a dysregulated host response to an infection resulting in organ dysfunction (10). Sepsis is a major global health problem with an estimated 48.9 million incident cases recorded worldwide and 11 million sepsis-related deaths in 2017, representing a fifth of all global deaths that year (11). Investigating functional differences between adherent and nonadherent monocytes upon activation may provide some insight into the phenotype of circulating monocytes and, therefore, the pathophysiology of this and other systemic diseases, although caution is warranted to translate in vitro findings to responses in humans in vivo.

In the last decade, the importance of cellular metabolism for immune cell phenotype and function has been established (12, 13). Cells with different immunological functions use distinct metabolic pathways to generate the amount of energy and biosynthetic intermediates required for their function. Generally, proinflammatory responses are associated with a shift toward the oxygen-independent glucose metabolism pathway called glycolysis. In contrast, anti-inflammatory responses are associated with oxygen-dependent energy metabolism pathways, including the tricarboxylic acid (TCA) cycle and oxidative phosphorylation (OXPHOS). Although originally studies mainly focused on lymphocytes, more recent research has pointed at a key role for cellular metabolism in myeloid cell function (12). In this regard, macrophages were reported to alter their metabolic profile in response to LPS, a proinflammatory component of the Gram-negative bacterial cell wall and ligand to TLR4, in a way that depended on their origin (14). Bone marrow–derived macrophages stimulated with LPS responded with a profound upregulation of glycolysis and downregulation of OXPHOS, whereas peritoneal macrophages showed upregulation of both glycolysis and OXPHOS. This suggests that differences in metabolic reprogramming exist, even in closely related cells. Human monocytes have been shown to induce glycolysis upon LPS stimulation (15, 16); however, these studies used adherent monocytes. The influence of adherence on metabolic reprogramming of monocytes is still unknown. The aim of our study was to elucidate the effect of adherence on the functional and metabolic response of monocytes to LPS.

Healthy volunteers for blood sampling were recruited in accordance with a study protocol that was reviewed by the Academic Medical Center Medical Ethical Committee (no. 2015_074). The need for ethical approval was waived. Prior to sample donation, all donors gave informed consent. Heparinized blood was diluted (1:1) in PBS. Isolation of PBMCs was performed by density-gradient centrifugation with Ficoll-Paque PLUS (GE Healthcare, Chicago, IL). CD14⁺ monocytes were subsequently purified using MACS CD14 microbeads for positive selection, according to the manufacturer’s instructions (Miltenyi Biotec, Bergisch Gladbach, Germany). All isolation procedures were started within 30 min after blood collection. Monocyte purity was analyzed at random (∼25% of all samples) by flow cytometry and was always higher than 95%.

Monocytes were plated in 48-well, tissue culture–treated plates for adherent monocytes or 48-well plates with cell-repellent surface for nonadherent monocytes (both 5 × 10⁵ cells per well; Greiner Bio-One, Kremsmünster, Austria). Monocytes were cultured in RPMI 1640 medium (no glutamine, 31870074; Life Technologies; Thermo Fisher Scientific, Waltham, MA) supplemented with 10 μg/ml gentamicin (Life Technologies), 1 mM sodium pyruvate (Life Technologies), 2 mM GlutaMAX (Life Technologies), 20 mM HEPES (Life Technologies), and 10% FBS (HyClone; GE Healthcare) or glucose-free RPMI 1640 (no glucose, with glutamine, 11879020; Life Technologies) supplemented with gentamicin, sodium pyruvate, HEPES, and FBS. After 1 h of incubation, monocytes were stimulated for 24 h with 10 ng/ml ultrapure LPS (from Escherichia coli 0111:B4; InvivoGen, Toulouse, France). To measure the production of reactive oxygen species (ROS), monocytes were cultured for 20 h prior to a 2.5-h stimulation with heat-killed Candida albicans (UC820, generously provided by Dr. L. Joosten, Radboud University Medical Center [UMC], Nijmegen, the Netherlands) or heat-killed Klebsiella pneumoniae (American Type Culture Collection 43816) at the equivalent of 5 × 10⁵ CFU/ml and subsequently treated with carboxy-H₂DCFDA (Invitrogen; Thermo Fisher Scientific). To measure phagocytic capacity, monocytes were cultured for 20 h prior to a 2.5-h incubation with 0.5 mg/ml pHrodo Red E. coli BioParticles (Invitrogen). To limit glycolysis, monocytes were treated with 2 mM 2-deoxy-d-glucose (2DG; Sigma-Aldrich) or cultured in glucose-free medium. The glucose-free medium was supplemented with all the supplements above, including 10% FBS, resulting in a concentration of ∼0.3 mM glucose.

Supernatants were stored at −20°C until cytokine measurements were performed. TNF, IL-1β, IL-10, and IL-6 were measured using commercially available ELISAs according to the protocol supplied by the manufacturer (R&D Systems, Minneapolis, MN).

Surface marker expression was analyzed by flow cytometry. Monocytes were washed twice with PBS and then incubated for 20 min with TrypLE Select (Life Technologies) to detach cells. After a wash with medium, monocytes were suspended in FACS buffer (5% BSA, 0.35 mM EDTA, and 0.01% NaN3) and incubated with fixable viability dye eFluor 780 (Invitrogen), mouse anti-human CD14 (clone M5E2), mouse anti-human CD16 (clone 3G8), mouse anti-human CD18 (clone 6.7), mouse anti-human CD11b/Mac-1 (clone ICRF44), and mouse anti-human HLA-DR (clone G46-6) (all from BD Biosciences, Franklin Lakes, NJ). Flow cytometry was performed using an FACSCanto II (BD Biosciences), and data were analyzed using FlowJo software (BD Biosciences).

Lactate was quantified using an enzymatic assay, as described before (16). Briefly, lactate was oxidized by lactate oxidase (Sigma-Aldrich, Saint Louis, MO) and the resulting H₂O₂ was coupled to the conversion of Amplex Red (Life Technologies, Carlsbad, CA) reagent to fluorescent resorufin by HRP (Sigma-Aldrich). Samples were diluted 200 times in PBS and incubated for 20 min at room temperature. Fluorescence was measured using a 96-well plate reader (BioTek Instruments, Winooski, VT). To quantify glucose levels, lactate oxidase was replaced by glucose oxidase (Sigma-Aldrich). To calculate the glucose consumption, the measured values were subtracted from the reference value (medium without cells).

Viability was assessed by CellTiter-Blue assay (Promega, Madison, WI) according to the protocol supplied by the manufacturer.

For glucose-labeling experiments, glucose-free RPMI media was supplemented with 5.55 mM [U-[¹³C]]-glucose (Cambridge Isotope Laboratories, Tewksbury, MA). Samples were separated in two phases as described previously (17, 18). Briefly, the aqueous phase was evaporated, and the metabolite residue was reconstituted in 100 μl methanol/water (6:4; v/v). Metabolites were analyzed using a Waters ACQUITY Ultra-HPLC System containing a SeQuant ZIC-cHILIC column (100 × 2.1 mm, 3-mm particle size; Merck, Darmstadt, Germany) coupled to a Bruker Impact II mass spectrometer (Bruker Daltonic, Bremen, Germany). Data were analyzed using Bruker TASQ 2.1.22.3 software (Bruker Daltonic). [¹³C] enrichment was calculated based on mass distribution isotopomer analysis; results were corrected for their natural [¹³C] abundance by solving the associated set of linear equations for lactate using nonnegative least squares (19).

RNA was isolated using the NucleoSpin RNA Isolation Kit according to the protocol supplied by the manufacturer (Macherey-Nagel, Düren, Germany). Total RNA concentrations were measured by Qubit RNA HS Assay Kit (Thermo Fisher Scientific). RNA integrity was measured using the Agilent 2100 Bioanalyzer instrument (Agilent, Santa Clara, CA); samples with RNA integrity number >7 were included for RNA sequencing (RNAseq) analysis. Stranded cDNA libraries were prepared from 100 to 200 ng total RNA per sample using the KAPA mRNA HyperPrep Kit (Roche Diagnostics, Bazel, Switzerland). Subsequently, 3 nM input per sample was sequenced on an Illumina HiSeq 4000 instrument (Illumina, San Diego, CA) to a depth of ∼40 million single-ended, 50-bp reads. For RNAseq, material from the first six donors were used (primarily to control costs).

The sequence read quality was assessed using FastQC methods (version 0.11.5; Babraham Institute, Babraham, Cambridgeshire, U.K.). Trimmomatic version 0.32 (20) was used to trim the Illumina adapters and filter low-quality reads and ambiguous nucleotide-containing sequences. Low-quality leading (3-nt) and trailing (3-nt) bases were removed from each read. A sliding window trimming using a window of four and a Phred score threshold of 15 nt was used to assess the quality of the reads. After preprocessing, the remaining high-quality reads were aligned against the Genome Reference Consortium Human Genome Build 38 patch release 7 using Bowtie2 version 2.3.4.3 (21) with default parameters. Count data were generated by means of the FeatureCounts method (22) and differential expression analyzed using the DESeq2 method (23) in the R statistical computing environment (R Core Team 2014. R: A language and environment for statistical computing; R Foundation for Statistical Computing, Vienna, Austria). Sequence libraries are publicly available through the National Center for Biotechnology Information Gene Expression Omnibus under the following accession number: GSE161839 (https://www.ncbi.nlm.nih.gov/geo).

Throughout significance was calculated using Benjamini–Hochberg (BH)–adjusted p values (24). Gene set enrichment analysis (GSEA) approach (25) was applied to determine the level and direction of enrichment for each pathway. For each pathway, GSEA calculates a normalized enrichment score (NES), this is the normalized version of the enrichment score, which reflects the degree to which a gene set is overrepresented by up- or downregulated genes. To determine the differences between the response to LPS on gene transcription in nonadherent and adherent monocytes, an interaction model was fitted for each gene of the indicated pathways: expression = β₁ donor identifier (ID) + β₂ adherence condition + β₃ LPS condition + β₄ adherence condition/LPS condition, where donor ID is a categorical variable indicating the donor, adherence condition is a binary variable indicating the adherence condition, LPS condition is an binary variable indicating the LPS condition, and adherence condition/LPS condition is the binary variable indicating the interaction between adherence and LPS. From this, we tested if a gene’s expression is influenced by an interaction between LPS and adherence (null hypothesis: there is no interaction between adherence and LPS, β₄ = 0). To determine if a gene exhibited a different transcriptional response to different glucose conditions, a likelihood-ratio χ² test was carried out for each gene (comparing the following model: expression − donor ID + glucose condition versus the nested model: expression − donor ID).

The cytokine-and-chemokine pathway was compiled by selecting all genes with the string IL, CCL, CXC, and TNF in their gene symbol and then checked for applicability. All other pathways are based on the Reactome database (26): TLR4 signaling cascade (TLR4 cascade; R-HSA-166016), ROS production (ROS and reactive nitrogen species [RNS] production in phagocytes; R-HSA-1222556), glycolysis (R-HSA-70171), lipid metabolism (metabolism of lipids; R-HSA-556833), TCA cycle, and OXPHOS (the citric acid cycle and respiratory electron transport; R-HSA-1428517). The M1 and M2 marker genes were selected based on previous publication (27).

In all experiments, nonadherent and adherent monocytes from the same donor were analyzed in a paired manner. Nonnormally distributed variables were analyzed using the nonparametric Wilcoxon test. Normally distributed variables were analyzed using the paired t test or a paired two-way ANOVA with Sidak multiple comparisons test in case of three or more groups. Analyses were done using GraphPad Prism version 8 (GraphPad Software, San Diego, CA).

To identify the differences between adherent monocytes and nonadherent monocytes from the same donors, we made use of different culturing techniques. To keep monocytes under nonadherent conditions, we cultured them in cell-repellent culture plates, whereas monocytes were let to adhere in tissue culture–treated plates. These different culturing conditions did not affect viability (Supplemental Fig. 1A); however, surface expression of certain markers was altered. As early as 20 h after plating, adherent monocytes showed enhanced expression of CD16 and reduced expression of CD18, CD11b/Mac1, and HLA-DR when compared with nonadherent monocytes (Supplemental Fig. 2). To determine the effect of adherence on cytokine production capacity, we stimulated nonadherent and adherent monocytes for 24 h with LPS. Visually, we found that LPS stimulation resulted in some monocyte aggregate formation under nonadherent conditions (Supplemental Fig. 1B). On a functional level, we found that adherent monocytes secreted less TNF and IL-1β (proinflammatory cytokines) and more IL-10 (anti-inflammatory cytokine), whereas IL-6 production was unaffected by adherence (Fig. 1A). Although interindividual variation in monocyte responses was evident, the direction per cytokine was consistent, and therefore, differences were significant. Cytokines were very low or undetectable in nonstimulated medium controls and not different between nonadherent and adherent monocytes. To further assess the functional differences induced by adherence, we studied the capacity to phagocytose E. coli particles and ROS production of monocytes. Adherence increased the phagocytic capacity of monocytes (Fig. 1B) and ROS production in response to heat-killed K. pneumoniae and heat-killed C. albicans (Fig. 1C). LPS did not induce ROS production in monocytes.

To obtain more insight in the effect of adherence on monocytes, we compared the transcriptional responses between LPS-stimulated adherent and nonadherent monocytes by RNAseq analysis (Supplemental Fig. 3). We found that LPS stimulation resulted in the upregulation of 5997 genes and the downregulation of 5538 genes in nonadherent monocytes and the upregulation of 5872 genes and the downregulation of 5518 genes in adherent monocytes. To further investigate the effect of adherence on monocyte immune functions, and partially guided by the results presented in Fig. 1A–C, we focused on genes related to the TLR4 signaling cascade (Fig. 1D), cytokines and chemokines (Fig. 1E), and ROS and RNS production (Fig. 1F). Consistent with enhanced release of the proinflammatory cytokines TNF and IL-1β upon LPS stimulation (Fig. 1A), nonadherent monocytes (relative to adherent monocytes) showed increased LPS-induced expression of proinflammatory genes involved in the TLR4 signaling cascade, including pathogen sensors (CD14, TLR2, and NOD2), NF-κB signaling (IRAK4, PELI1, PELI2, NFKB1, NFKB2, RELA, and IKBKB), and MAPK signaling (MAPK3, MAPK7, MAP2K1, MAP3K1, and MAP3K8) (Fig. 1D). Adherence had differential effects on the LPS-induced expression of chemokines, with adherent monocytes showing enhanced expression of genes encoding CC chemokines such as CCL2, CCL3, CCL7, CCL22, and CCL24 and reduced expression of genes encoding CXC chemokines and their receptors and ILs, such as CXCL1, CXCL2, CXCL3, CXCR2, CXCR3, CXCR4, IL7, IL9, IL10, and IL15 (Fig. 1E). Adherent monocytes showed more induction of genes involved in ROS and RNS production (Fig. 1F), especially RAC2, which was highly upregulated in LPS-stimulated adherent monocytes, but not in nonadherent monocytes, and is associated with production of ROS (28). In addition, adherent monocytes displayed a marked upregulation of genes encoding components of vacuolar ATPase (ATP6V1A, ATP6V1B2, ATP6V1D, ATP6V1E1, ATP6V1G1, ATP6V1H, and ATP6V1OA2), an enzyme transporter that functions to acidify intracellular compartments, and is a central protein complex in the killing and destruction of intracellular microorganisms (29). The expression of M1 and M2 genes did not clearly differ between adherent and nonadherent monocytes (Supplemental Fig. 4).

Together, these results indicate that adherence modifies the immune function of monocytes, mitigating LPS-induced proinflammatory TLR4 signaling and changing the cytokine profile in an anti-inflammatory direction while enhancing responses implicated in pathogen elimination.

To determine the impact of adherence on monocyte glucose metabolism, we measured glucose consumption and lactate secretion by resting and LPS-stimulated nonadherent and adherent monocytes. Although glucose consumption was not induced by LPS stimulation in either nonadherent or adherent monocytes (Fig. 2A), only adherent monocytes produced significantly more lactate upon LPS stimulation compared with medium control (Fig. 2B). Direct comparison of the induction of glycolysis in nonadherent and adherent monocytes displayed a significantly stronger glycolytic response in adherent monocytes (Fig. 2B). This was supported by GSEA of the glycolysis pathway, showing that the majority of genes involved in glycolysis was upregulated in adherent monocytes, resulting in a positive NES, but downregulated in nonadherent monocytes, resulting in a negative NES (Fig. 2C). A direct comparison of LPS-induced expression of glycolytic genes demonstrated that most genes were significantly more induced in adherent monocytes (Fig. 2D). The glycolytic flux was further assessed by measuring the incorporation of stable isotope-labeled [¹³C]-glucose in glycolysis intermediates. We found more [¹³C] incorporation in glucose-6-phophate and dihydroxyacetone phosphate in adherent monocytes, suggesting a higher glycolytic flux in adherent monocytes compared with nonadherent monocytes in the upper part of the pathway (Fig. 3). Incorporation of [¹³C] into lactate also tended to be higher in adherent monocytes; however, this was not significant. Of note, this method measures intracellular lactate only, and therefore, the secreted lactate is not analyzed in this assay. The [¹³C] incorporation in ribose-5-phosphate, the metabolite branching from the upper part of the glycolysis pathway representing the first step of the pentose phosphate pathway, was also higher in adherent monocytes. This suggests more flux toward the pentose phosphate pathway in adherent monocytes compared with nonadherent monocytes. Concordantly, adherent monocytes had a stronger induction of genes involved in the pentose phosphate pathway (Fig. 2E). Collectively, these results indicate that adherent monocytes upregulate the glycolysis pathway in response to LPS both transcriptionally and functionally, whereas nonadherent monocytes do to a far lesser extent.

We next sought to determine the role of glycolysis in the LPS-induced response of nonadherent and adherent monocytes. For this, we used two interventions to limit glycolysis: 1) glucose deprivation from the culture medium, and 2) treatment with the glucose analogue 2DG. The latter is commonly used as an inhibitor of glycolysis, as it competes with glucose but essentially blocks the second step in the glycolysis pathway (30). Surprisingly, glucose deprivation did not affect LPS-induced cytokine production (Fig. 4A) or heat-killed K. pneumoniae–induced ROS production (Fig. 4B) by either nonadherent or adherent monocytes. In contrast, 2DG reduced TNF (p = 0.012) and IL-1β (p = 0.055) release by nonadherent (but not adherent) monocytes and diminished IL-10 release by both nonadherent and adherent monocytes. 2DG also limited the ROS production of adherent, but not nonadherent, monocytes. Furthermore, 2DG, but not glucose deprivation, exerted strong effects on innate immune signaling pathways in LPS-stimulated nonadherent (Fig. 4C) and adherent monocytes (Supplemental Fig. 5A) relative to normal medium controls. 2DG reduced the expression of several genes encoding pathogen sensors (CD14, LY96, and TLR1), proximal pattern-recognition signaling proteins (MYD88, IRAK4, and BTK), and MAPK signaling (MAPK1, MAPK3, MAPK14, and MAP3K1), suggestive of an inhibitory effect on immune signaling. More downstream, within the cytokine-and-chemokine pathway (Fig. 4D, Supplemental Fig. 5B), effects of 2DG that reached statistical significance entailed reduced expression of genes encoding anti-inflammatory cytokines (IL-10 family members IL10, IL19, and IL24) and their receptors (IL4R, IL10RA, and IL10RB) as well as chemokines (CCL2, CCL7, CCL8, CCL13, CCL23, CXCL1, CXCL5, and CXCL6). This is in line with the strongly reduced IL-10 production by 2DG-treated monocytes (Fig. 4A), suggesting this might be primarily transcriptionally regulated. Finally, 2DG also reduced the LPS-induced expression of genes involved in ROS production and pathogen elimination in nonadherent (Fig. 4E) and adherent (Supplemental Fig. 5C) monocytes. Taken together, these data suggest that limiting glycolysis by glucose deprivation does not affect cytokine production or key innate immune pathways in either nonadherent or adherent monocytes. However, 2DG treatment results in reduced TNF, IL-1β, and IL-10 production in nonadherent monocytes and reduced expression of genes encoding essential innate immune pathways (i.e., TLR signaling, cytokines and chemokines, and ROS production), suggesting other effects of 2DG next to inhibition of glycolysis or differences in adaptation to these two interventions.

Glucose deprivation and 2DG inhibited glycolysis to a similar extent as indicated by reduced lactate secretion compared with normal medium (Fig. 5A). The discordant effects of glucose deprivation and 2DG on LPS-induced cytokine production by especially nonadherent monocytes (Fig. 4A) led us to hypothesize that these monocytes adapt differently to these two interventions. Indeed, although glucose deprivation resulted in a positive enrichment of genes encoding for compensatory metabolic pathways like the TCA cycle/OXPHOS and lipid metabolism compared with control (Glu⁺) medium, 2DG treatment resulted in negative enrichment scores for these pathways (Fig. 5B). In accordance, 2DG reduced the expression of a majority of genes encoding proteins involved in the TCA cycle and OXPHOS in LPS-stimulated nonadherent monocytes when compared with cells treated with normal medium and glucose-depleted medium (Fig. 5C). Notably, although the majority of genes in this pathway were downregulated by 2DG, some, like encoding citrate synthase (CS), were upregulated. Citrate synthase is one of the first enzymes of the TCA cycle and converts mitochondrial acetyl-CoA to citrate; although important for the TCA cycle, citrate synthase is also associated with fatty acid synthesis for which citrate is shuttled out of the TCA cycle and mitochondria to obtain cytosolic acetyl-CoA, a requirement for fatty acid synthesis (31, 32). In adherent monocytes, 2DG treatment also resulted in the downregulation of genes involved in the TCA cycle and OXPHOS when compared with glucose-depleted and normal medium (Supplemental Fig. 5D). Overall, these data suggest that although glucose deprivation selectively reduces glycolysis, 2DG inhibits both glycolysis and the TCA cycle/OXPHOS. To determine whether inhibition of OXPHOS by 2DG contributes to the reduced cytokine production by monocytes, we treated nonadherent and adherent monocytes with oligomycin, a specific inhibitor of complex V (ATP synthase) of OXPHOS (13, 33). Treatment with oligomycin significantly reduced the production of TNF in nonadherent monocytes and of IL-10 in nonadherent and adherent monocytes (Fig. 5D). In adherent cells, oligomycin did not impact TNF production and modestly, but significantly, enhanced IL-1β release. Together, these data suggest that 2DG-induced inhibition of TNF and IL-10 production by nonadherent monocytes could be (partially) caused by a 2DG effect on OXPHOS.

Although earlier investigations documented that adherence influences monocyte function, knowledge of the underlying mechanisms is limited. With the development of new techniques like RNAseq and the emerging field of immunometabolism, we decided to revisit this topic and to study the effect of adherence on monocyte function and metabolic reprogramming. We found that adherent monocytes (relative to nonadherent monocytes) produced less TNF and IL-1β (proinflammatory cytokines) and more IL-10 (anti-inflammatory cytokine) upon LPS stimulation and had an increased capacity to phagocytose and to produce ROS. Consistently, RNAseq analysis pointed at a reprogramming of monocytes after adherence, showing reduced expression of proinflammatory genes involved in TLR signaling and increased induction of genes involved in pathogen elimination in LPS-stimulated adherent monocytes when compared with nonadherent monocytes. These immune responses induced by LPS in adherent monocytes were accompanied by an increased glycolytic response, as indicated by lactate release, GSEA, and [¹³C]-glucose flux analysis. To obtain insight in the involvement of glycolysis in LPS-induced immune responses, this pathway was inhibited by glucose deprivation or 2DG. Remarkably, although these interventions similarly inhibited lactate production, only 2DG influenced monocyte functions, inhibiting expression of genes involved in TLR signaling and pathogen elimination, as well as (especially in nonadherent monocytes) cytokine release. 2DG, but not glucose deprivation, also reduced the expression of genes involved in OXPHOS. Chemical inhibition of OXPHOS influenced TNF and IL-10 release in a way similar to 2DG. Collectively, these data suggest that adherence modifies the metabolic and immunological profile of monocytes and that inhibition of glycolysis and OXPHOS (by 2DG), but not inhibition of glycolysis alone (by glucose deprivation), has a profound effect on immune functions of monocytes exposed to LPS. Our results in addition argue for careful consideration of (adherent or nonadherent) culturing conditions when studying immunometabolism in monocytes or in the context of specific diseases like systemic inflammation and sepsis.

Our finding that LPS stimulates the glycolytic rate in monocytes corroborates previous studies (15, 16, 34, 35). In extension to previous reports, we show that adherent monocytes exhibit a much stronger upregulation of glycolysis upon LPS stimulation when compared with nonadherent monocytes. Of interest, a recent investigation reported that 2DG reduces adherence of monocytes to fibrinogen (15), suggesting that glycolysis and possibly other energy metabolism pathways impacted by 2DG are mobilized for this process. Adherence is the first step in monocyte extravasation and differentiation toward macrophages. The main function of macrophages is to locate and phagocytose small particles, dead cells, and pathogens like bacteria (36). We found that adherence primes monocytes for this function, likely in part through induction of CD16, also known as FcγRIII, expression on the surface. Adherent monocytes indeed showed an enhanced phagocytic capacity compared with nonadherent monocytes. Interestingly, macrophages have been shown to require a secondary signal to induce inflammasome activation to enable IL-1β production, whereas monocytes activate the inflammasome pathway via an alternative route without a need for a secondary signal (37, 38). Our finding that adherent monocytes produce less IL-1β would therefore fit with the idea that adherence induces differentiation toward macrophages, reducing the IL-1β produced in the absence of a secondary signal. Additionally, adherence also resulted in enhanced ROS production upon stimulation with heat-killed K. pneumoniae and C. albicans (LPS barely induced ROS production). This too could be explained by adherence-induced differentiation toward macrophages. Monocytes are impaired in DNA repair, rendering them vulnerable to ROS-induced genotoxic stress, whereas monocyte-derived macrophages are DNA repair competent and genotoxic stress-resistant (39, 40), enabling them to safely induce ROS. However, the metabolic adaptation in adherent monocytes could also be a driving factor in this finding. The induction of glycolysis is highly associated with the capacity to produce ROS (15, 41–43). The ATP generated by glycolysis allows for the repurposing of the electron transport chain from ATP to ROS production (43). Furthermore, the induction of the pentose phosphate pathway and the subsequent generation of NADPH protects the cell from harmful effects elicited by ROS (13, 41, 44). Of note, we show in this study that 2DG treatment, but not glucose deprivation, resulted in reduced ROS production in adherent monocytes, suggesting cells can adapt to glucose deprivation in a way that ROS production is not affected or glycolysis is not as important as previously thought.

Adherence also enhanced the induction of genes involved in pathogen elimination. Adherent monocytes exhibited a dramatic upregulation of RAC2 in response to LPS, which is associated with production of ROS (28). In addition, adherent monocytes demonstrated a marked upregulation of genes encoding components of vacuolar ATPase, an enzyme transporter that acidifies intracellular compartments and is instrumental for the killing and destruction of intracellular microorganisms (29). Overall, these data suggest that adherence primes monocytes for a hallmark function of macrophages: phagocytosis and pathogen elimination.

2DG is a widely used inhibitor of glycolysis, which competes with glucose but cannot be metabolized further by the glycolysis pathway, leading to the accumulation of 2DG-6-phosphate, essentially blocking the second step in the pathway (30). Although glycolysis was strongly induced by LPS in adherent monocytes, 2DG affected the release of TNF and IL-1β of only nonadherent monocytes. Together with the fact that glucose deprivation did not affect cytokine release, this suggests that 2DG has effects besides inhibition of glycolysis impacting the production of cytokines. Indeed, because of its structural similarity to mannose, 2DG interferes with N-linked glycosylation, resulting in endoplasmic reticulum stress and activation of the unfolded protein response (45, 46), which in turn activates autophagy (47), and the interference with the glycosylation of cytokines could impair the secretory pathway and/or stability of these mediators (48, 49). Another possible explanation for the discordant results obtained with 2DG and glucose deprivation might be due to glucose obtained via glycogenolysis, which could be used in the glucose deprivation condition but would still be inhibited by 2DG (50). Although glycogen stores are mainly found in the liver and skeletal muscle tissue, blood leukocytes, especially neutrophils, also contain some glycogen (51). Cell-intrinsic glycogen metabolism can support TLR-induced functions, including TNF and IL-10 production of bone marrow–derived dendritic cells (50). However, little is known about the role of glycosylation or glycogenolysis on cytokine production in monocytes. Thus, the mechanism by which 2DG affects TNF and IL-1β production only in nonadherent monocytes remains to be established. 2DG reduced the expression of many genes involved in innate immune pathways such as pathogen sensing, TLR signaling, cytokine and chemokine expression, and pathogen elimination in both nonadherent and adherent monocytes, and its inhibitory effect on IL10 transcription could account for the reduced IL-10 production by both adherent and nonadherent monocytes.

Our results show that 2DG also inhibited the expression of genes involved in OXPHOS in monocytes. In accordance, a previous study reported that 2DG blocked not only glycolysis but also inhibited glycolytic-dependent OXPHOS in a mantle cell lymphoma cell line (52). Furthermore, an earlier study examined the impact of glucose deprivation on immune function and metabolism of monocytes within the first 12 h after LPS stimulation (35). Although monocytes incubated in glucose-deficient medium responded to LPS with a short-lived and flat increase in lactate production, these cells exhibited an increase in OXPHOS for up to 12 h, hinting at a compensation mechanism for the absence of glycolysis (35). This is in agreement with the significant induction of genes involved in the TCA cycle and OXPHOS we found in nonadherent monocytes cultured in glucose-deprived medium. Concordant with our results, glucose-deprived monocytes had a similar capacity to produce cytokines such as TNF and IL-1β upon LPS stimulation and also to phagocytose E. coli when compared with monocytes cultured in control medium (35).

Considering the reduced expression of genes involved in the TCA cycle and OXPHOS when treated with 2DG and the fact that nonadherent monocytes, in contrast to adherent monocytes, barely induced glycolysis upon LPS stimulation, we hypothesized that OXPHOS might be important for cytokine production by LPS-stimulated nonadherent monocytes. Indeed, addition of the OXPHOS inhibitor oligomycin reduced the production of TNF by nonadherent monocytes and that of IL-10 by monocytes independent of adherence, thereby mimicking the effect of 2DG. These data are partially in accordance with earlier investigations. A previous study reported similar inhibitory effects of 2DG and oligomycin on TNF and IL-1β production by adherent monocytes stimulated with C. albicans (53). In addition, the OXPHOS inhibitor rotenone inhibited LPS-induced IL-10 production (16), but not that of TNF (34) or IL-1β (16), by adherent monocytes. Notably, in contrast to 2DG, oligomycin modestly enhanced LPS-induced IL-1β release by adherent monocytes. It has been shown that oligomycin induces ROS production in T cells (54) and chondrocytes (55). In macrophages, ROS stimulates IL-1β production via increased activation of the classical inflammasome pathway (56). Monocytes are not dependent on the classical inflammasome pathway but can activate the inflammasome via an alternative pathway (37). However, because adherence is the first step toward macrophage differentiation, it is possible that ROS-induced activation of the classical inflammasome pathway promotes IL-1β production in adherent monocytes, but not in nonadherent monocytes. The mechanism by which OXPHOS might impact cytokine production in monocytes remains to be determined. Together, these results indicate that 2DG modifies cellular processes independent of inhibition of glycolysis.

Our study has limitations. The RNA expression analysis is based on a 24-h incubation with LPS and medium control, which likely is not optimal for measuring the expression of all TLR4 signaling genes and/or the expression of genes encoding chemokines and cytokines. Additionally, we found that some monocyte aggregates were formed after LPS stimulation under nonadherent conditions (Supplemental Fig. 2B). At present, it is unclear whether this can affect cytokine production by monocytes. We were not able to directly measure OXPHOS because of the fact that monocytes adhere to culture plates used in the Seahorse [the most commonly used assay system to measure OXPHOS (57)]. Finally, it is should be noted that the use of a cell-repellent surface to avoid adherence may induce artificial effects that do not relate to differences between adherent and nonadherent monocytes in humans in vivo.

Adherence of monocytes is associated with profound functional and metabolic reprogramming. 2DG has a major effect on the transcription of gene sets encoding proteins involved in key innate immune functions, which are not reproduced by selective inhibition of glycolysis through glucose deprivation. Although the effect of OXPHOS inhibition on the LPS-induced TNF and IL-10 release resembled that of 2DG, more studies are warranted to elucidate the relative contribution of distinct energy metabolism pathways to immune functions of nonadherent and adherent monocytes.

We thank Linda Koster from the Core Facility Genomics, Amsterdam UMC, for the technical assistance regarding the preparations for the RNAseq analysis. We thank Leo Joosten from Radboud UMC, Nijmegen, for kindly providing the heat-killed C. albicans.

The authors have no financial conflicts of interest.