Nicotinamide Antagonizes Lipopolysaccharide-Induced Hypoxic Cell Signals in Human Macrophages

Immune activation and suppression respectively protect the host from disease and inflammatory damage. An imbalance in these responses, as identified in sepsis, can lead to excessive inflammation, tissue injury, disease progression, secondary infections, organ failure, and death (1, 2). Gram-negative bacteria are common sepsis pathogens, and the outer membrane molecule LPS is a recognized conserved motif in innate immunity (3). Resident macrophages exist in nearly every organ and are therefore early responders to invading pathogens (4). LPS binding to TLR4 on macrophages promotes inflammation through various cell signaling pathways that generate downstream hypoxic responses (5), including PI3K/AKT and NF-κB (6), which promote hypoxia-inducible transcription factor-1α (HIF-1α) stabilization (7) and HIF-1α transcription (8), respectively. These downstream responses induce hypoxic metabolism, promote the expression of cell surface receptors, and stimulate the production of cytokines/chemokines involved in macrophage polarization to an M1 (classical) phenotype (6).

In addition to cell signals, biomolecular cues regulate HIF-1α stability. For example, microenvironmental oxygen, iron, and the metabolite α-ketoglutarate decrease HIF-1α stability, as these molecules promote prolyl hydroxylase domain (PHD) activity (9), which adds hydroxyl groups to HIF-1α for the recruitment of the von Hippel–Lindau tumor suppressor protein (VHL) E3 ubiquitin ligase (10). Subsequently, ubiquitination of HIF-1α through VHL or additional PHD2-independent ligases promotes HIF-1α degradation via the proteasome or lysosome (11). Various factors can also alter HIF-1α activity by interacting directly with HIF-1α (e.g., poly(ADP-ribose) polymerases [PARPs], sirtuins [SIRTs]) (12–14).

Nicotinamide (NAM) is the amide form of dietary niacin (15). Enzymes in the cytosol, mitochondria, and nucleus catalyze NAM conversion to NAD, which is used in NADPH redox reactions or is consumed as a substrate by PARPs, CD38/CD157 ectoenzymes, and SIRTs (16). In a murine skin infection model, NAM increased neutrophil clearance of the Gram-positive bacteria Staphylococcus aureus (17). In the context of Gram-negative bacteria, NAM antagonized LPS-induced oxidative stress and the activity of the proinflammatory transcription factor p65/RelA in macrophages (18, 19). Because oxidative stress and p65/RelA promote the stability and transcription of HIF-1α (20), we therefore hypothesized that NAM inhibits mechanisms involved in HIF-1α stabilization, activity, and subsequent immunological responses in macrophages.

In experiments using human monocyte-derived macrophages (HMDMs) reported in the present study, we demonstrate that NAM antagonized LPS-induced activation of HMDMs by promoting NAD production, predominantly through the salvage pathway, and reducing phagocytosis, NOX2 activity, and the production of proinflammatory mediators and cell surface receptors. Hypoxic/proinflammatory cell signals (AKT, NF-κB) associated with these responses and downstream activation of HIF-1α were also downregulated by NAM. The mechanisms involved reduced phosphorylation of cell signals, increased deacetylation and ubiquitination of transcription factors, and increased production of molecules that degrade HIF-1α.

Human PBMCs were obtained from healthy donors on an Institutional Review Board–approved National Institutes of Health protocol (99-CC-0168). Research blood donors provided written informed consent and blood samples were deidentified prior to distribution (ClinicalTrials.gov identifier NCT00001846). Cells (3 × 10⁶/ml) were cultured overnight in RPMI 1640 medium plus l-glutamine (Life Technologies) supplemented with 10% FBS (Life Technologies). Cells were resuspended (2 × 10⁶/ml) in IMDM (Life Technologies) supplemented with 10% FBS and 30 ng/ml M-CSF (PHC9504, Thermo Scientific) for 6 d. HMDMs were lifted and plated at 5 × 10⁵/ml in 10% FBS/IMDM overnight prior to additional cell stimulation with NAM (N0636, Sigma-Aldrich, suspended in media), MG-132 (M7449, Sigma-Aldrich, suspended in DMSO), IOX4 (S6684, Selleck Chemicals, suspended in DMSO), IDO1 inhibitor (S8557, Selleck Chemicals, suspended in DMSO), deferoxamine mesylate (DFOM; sc-203331, Santa Cruz, suspended in water), 1-methyl-d-tryptophan (1-MT; 452483, Sigma-Aldrich, 1 mM intermittently vortexed and incubated in warm media [37°C] until soluble), or LPS (433, List Biologicals, suspended in media) as detailed in representative figures. RAW 264.7 murine macrophages stably transfected with GFP and mCherry reporters fused to p65 and TNF-α genes, respectively, were provided by Dr. Iain Fraser (21) and cultured in 10% FBS/RPMI 1640. EAhy926 cells were purchased from American Type Culture Collection and cultured in 10% FBS/DMEM (Life Technologies).

Cells were treated with 0.25% Trypsin/EDTA (1×, Life Technologies) for 3 min and neutralized with equivalent media prior to scraping. Cells were washed with 2% FBS/PBS (400 × g, 5 min, 4°C), incubated (15 min, 4°C) with Hu FcR binding inhibitor (14-9161-73, eBioscience), washed, and treated with Abs (30 min, 4°C) prior to treatment with 2% methanol-free formaldehyde (04018, Polysciences) in PBS (15 min, 23°C) and suspension in PBS. A total of 30,000 events were assessed on an LSRFortessa flow cytometer (BD Biosciences). Data were analyzed with FlowJo data analysis software (Tree Star, Ashland, OR). Abs included PerCP/Cy5.5-CD38 (356614), BV510-CD40 (334330), allophycocyanin-CD80 (305220), and PE/Cy7–programmed death ligand 1 (PD-L1; 329718) (all from BioLegend).

Total RNA was extracted using the RNeasy kit (Qiagen), and cDNA was synthesized with iScript cDNA synthesis kits (Bio-Rad). SYBR Green real-time PCR was performed with SsoAdvanced Universal SYBR Green supermix (Bio-Rad) on an Applied Biosystems ViiA 7 instrument. Real-time PCR was performed with the following primers: CD38, 5′-GCA GCA ACA ACC CTG TTT CA-3′ (forward); 5′-CAC ACT CCC AAA AGT GCT GT-3′ (reverse); CD80, 5′-AGG CAG GGA ACA TCA CCA TC-3′ (forward); 5′-TCA CGT GGA TAA CAC CTG AAC A-3′ (reverse); NAMPT, 5′-TGG CCT TGG GAT TAA CGT CT-3′ (forward); 5′-CAA AAT TCC CTG CTG GCG TC-3′ (reverse); CD40, 5′-TGA TGT TGT CTG TGG TCC CC-3′ (forward); 5′-GAT AAA GAC CAG CAC CAA GAG G-3′ (reverse); PDL1, 5′-CCT GCA GGG CAT TCC AGA AA-3′ (forward); 5′-TAG GTC CTT GGG AAC CGT GA-3′ (reverse); HIF1A, 5′-TTC CTT CTC TTC TCC GCG TG-3′ (forward); 5′-ACT TAT CTT TTT CTT GTC GTT CGC-3′ (reverse); EGLN1, 5′-CAA AGC CCA GTT TGC TGA CA-3′ (forward); 5′-CCA AAC AGT TAT TGC GTA CCT TG-3′ (reverse); PUM1, 5′-GGC TTT GGC AGA ACG GAT TC-3′ (forward); 5′-TCT CAT TAA TTA CCT GCT GGT CTG-3′ (reverse). Quantitation was performed using the ΔΔCt method relative to the PUM1 internal control.

Four male donor HMDMs (1 × 10⁶ cells/2 ml media in six-well plates) were treated with four conditions (vehicle control, 10 ng/ml LPS, 10 mM NAM, or LPS+NAM) for 4 h. Samples were validated prior to microarray assessment by affirming NAM downregulation of LPS-induced responses. Specifically, quantitative PCR of HIF-1α, PD-L1, CD38, CD40, and CD80 was examined at 4 h and immunoblots of HIF-1α protein were examined at 24 h for each human donor. The 16 samples of total RNA were extracted using the RNeasy kit (Qiagen) with processing through QIAshredder columns (Qiagen). Total RNA was quantified by a NanoDrop 8000 spectrophotometer (Thermo Fisher Scientific) and quality was examined by RNA 6000 Nano LabChip (Agilent 2100 Bioanalyzer). RNA samples were then processed for microarray expression analysis using PrimeView Chip 30 microarrays (901838, Thermo Fisher Scientific), a GeneChip 3′ IVT Plus reagent kit (902416, Thermo Fisher Scientific), and a GeneChip hybridization wash and stain kit (900720, Thermo Fisher Scientific). Scanning and image acquisition was performed on a GeneChip scanner 3000 7G.

Microarray signal intensity values were assessed using Thermo Fisher Scientific Transcriptome Analysis Console (version 4.0.1.36) and they were normalized and log₂ transformed using the robust multiarray average method. Probe set IDs were annotated using PrimeView Human Gene Expression Array annotation file version 36 (release date March 30, 2016). The transformed data matrix with resulting signal intensity values for each of 49,372 probe sets was imported to R and subjected to principal component analysis to visualize the relative location of chips in a low-dimensional space allowing for detection of outliers or other relevant patterns. Raw and robust multiarray average–normalized data were submitted to the National Center for Biotechnology Information Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE188206; accession no. GSE188206), and all data are minimum information about a microarray experiment (MIAME) compliant.

A linear model with fixed effects for donor, LPS (LPS, no LPS), NAM (no NAM, NAM), and interaction between LPS and NAM were fit to the expression data for each gene using the Bioconductor R package limma. As part of each linear model analysis, p values were obtained for the main NAM effect, the main LPS effect, and the interaction between LPS and NAM. Because there was evidence of an interaction between LPS and NAM, simple effect tests were performed to compare either NAM or LPS to the control and NAM+LPS to LPS. A false discovery rate (FDR) of 5% and a 1.5 fold-change cutoff yielded a total of 889, 1169, and 1115 unique transcripts, respectively.

Cells (5 × 10⁵/ml) in 10% FBS/IMDM were plated (80 μl/well) in a 96-well SF96 cell culture microplate (103729-100, Seahorse FluxPaks, Agilent Technologies). The following day, cells were treated with and without NAM or LPS. After 24 h, the medium was removed and the plate was washed with 200 μl of RPMI 1640 (103571-100, Agilent Technologies) supplemented with 2 mM l-glutamine (103579-100, Agilent Technologies) prior to adding 180 μl of this same medium to each well. Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were assessed using XFe96 cartridges on an XF96 extracellular flow analyzer. The baseline was measured over five cycles (6 min/cycle) with the last three cycles incorporated in calculations. Injections included reagents from the XF Cell Mito Stress kit (1030150-100, Agilent Technologies), 25 mM glucose (103577-100, Agilent Technologies), 1.5 μM oligomycin (Mito Stress kit), 1 mM pyruvate (103578-100, Agilent Technologies) combined with 1 μM carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, Mito Stress kit), and 0.5 μM rotenone and antimycin A (Mito Stress kit) combined with 20 μg of Hoechst 33342 (H21492, Thermo Fisher Scientific) for three cycles (6 min/cycle). The plate was then imaged for cell nuclei as a measure of cell numbers on an Agilent Cytation 5 cell imaging multimode reader. ECAR and OCR measurements were normalized to experimental cell numbers.

Cells were solubilized with RIPA lysis and extraction buffer (80099, Thermo Fisher Scientific) supplemented with Halt protease and phosphatase inhibitor cocktail (78445, Thermo Fisher Scientific). Protein, equivalent to 3.3 × 10⁵ cells, was added to 1× sample reducing agent (NP0009, Thermo Fisher Scientific) and 1× loading dye (NP0007, Thermo Fisher Scientific), sonicated, boiled (5 min, 100°C), and loaded onto NuPAGE Novex 4–12% Bis-Tris protein gels (NP0335, Thermo Fisher Scientific). For phosphorylated and total protein blots, equal amounts of protein lysate from the same donor were loaded onto two separate gels. Proteins were transferred with a nitrocellulose Trans-Blot Turbo transfer pack (1704159, Bio-Rad) using a Bio-Rad Trans-Blot Turbo system. Blots were cut for multiple detection and incubated with Abs raised against AKT1 (s473) (9271S, Cell Signaling Technology, 1:1000), AKT (9272S, Cell Signaling Technology), acetylated p65 (3045, Cell Signaling Technology, 1:1000), actin (612656, BD Biosciences, 1:5000), FOXO1 (s256) (9461S, Cell Signaling Technology, 1:1000), FOXO1 (2880S, Cell Signaling Technology, 1:1000), histone deacetylase (HDAC)4 (15164S, Cell Signaling Technology, 1:1000), HIF-1α (MAB1536, R&D Systems, 2 μg/ml), IDO1 (86630S, Cell Signaling Technology, 1:1000), lactate dehydrogenase A (LDHA; 2012S, Cell Signaling Technology, 1:1000), NAM phosphoribosyltransferase (NAMPT; MAB4044, R&D Systems, 1 μg/ml), p65 (8242, Cell Signaling Technology, 1:1000), PHD2 (4835S, Cell Signaling Technology, 1:3000), PSMF1 (HPA041300, Sigma-Aldrich, 0.4 μg/ml), SART1 (ab8858, Abcam, 1 μg/ml), SIRT1 (s47) (2314S, Cell Signaling Technology, 1:1000), SIRT1 (2310S, Cell Signaling Technology, 1:1000), ubiquitin (13724, Cayman Chemical, 1:500), or VHL (ab140989, Abcam, 1:1000). Blots were washed in 0.5% PBS with Tween 20 (PBST) and subsequently incubated with HRP-conjugated secondary Abs. Bound secondary Ab was visualized following incubation of the membrane with SuperSignal West chemiluminescent HRP substrate (Thermo Scientific Pierce) and using a ChemiDoc MP imaging system (Bio-Rad). Luminescence was quantified and evaluated via the application of ImageJ software (National Institutes of Health).

Protein G Sepharose 4 Fast Flow (GE17-0618-01, Sigma-Aldrich) were prepared by transferring 40 μl of the Sepharose beads/sample and adding 10× volume lysis buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% Triton X-100, 0.2 mg/ml BSA, sterile culture grade H₂O) to a 15-ml tube. The Sepharose beads were then centrifuged at 4000 rpm for 3 min and the supernatant was aspirated. The same volume of lysis buffer was added and tubes were rotated for 5 min at 4°C, centrifuged and aspirated four times. After the final aspiration, the Sepharose beads were suspended in 10% nonfat dry milk in 0.5% PBST at 10× volume and rotated for 2 h at 4°C. The tube was centrifuged and the milk was aspirated. The Sepharose beads were resuspended in 10% milk/0.5% PBST at a volume 2× of the original Sepharose bead volume and stored at 4°C until use. In preparing lysates, monocytes (3 × 10⁶/ml) were cultured overnight in 10% FBS/RPMI and resuspended (15 × 10⁶ 10 ml) in 10% FBS/IMDM in 100-mm plates supplemented with 30 ng/ml M-CSF for 6 d. The medium was replaced and 2 plates/sample were stimulated with 10 ng/ml LPS or 10 ng/ml LPS and 10 mM NAM for 24 h. The plates were washed twice with PBS and treated with 225 μl of CytoBuster protein extraction reagent (71009, Millipore, Billerica, MA) supplemented with Halt protease and phosphatase inhibitor cocktail (78445, Thermo Fisher Scientific) and similar treatments were combined. Lysate (10%) was removed from each treatment and supplemented with 1× sample reducing agent and 1× loading dye to measure input protein responses. The remaining lysates were incubated with mouse IgG-agarose beads (A0919, Sigma-Aldrich) for 2 h at 4°C on a tube rotator. Lysates were centrifuged (5000 rpm) for 2 min at 4°C and the supernatant was transferred to fresh tubes and incubated with 15 μg/ml Abs (HIF-1α, Ab1, Abcam; p65, 8242, Cell Signaling Technology) on a tube rotator at 4°C overnight. The lysates are then treated with 50 μl of suspended and preblocked Sepharose beads and placed on a tube rotator at 4°C for 2 h. Samples were centrifuged (5000 rpm, 4°C, 60 s), supernatant was aspirated, and 1 ml of lysis buffer is added to the beads that are washed four times with 1 ml of lysis buffer. After the final wash, the beads are treated with sample reducing agent and 1× loading dye and CytoBuster at a volume equivalent to the input and sufficient for three aliquots to a gel. Heat-treated beads (70°C, 10 min) were centrifuged (13,000 rpm, room temperature, 5 min) and the supernatant (output) was stored at −20°C for use in immunoblots.

Analytes from conditioned media were assessed with a Bio-Plex Pro human chemokine panel, 40-plex (171AK99MR2, Bio-Rad) and a Bio-Plex pro human inflammation panel 1, 37-plex (171AL001M, Bio-Rad) on a Bio-Rad Bio-Plex 200 system. For each of the 33 analytes, differences between experimental conditions were assessed by a Wilcoxon signed-rank test and a FDR of 5% was used to declare significance.

HMDMs were cultured as above for 24 h in the presence or absence of 10 mM NAM and treated with 1 mg/ml dextran labeled with FITC (FD40, Sigma-Aldrich, 100 mg) for 30 min. Cells were lifted, stained for viability, fixed, and assessed by flow cytometry.

P. aeruginosa PAO1 encoding GFP (22) was serially diluted, plated on Remel contact I blood agar (tryptic soy agar with 5% sheep blood, Thermo Fisher Scientific), and counted. Log phase cultures at a macrophage/bacteria count of 1:25 were added to macrophages in six-well plates for 30 min. Cells were lifted, stained for viability, fixed, and assessed by flow cytometry.

In evaluating the effects of LPS and NAM in HMDMs, we performed microarray analysis of HMDMs after a 4-h treatment with vehicle control, NAM, LPS, or NAM+LPS. In Fig. 1A, principal component analysis of HMDMs from four different healthy donors revealed global treatment-related shifts in transcriptional profiles in response to the four conditions. In Supplemental Table I, we identified unique transcripts comparing either NAM or LPS to the control and NAM+LPS to LPS with a cutoff of 1.5 fold change and FDR of <0.05.

NAM is a primary component in the metabolism of NAD, which is synthesized from tryptophan in the de novo pathway or recycled from NAD precursors (e.g., NAM mononucleotide [NMN] or NAM riboside [NR]) in the salvage pathway (23) (Fig. 1B). NAD can also be catabolized on the cell surface into NR in a series of reactions involving CD38, extracellular NAMPT (also known as pre–B cell colony enhancing factor [PBEF] or visfatin), which additionally serves as a macrophage prosurvival cytokine, and CD73 (24, 25). NR may enter the cell through a transporter, and upon NR kinase phosphorylation (26) forms NMN, which is also a product of intracellular NAMPT in the salvage pathway (23). NAM/nicotinic acid mononucleotide adenylyltransferase (NMNAT)1–3 (NMNAT1-3) transfers adenylate from ATP to NMN or nicotinic acid mononucleotide (NMAN), forming NAD (23) (Fig. 1B).

LPS has previously been shown to induce NAMPT, the rate-limiting enzyme in the NAD salvage pathway and a HIF-2α target gene (27). In Fig. 1C, we also show that LPS induced NAMPT production that was not modified by NAM despite a small but significant reduction at the gene level (−1.18 fold change, FDR < 0.002) for the NAM effect in the LPS response. Because reductions in NAD levels have been previously reported in murine macrophages in response to the highly specific noncompetitive NAMPT inhibitor FK866 (28, 29), we treated HMDMs with FK866.

In Fig. 1D, the levels of NAD significantly increased in response to each treatment, and these levels were reduced by FK866. The significant reductions in the control and LPS treatment by FK866 identify dependence on the salvage pathway. The lack of significance but clearly reduced levels of NAD in the FK866 NAM treatments may reflect NAM competition with FK866 (29). As shown in Fig. 1E, the levels of NADH were also induced in response to each treatment, but FK866 only significantly inhibited NADH levels in the LPS treatment, further supporting a salvage pathway dependence in the LPS response. In Fig. 1F, NAD/NADH levels increased in the presence of NAM and each treatment was inhibited by FK866. Taken together, these data suggest that both NAM and LPS produce NAD through the salvage pathway.

Because NAD may also be formed through tryptophan catabolism in the de novo pathway, we examined IDO-1, the rate-limiting enzyme in the de novo pathway (30). As shown in Fig. 1G, LPS-induced production of IDO1 protein was significantly reduced by NAM cotreatment, indicating that NAM may alter the production of kynurenine metabolites.

In assessing the potential function of IDO1 in the NAM effect, we treated cells with a potent IDO1 inhibitor (IDO1i). In contrast to FK866 treatment (Fig. 1D), IDO1i increased NAD levels in the control and NAM treatment (Fig. 1H), suggesting increased activation of the salvage pathway. In the LPS treatments, inhibiting IDO1 did not affect NAD levels, similar to a previous study in human monocytes (31). IDO1 also oxidizes NADH to form NAD (32), and in Fig. 1I, inhibition of IDO1 with IDO1i increased NADH levels in each treatment. In Fig. 1J, IDO1i inhibited NAD/NADH only in the LPS treatments, predominantly because of changes in NADH not NAD. To support these findings, we also tested the more common competitive IDO1 inhibitor, 1-MT (33), which performed comparably to IDO1i (Fig. 1K–M).

Taken together, these data indicate that 10 mM NAM or 10 ng/ml LPS, supplemented to media containing 800 µM tryptophan and 33 µM NAM, promotes NAD production through the salvage pathway given that NAD levels are reduced by FK866 but not by IDO1i or 1-MT. Moreover, NADH levels were also induced by NAM or LPS and enhanced by IDO1 inhibition, likely as a result of reduced IDO1-mediated oxidation of NADH (32). Differential NADH levels when comparing LPS to NAM treatments involving FK866 highlight the potential noncompetitive functions of this inhibitor with NAM as previously described (29).

Because LPS induces the expression of cell surface markers involved in Ag presentation (CD80, CD40) (34), myeloid suppression (PD-L1) (35, 36), and NAD metabolism (CD38) (37, 38), we assessed these markers in the presence or absence of NAM and/or LPS at the mRNA and protein levels at 4 and 24 h, respectively. In Fig. 2A–D, we demonstrate by quantitative real-time PCR increased levels of mRNA in response to LPS that were antagonized by NAM, similar to results in our microarray. In Fig. 2E–H, we identified increased cell surface expression of CD40, CD38, CD80, and PD-L1 in the presence of LPS that was significantly reduced by NAM. This NAM anti-inflammatory effect is also identified in Fig. 2I–L and Supplemental Figs. 1 and 2, which show reduced production of cytokines and chemokines in the presence of NAM. To understand these responses, we examined cell signaling pathways common to LPS and the receptors and mediators produced.

LPS activates the PI3K/AKT pathway (6), which alters metabolism either directly, through phosphorylation-mediated regulation of metabolic enzymes (e.g., 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase, ATP citrate lyase), or indirectly, through the control of various cell signals (e.g., NAD kinase, tuberous sclerosis complex 2, p47^phox, pyruvate dehydrogenase kinase 1) and transcription factors (e.g., FOXO1, NF-κB, HIF-1a) (6, 39). NAM inhibits AKT phosphorylation in the murine lung and RAW 264.7 murine macrophages (40), possibly by altering the acetylation or ubiquitination of the signaling molecule (41). In Fig. 3A–C, we show LPS-induced phosphorylation of AKT (serine 473) and NAM-mediated inhibition relative to actin and total AKT in HMDMs, indicating that NAM inhibits this proinflammatory molecule and potentially its downstream signals.

To further demonstrate AKT activation, we examined FOXO1, which is an established factor in the production of antioxidant molecules (42). Acetylated FOXO1 inhibits FOXO1 DNA binding activity and promotes AKT-induced FOXO1 phosphorylation at serine 256 (43). Phosphorylated FOXO1 (serine 256) is exported from the nucleus, ubiquitinated, and targeted to the proteasome for degradation (44). In Fig. 3D, NAM induced modest but significant decreases in p-FOXO1 levels compared with the control. In Fig. 3E, the levels of total protein are significantly induced by NAM and inhibited by LPS, highlighting increased phosphorylation/degradation in the presence of LPS and reduced phosphorylation/degradation in the presence of NAM, as displayed in Fig. 3F. FOXO1 ubiquitination by the E3 ligase MDM2 promotes FOXO1 proteasome degradation (44), and in Supplemental Fig. 3F, MDM2 gene expression was induced by LPS and inhibited by NAM, supporting the reduced degradation of FOXO1 by NAM (Fig. 3E).

An additional downstream signal of AKT is NF-κB (45). Posttranslational modifications of the NF-κB subunits and cofactors regulate NF-κB DNA binding, transcriptional activity, and release from the bound NF-κB inhibitor IκBα (46). Acetylation is a posttranslational modification that has a role NF-κB activation. Deacetylases consist of seven SIRTs and 11 HDACs (Table I). HDAC4 acts as a SUMO E3 ligase, which promotes the stability of IκBα (47). HDAC1, HDAC3, and SIRT1 promote the deacetylation of p65/RelA, which promotes p65/RelA binding to IκBα (46). In Table I, NAM promoted the expression of HDAC1, HDAC3, HDAC4, and SIRT1 in the LPS+NAM treatment response. In Fig. 4A, LPS inhibited HDAC4 production and NAM induced a small but significant increase in HDAC4 production, in association with shifts in gene expression (Table I). In Fig. 4B, activated SIRT1 (noted by increased S47 phosphorylation) is induced by either NAM or LPS alone and in combination compared with controls. Levels of total SIRT1 did not significantly change (Fig. 4C), resulting in an increased p-SIRT1/total SIRT1 ratio in the presence of LPS with and without NAM (Fig. 4D). Additional immunoblots demonstrating either a single or a double band for SIRT1 healthy donors are provided in Supplemental Fig. 3A. These data combined suggest that NAM may promote p65 deacetylation and IκBα stability in regulating NF-κB activation.

We subsequently performed immunoprecipitations of p65 from HMDM lysates and probed for acetylated p65 and ubiquitin. In Fig. 4E, immunoblots of immunoprecipitated p65 identified decreases in acetylated p65 but increases in ubiquitinated protein. Measuring the amount of total p65 input in these assays (Fig. 4F) revealed a small but significant NAM-induced loss of p65 production, which correlated with the NAM effect in the LPS response in our microarray (RelA, −1.4 fold change, FDR < 0.0005). Moreover, RAW 264.7 murine macrophages stably transfected with GFP and mCherry reporters fused to p65 and TNF-α genes, respectively (21), were treated with LPS and NAM and assessed for fluorescence by flow cytometry. In Fig. 4G and 4H both the GFP-p65 reporter and the mCherry-TNF-α reporter, which is activated by p65, were induced by LPS and antagonized by NAM.

Although NF-κB is a transcriptional regulator of HIF-1α (8), the effect of NAM in the LPS response in our microarray only revealed a 1.15-fold decrease in HIF-1α gene expression at an FDR <0.02. In examining HIF-1α protein levels, we first evaluated whether LPS induced HIF-1α protein in comparison with the positive control, DFOM, a characterized hypoxia mimetic. In Fig. 5A, we identified an ∼120-kDa band associated with HIF-1α production that was upregulated by DFOM and LPS compared with the control. We then evaluated whether NAM modulated the production of HIF-1α. We treated HMDMs with 10 ng/ml LPS in the presence or absence of increasing concentrations of NAM for 24 h. As shown in Fig. 5B, HIF-1α production decreased in response to increasing concentrations of NAM compared with the LPS control. NAM did not have an effect on DFOM-induced HIF-1α (Supplemental Fig. 3B).

To determine whether the effects of NAM on HIF-1α were specific to HMDMs, the endothelial cell line EAhy926 was also assessed. As shown in Fig. 5C, HIF-1α protein was identified in EAhy926 cells and was significantly reduced by NAM after a 24-h treatment. Furthermore, EAhy926 cells transfected with luciferase reporter plasmids driven by the hypoxia response elements exhibited decreased luciferase activity following treatment with NAM compared with the control (Fig. 5D), correlating with the decreased HIF-1α production in Fig. 5C.

LPS-induced HIF-1α activation leads to increased glycolysis (48). To examine the effects of NAM on this metabolic pathway, we measured the ECAR of the surrounding media, similar to previous work performed in macrophages (49). In Fig. 5E and 5F, HMDMs cultured with NAM demonstrated reduced overall ECAR and reduced glycolysis compared with the control or LPS-treated cells, indicating that NAM alters the HIF-1α physiological response of glycolysis. The OCR in these same experiments was variable and did not reveal significant differences in mitochondrial respiration between treatments (data not shown). Lastly, a downstream target of HIF-1α is LDHA (50). In Fig. 5G, HMDM production of LDHA was reduced by coculture with NAM and LPS compared with LPS alone, supporting decreased HIF-1α activity and decreased glycolysis.

NAD fuels the citric acid cycle and generates α-ketoglutarate that is essential for PHD activity (16, 51). Three mammalian PHD isoforms exist (PHD1, PHD2, PHD3). HIF-1α is a preferred target of PHD2, the only PHD essential for life (11). PHD2 also hydroxylates and inactivates an upstream activator of hypoxia, AKT (52). To understand whether the effect of NAM on HIF-1α is mediated by PHD2 activity, we assessed HMDM PHD2 protein production following treatment with NAM with and without LPS. As shown in Fig. 6A, we identified NAM-induced production of PHD2 in the presence or absence of LPS.

To understand whether NAM-induced PHD2 production is responsible for the decreased production of HIF-1α (Fig. 5B), we treated HMDMs with a potent PHD2 inhibitor, IOX4, which competes with α-ketoglutarate in binding PHD2 (53). In Fig. 6B, we show that NAM inhibition of HIF-1α production is blocked by 10 μM IOX4, indicating that NAM regulates HIF-1α degradation through PHD2. We also demonstrated this effect with a less potent inhibitor, IOX2 (Supplemental Fig. 3C).

PHD2 hydroxylation of HIF-1α induces the recognition of HIF-1α by VHL, which is an E3 ubiquitin ligase that functions with a complex of proteins to ubiquitinate HIF-1α for targeted degradation by the proteasome (10). Neither VHL mRNA (Supplemental Fig. 3F) nor protein (data not shown) was significantly affected by NAM treatment.

Because various ubiquitin genes that activate, conjugate, ligate, and remove ubiquitin from target proteins were significantly and inversely modified by LPS compared with NAM (Supplemental Fig. 3D–G), we examined the production of ubiquitin in HMDMs. In Fig. 6C, a band corresponding to the molecular mass of free ubiquitin was identified in control and LPS treatments that was significantly reduced in the presence of NAM. This may suggest that the intracellular pool of free ubiquitin was relocated to additional proteins, similar to p65 (Fig. 4E). We therefore performed an immunoprecipitation of HIF-1α from HMDM lysates and probed for ubiquitin. In Fig. 6D, we identified NAM-induced loss of HIF-1α protein in the direct lysates (input) and NAM-induced ubiquitination of HIF-1α in the immunoprecipitation (output).

Because ubiquitination targets proteins to the proteasome for degradation, we examined the effect of NAM on HIF-1α in the presence of the common proteasome inhibitor MG-132. In Fig. 6E, HIF-1α was induced by MG-132 and was not blocked in the presence of NAM, despite increased PHD2 production, suggesting that NAM-induced inhibition of HIF-1α occurs through the proteasome.

PSMF1, also known as PI31, inhibits the assembly of the 19S structure in the 26S proteasome (54) by competing with activating particles for 20S binding (e.g., PA700) (55). In Fig. 6F, LPS-induced production of PSMF1 was significantly inhibited by NAM. Tankyrase (TNKS) is a PARP that inhibits PSMF1 affinity for the 20S proteasome (56) and promotes target protein ubiquitination and degradation through ADP ribosylation (57). In our microarray, NAM significantly (p < 0.0001, FDR < 0.0001) induced TNKS gene expression in the presence (1.50-fold increase) or absence (1.48-fold increase) of LPS. Taken together, these data support that NAM-induced degradation of HIF-1α requires the proteasome and PHD2.

Ubiquitin regulates various biological processes in the cell, including endocytosis of plasma membrane proteins (58) and phagocytic killing of bacteria (59). Acetylation is also important to the stability of cytoplasmic microtubules that are necessary for the phagocytic functions of macrophages (60). In Fig. 7A and 7B, HMDMs were treated for 24 h with NAM prior to 30-min stimulation with FITC-labeled dextran or GFP-labeled P. aeruginosa. In this study, we detected reduced endocytosis of FITC-labeled dextran and phagocytosis of GFP-labeled P. aeruginosa compared with the untreated control.

Bacteria phagocytosed by macrophages are directed to the phagosome where free radicals involved in microbial killing are produced via the activation of the NOX2 NADPH oxidase (61). The cytosolic protein p47^phox, essential for the organization and stabilization of the NOX2 NADPH oxidase (62), is activated by hypoxia and AKT (6, 39). This gene was reduced in our microarray (NCF1 (p47), −1.46 fold change, FDR < 0.03) in examining the NAM effect in the LPS response. In Fig. 7C, LPS-induced p47^phox was inhibited by NAM plus LPS compared with LPS alone.

In summary, the NAM effect in LPS-activated macrophages is associated with increased intracellular NAD production through the salvage pathway, posttranslational modifications of proteins involved in cell signaling (AKT, FOXO1, p65), proteasome degradation of HIF-1α, reduced phagocytosis, and decreased p47/NOX2 activity (Fig. 8).

NAM is the amide version of vitamin B₃/niacin. Intracellularly, NAM is an essential source of NAD, which is a required cofactor involved in energy production, anabolic and catabolic pathways, and cellular homeostasis for all life forms (63). In sepsis, hypoxia and dysoxia at the cellular level increase NADH relative to NAD (64, 65), indicating a deficiency of intracellular biochemical molecules and enzymes to generate NAD. In endotoxin animal models, NAM improved survival and reduced organ damage by potentially dampening inflammation-related tissue injury (40, 66, 67). Because macrophages are pivotal to the inflammatory response during sepsis (68), we examined the effects of NAM in HMDMs treated with the Gram-negative pathogen-associated molecular pattern molecule LPS.

In Fig. 1, we identified distinct global genetic shifts in response to LPS and/or NAM, which resulted in increased production of NAD, predominantly through the salvage pathway. Although the NAM inhibited LPS-induced production of the de novo pathway rate-limiting enzyme IDO1, levels of NAD increased in this treatment condition, indicating that NAM may affect molecules along the de novo pathway. Moreover, inhibition of IDO1 did not reduce detection of NAD in response to LPS and/or NAM, suggesting that molecules along the de novo pathway may be used for biochemical cues endogenously or exogenously; for example, activation of the aryl hydrocarbon receptor (AHR) (69) or hydroxycarboxylic acid receptor 3 (HCAR3) (70).

PD-L1 and CD38 are cell surface markers of monocyte exhaustion, associated with depleted cellular NAD and the inflammatory response to sepsis (71). In HMDMs in Fig. 2A–H, we identify LPS-induced PD-L1, CD38, and Ag-presenting molecules (CD40, CD80) at the mRNA level and protein at the cell surface, which are each inhibited by NAM. Sepsis is associated with the increased production of proinflammatory mediators (72). In Fig. 2I–L, we show that NAM reduced production of proinflammatory cytokines and chemokines in response to LPS. These cell surface molecules and mediators are predominantly regulated by NF-κB (PD-L1 [73], CD40 [74], CD80 [75], CD38 [76], IL-1β [77], IL-6 [78], eotaxin [79], CXCL12 [80]), which is a transcription factor highly associated with the pathophysiology of sepsis (81). NAM may therefore dampen excessive inflammation by regulating NF-κB cell signals.

AKT is an upstream activator of NF-κB (45). In Fig. 3, we show that NAM inhibited AKT phosphorylation and its direct downstream phosphorylation target, FOXO1. Phosphorylation (s256) of FOXO1 induces its nuclear export (82). NAM decreased FOXO1 phosphorylation and increased production of FOXO1 in cocultures with LPS, suggesting that FOXO1 activity is enhanced. Increased FOXO1 activity has been associated with decreased endotoxin-induced kidney injury in mice (83). In COVID-19 patients, NAM may improve kidney injury (84). Possibly, FOXO1-activated macrophages have a role in these responses.

LPS-induced AKT also promotes NF-κB activation (45), and in Fig. 4 we show NAM-induced p65 ubiquitination, deacetylation, and reduced p65 production. This decreased NF-κB activity may play a role in the reduction of inflammatory mediators in LPS and NAM cocultures compared with LPS alone (Fig. 2I–L, Supplemental Figs. 1, 2). LPS-induced NF-κB promotes transcriptional expression of HIF-1α (85), which we show is inhibited by NAM at the mRNA, transcriptional, and translational levels (Fig. 5).

The mechanisms that control LPS-induced HIF-1α stability are not fully elucidated (85). By examining the NAM response, we identified posttranslational molecules involved in altering LPS-induced HIF-1α stability. Specifically, we found that NAM-induced PHD2 production (Fig. 6A), reduced unbound ubiquitin (Fig. 6C), increased HIF-1α ubiquitination (Fig. 6D), and promoted proteasome assembly (Fig. 6F), all of which contributed to NAM-mediated HIF-1α degradation in the presence of LPS. These data were additionally supported by IOX4 (Fig. 6B), IOX2 (Supplemental Fig. 3C), and MG-132 (Fig. 6E) blockade of NAM-induced HIF-1α degradation.

HIF-1α is a primary transcription factor in generating enzymes in glycolysis (2). LPS increased glycolysis as demonstrated by increased ECAR (Fig. 5F), and this response was reduced by NAM. LDHA regulates the conversion of pyruvate formed in glycolysis into lactate (50). In Fig. 5G, NAM decreased LDHA production in response to LPS activation, suggesting that NAM may reduce the accumulation of serum lactate, an independent predictor of mortality in sepsis patients (86).

Glycolysis is positively associated with phagocytosis (87). In Fig. 7A and 7B, NAM significantly reduced the endocytosis of dextran and phagocytosis of the bacterium P. aeruginosa. NAM-mediated mechanisms involved in this response may include changes in the ubiquitination (58, 59) and acetylation (60) of plasma membrane proteins and cytoplasmic microtubules, respectively. Activation of TLRs and/or phagocytosis promotes the recruitment of cytoplasmic proteins (e.g., p67^phox, p47^phox, p40^phox, Rac2) to plasma membrane proteins (e.g., gp91^phox and p22^phox) to form the NADPH oxidase, NOX2 (88), which is an identified target in the treatment of sepsis (89). In Fig. 7C, NAM antagonized the LPS-induced levels of the NOX2 NADPH oxidase stability protein, p47^phox. Moreover, AKT phosphorylates p47 to promote NADPH activation (90), which we show was reduced by NAM (Fig. 3), supporting possible NAM-mediated inhibition of NOX2 via reduced AKT activity and p47 protein production.

This study has limitations. First, HMDMs were differentiated with M-CSF, forming more of an M2 phenotype versus differentiation with GM-CSF, which generates more of an M1 phenotype (91). The type of macrophage baseline phenotype may affect the NAM response. Second, we only tested one form of endotoxin (Escherichia coli O111:B3). The source of LPS can have differential effects on glucose metabolism (92). Lastly, we did not explore the effects of NAM on macrophages in endotoxin-challenged mice.

In conclusion, NAM inhibits LPS-induced hypoxic cell signals and hypoxic metabolism in HMDMs. NAM also dampens excessive inflammation identified in murine models (40, 66, 67), sepsis patients (1, 3), and in patients with COVID-19 (84), suggesting a role for macrophages in the response. NAM or NAM-associated molecules (e.g., NR or NMN) may support the effects of drugs that target HIF-1α in preclinical and clinical trials (93, 94). Alternatively, the lack of activated macrophages capable of clearing pathogens may promote injury. Further exploration of NAM-induced crosstalk between the NAD salvage pathway and the de novo pathway, FOXO1 activation, and enzymes involved in deacetylation and ubiquitination of proteins in vitro and in vivo may provide insight into mechanisms that intercede in the host–pathogen response.

The authors have no financial conflicts of interest.

We thank Dr. Teruhiko Yoshida and Dr. Jeffrey B. Kopp for assistance and use of the Seahorse XF analyzer and Dr. Abdel Elkahloun for performing the microarray.