IFN Regulatory Factor 2 Inhibits Expression of Glycolytic Genes and Lipopolysaccharide-Induced Proinflammatory Responses in Macrophages

Rapid inflammatory response is a critical host defense against invading microbial pathogens. However, immune response should be also under tight controls in that excessive inflammation often leads to tissue injury and diseases, including acute lung injury (1–6). Macrophages are one of the most important inflammatory mediators. This type of cell demonstrates a remarkable plasticity, with the ability to undergo dynamic transitions across distinct functional phenotypes (7–10). The differential activation of macrophages has been the focus of recent studies, particularly with regard to the regulations by transcriptional factors (7, 11).

There have also been emerging concepts that metabolic reprogramming plays a critical role in regulating macrophage activation and immune response (12–15). For example, LPS-stimulated macrophages demonstrate enhanced aerobic glycolysis, whereas IL-4–activated cells undertake augmented fatty acid oxidation and glutaminolysis (13, 15–19). More importantly, the newly adopted metabolic programs are essential for macrophages to develop differential activation-associated phenotypic traits and functions (12, 13, 16, 17).

The IFN regulatory factor (IRF) family plays a myriad of roles in host responses to bacterial and viral infections (20–22). There have been a number of studies showing that the IRF family members, particularly IRF4 and IRF5, participate in differential macrophage activations (23–25). Although it is evident that IRFs can regulate these processes by directly trans-activating phenotypic markers in macrophages, it is less clear if this family participates in metabolic reprogramming and if so, what the role of the specific metabolic program is in IRF-mediated macrophage activations.

During our systemic effort to delineate the role of IRF family in inflammatory lung diseases, we found that IRF2 demonstrated unique characteristic distinct from other IRF members. We found that IRF2 underwent early proteasome-mediated degradation, which was necessary for HIF-1α–dependent glycolytic gene expression and glycolysis, in LPS-treated macrophages. We found that IRF2 was anti-inflammatory and this activity required IRF2 suppression of glycolysis. We also found that IRF2 promoted alternative activation of macrophages by IL-4. Together, our study indicates IRF2 as a new therapeutic target for treating inflammatory disorders associated with dysregulations of macrophage activations.

PAM3CSK4 was from InvivoGen. Mouse recombinant IL-4 was from PeproTech. Mouse recombinant IFN-β was from R&D Systems. Ultrapure LPS from Escherichia coli O111:B4, 2-deoxy-d-glucose (2-DG), cycloheximide, actinomycin D, and MG132 were from Sigma-Aldrich. Chetomin was from Santa Cruz Biotechnology. Bafilomycin A1 was from Cayman Chemical.

The mouse macrophage cell line MH-S was purchased from the American Type Culture Collection and cultured according to their instructions.

Mouse bone marrow–derived macrophages (BMDMs) were derived from bone marrow cells of C57BL/6 mice as previously described (26). Briefly, following erythrocyte lysis, bone marrow cells were cultured in DMEM containing 10% FBS and 50 ng/ml murine M-CSF (R&D Systems) for 5 d. The differentiated cells were then split and plated for following experiments. Peritoneal macrophages were elicited in 8-wk-old mice by i.p. injection of 1 ml 4% Brewer thioglycollate (Sigma-Aldrich). Peritoneal cells were harvested 4 d later by lavage and plated for 2 h, followed by extensive wash to remove nonadherent cells. The adherent cells were used as peritoneal macrophages. The animal protocol was approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee.

Small interfering RNA (siRNA) transfection was performed using HiPerFect reagents (Qiagen) according to the manufacturer’s instructions. ON-TARGETplus negative control siRNA pool, specific mouse IRF2 siRNA pool and individual IRF2 siRNA oligos, KLF4 siRNA pool, and Accell negative control and IRF2 siRNAs were from Dharmacon.

RNA sequencing was performed by the University of Alabama at Birmingham Heflin Center for Genomic Science. RNA sequencing data were submitted to the Gene Expression Omnibus and are unrestrictedly accessible with accession number GSE106895 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE106895).

mRNA levels were determined by real-time PCR using SYBR Green Master Mix Kit (Roche). Primer sequences for mouse genes were: tubulin α1: sense, 5′-GGATGCTGCCAATAACTATGCTCGT-3′, antisense, 5′-GCCAAAGCTGTGGAAAACCAAGAAG-3′; IRF2: sense, 5′-ATGAAGAGAACGCAGAGGGGAG-3′, antisense, 5′-GCTTGTTGGAGGTGACAAAGGT-3′; IL-1β: sense, 5′-AAGGAGAACCAAGCAACGACAAAATA-3′, antisense, 5′-TTTCCATCTTCTTCTTTGGGTATTGC-3′; iNOS: sense, 5′-ATCTTTGCCACCAAGATGGCCTGG-3′, antisense, 5′-TTCCTGTGCTGTGCTACAGTTCCG-3′; ATF3: sense, 5′-ACACCTCTGCCATCGGATGTCCTCT-3′, antisense, 5′-ATTTCTTTCTCGCCGCCTCCTTTTC-3′; CEBP-δ: sense, 5′-AACTTGATTCCTCGTTGCCTCTACTTTC-3′, antisense, 5′-CCGCAAACATTACAATTACTGGCTTTT-3′; HK2: sense, 5′-CTTACCGTCTGGCTGACCAACAC-3′, antisense, 5′-CTCCATTTCCACCTTCATCCTTCT-3′; PFKFB3: sense, 5′-CTGACAAAGGAAGGTGGACAGATTG-3′, antisense, 5′-CCACAACAGTAGGGTCGTCACAGAC-3′; Arg-1: sense, 5′-TGACTGAAGTAGACAAGCTGGGGAT-3′, antisense, 5′-CGACATCAAAGCTCAGGTGAATCGG-3′; YM-1: sense, 5′-ATGAAGCATTGAATGGTCTGAAAG-3′, antisense, 5′-TGAATATCTGACGGTTCTGAGGAG-3′; FIZZ1: sense, 5′-AGGTCAAGGAACTTCTTGCCAATCC-3′, antisense, 5′-AAGCACACCCAGTAGCAGTCATCCC-3′; MGL1: sense, 5′-GCTTCGAAAAAGGGATCAGTTCTCT-3′, antisense, 5′-CTCTTCTCCACTGTGCTCTCCAGAG-3′; KLF4: sense, 5′-TACCCTCCTTTCCTGCCAGACCA-3′, antisense, 5′-GCCACGACCTTCTTCCCCTCTTT-3′. To calculate fold change in the expression of these genes, ΔCt = Ct of tubulin − Ct of individual genes (where Ct denotes cycle threshold) was first obtained. ΔΔCt = ΔCt of treated groups − ΔCt of untreated control groups was then obtained. Fold change was calculated as 2^ΔΔCt, with control groups as 1 fold.

Western blotting was performed as previously described (27). Mouse anti–α-tubulin Ab was from Sigma-Aldrich. Rabbit anti-IRF2, lamin B1, Arg-1, and iNOS Abs were from Santa Cruz Biotechnology. Rabbit anti-HK2, p65, p-p65, ERK1/2, p-ERK1/2, p38, p-p38, JNK, p-JNK, IκBα, KLF4, and STAT6 Abs were from Cell Signaling Technology. Rabbit anti–HIF-1α Ab was from Novus Biologicals. Rabbit anti-PFKFB3 Ab was from Abcam. Goat anti–mIL-1β Ab was from R&D Systems.

Levels of secreted YM-1 in cell culture supernatants were determined using DuoSet ELISA development kits (R&D Systems) according to the manufacturer’s instructions.

Bacterial killing assay was performed as described previously (26, 28). Briefly, 0.1 × 10⁶ CFU/ml Escherichia coli (BL21DE3pLysS) was added to macrophages in 96-well plates and incubated at 37°C for 1 h. Supernatants were serially diluted and plated on Luria broth agar plates. The plates were incubated overnight at 37°C, and bacterial colonies were counted. Data were presented as CFU per milliliter.

XF-24 Extracellular Flux Analyzer (Seahorse Bioscience) was used for real-time recording of extracellular acidification rate (ECAR). Briefly, BMDMs were seeded in Seahorse XF-24 microplates (1.5 × 10⁵ cells per well), and treated with or without 100 ng/ml LPS for 6 h. Before analysis, the cells were incubated in ECAR media for 1 h at 37°C in room air. Cells were sequentially treated with 10 mM glucose, 2 μg/ml oligomycin, and 5 mM 2-DG. Real-time ECARs were recorded according to the manufacturer’s manual.

Glucose levels in the cell culture supernatants were determined using a glucose colorimetric/fluorometric assay kit (BioVision) according the manufacturer’s instructions. Glucose consumption was calculated as (glucose level of fresh medium − glucose levels of supernatants)/micrograms protein/incubation hours.

Extracellular levels of lactate were determined using lactate assay kit (BioVision) according to the manufacturer’s instructions.

Chromatin immunoprecipitation (ChIP) assays were performed as previously described (29). Briefly, BMDMs were fixed with 1% formaldehyde in PBS for 10 min and collected in RIPA buffer. Genomic DNA was then sheared by sonication to length around 200–500 bp. One percent of the cell extracts was taken as input, and the rest of the extracts were incubated with either rabbit anti–HIF-1α Ab, anti-KLF4 Ab, or rabbit IgG overnight, followed by precipitation with protein G agarose beads (Invitrogen). Genomic DNA in the immunocomplexes was purified using Qiagen miniprep columns (Qiagen). Primer sequences to specifically amplify the HK2 promoter or PFKFB3 promoter region spanning the putative HIF-1α binding site/hypoxia-response element (HRE) were: HK2-HRE: sense, 5′-GTTGAGCTACAATTAAGATGAGAATCA-3′, antisense, 5′-CCCAAGCAGGCGGCGGGAGAGA-3′; PFKFB3-HRE: sense, 5′-GTGTGTGTGTGTAGGGGTGGAGGA-3′, antisense, 5′-ACGTGGAGAGAAGGGTGGGCAAGG-3′. Primer sequences to amplify the KLF4 promoter region spanning the putative IRF2 binding elements (IRF-E) were: KLF4-IRF-E-#1 sense, 5′-GGTAGTGGGGAATGGGAAAAGGAGT-3′, antisense, 5′-ATAAAGAGGTGAAGCGGCGAGGTAA-3′; KLF4-IRF-E-#2 sense, 5′-AGGTGTTCTTTTTGTTGTTCCTTTGT-3′, antisense, 5′-TTCATCCTTTCTCTTGGTTTTGGC-3′.

One-way ANOVA followed by the Bonferroni test was used for multiple group comparisons. The Student t test was used for comparison between two groups. A p value < 0.05 was considered statistically significant.

To determine the role of IRF2 in inflammatory response, we initially characterized its expression in LPS-treated macrophages. As shown in Fig. 1A, the protein level of IRF2 was rapidly decreased prior to its recovery at 6 h after LPS stimulation. IRF2 protein level in macrophages treated with the TLR2 ligand Pam3CSK4 demonstrated a similar pattern of alterations, although at a delayed kinetics (Fig. 1B). Intriguingly, IRF2 reduction appeared coincident with the induction of the proinflammatory cytokine IL-1β in LPS- and Pam3CSK4-treated cells (Fig. 1A, 1B). To determine if the decrease in the IRF2 protein level was caused by a downregulated IRF2 transcription, we examined IRF2 mRNA levels and found surprisingly that IRF2 was transcriptionally upregulated in LPS-treated macrophages (Fig. 1C). In the aggregate, these data clearly suggest that IRF2 expression undergoes a dynamic regulation at both the transcriptional and posttranscriptional levels in response to LPS stimulation. To delineate the mechanism underlying the early decline in IRF2 protein in LPS-treated macrophages, we examined its stability by blocking cellular translation and found that IRF2 was short-lived in that nearly half of the protein was degraded within 3 h (Fig. 1D). More importantly, LPS treatment promoted IRF2 degradation (Fig. 1D). Of note, both findings resembled what was observed with the principal anti-inflammatory mediator IκBα in LPS-treated macrophages (Fig. 1D) (30). Next, we demonstrated that the physiological turnover of the IRF2 protein was mediated by proteasomes as the proteasome inhibitor MG132 completely blocked this process (Fig. 1E). The proteasome-dependent degradation was found to be solely responsible for the LPS-induced early downregulation of the IRF2 protein because MG132, but not the lysosome inhibitor bafilomycin A1, completely abolished this phenomenon (Fig. 1F, 1G). As expected, autophagy mediator LC3-II was accumulated because of lysosome inhibition in bafilomycin A1 treated cells (Fig. 1G). Additionally, to determine if LPS-induced IRF2 transcription accounted for the later restoration of the protein expression, we inhibited cellular transcription with actinomycin D and found that the IRF2 protein level failed to recover when transcription was halted in LPS-treated macrophages (Fig. 1H). Collectively, these data suggest that the expression of IRF2 bears the trait typical of that of key inflammatory regulators that function in a negative feedback manner in response to LPS.

The resemblance of IRF2 expression dynamics to that of typical negative regulators in inflamed macrophages prompted us to ask if this protein also functioned in a similar manner to regulate immune response (2, 30). To test this hypothesis, we knocked down IRF2 in BMDMs and examined the proinflammatory cytokines and mediators in these cells after LPS stimulation. We found that IRF2 knockdown significantly augmented the LPS-induced expression of IL-1β, iNOS, TNF-α, and IL-12 at both mRNA and protein levels (Fig. 2A–D, data not shown). The increased expression of the proinflammatory mediators was also observed in LPS-treated peritoneal macrophages and macrophage line MH-S cells with IRF2 knockdown, suggesting a general role of IRF2 in macrophages in response to LPS (Supplemental Fig. 1A, 1B). To rule out the possibility that the effect of IRF2 knockdown in LPS-treated macrophages was caused by any off-target effect from a single IRF2 siRNA, we replicated the experiments with siRNAs separately targeting two independent locations in the IRF2 transcripts. As shown in Supplemental Fig. 1C, both siRNAs efficiently knocked down IRF2. More importantly, both siRNAs enhanced the expression of IL-1β and iNOS to the same levels, indicative of the specificity of the findings with IRF2 knockdown. This conclusion was further reinforced after achieving similar results using a different type of IRF2 siRNA molecule, Accell siRNAs, which required drastically reduced transfection reagents (Supplemental Fig. 1D). There was also evidence showing that IRF2 could competitively inhibit IRF1-mediated type I IFN response that is known to be present in LPS-treated macrophages (1). To determine the role of IRF1 and the resulting IFN activation in the IRF2 anti-inflammatory activity, we cotreated the macrophages with IFN-β. As expected, IFN-β significantly induced IRF1-dependent expression of ISG15 and Gbp2 (Supplemental Fig. 2A). However, IFN-β had no effect on the increased expression of IL-1β and iNOS in LPS-treated IRF2 knockdown cells (Supplemental Fig. 2B, 2C). Together, these data indicate that the anti-inflammatory effect of IRF2 is not achieved through regulating IRF1-associated type I IFN response. Additionally, as functional activities of inflamed macrophages, such as bacteria killing capability, are often commensurate with the levels of inflammatory response (10, 31), we investigated if IRF2 knockdown promoted such activities of the cells. Consistent with LPS-induced expression of iNOS, we found that the bactericidal activity of LPS-treated macrophages was significantly increased when IRF2 was knocked down (Fig. 2E, 2F). Collectively, the results suggest that IRF2 is a negative regulator of immune response in macrophages. Moreover, it appears that the early degradation of the IRF2 protein is to facilitate macrophages mounting swift proinflammatory reaction to microbial infections.

We had demonstrated that knockdown of IRF2 promoted the inflammatory response to LPS in macrophages. Next we sought for the mechanisms underlying these observations. We first examined the principal signaling cascade upon LPS stimulation, which includes phosphorylation and activation of NF-κB subunit p65 and MAPKs (1, 3). As shown in Fig. 3A, IRF2 knockdown exerted no effect on the phosphorylation of p65, ERK1/2, p38, or JNK in LPS-treated macrophages, suggesting that the immediate signaling events downstream of LPS engaging with its receptor TLR4 were not the targets of IRF2. Furthermore, as an essential step of the LPS-induced proinflammatory mediators, including IL-1β and iNOS, p65 bindings to the promoters of these genes remained unchanged when IRF2 was knocked down (Fig. 3B). Additionally, we examined if IRF2 had a role in the expression of ATF3 and CEBP-δ, two well-recognized key regulators of the LPS-induced inflammatory response (32–34), and found that the level of neither ATF3 nor CEBP-δ was affected in the naive and LPS-treated macrophages with IRF2 knockdown (Fig. 3C, 3D). Taken together, these data suggest that IRF2 may not have a direct role in LPS-induced proximal signaling cascade. These results also indicate that IRF2 does not act via the investigated regulators (i.e., ATF3 and CEBP-δ) that have a major role in LPS-induced inflammatory responses.

The lack of effects of IRF2 knockdown on the LPS-induced cytoplasmic signaling events led us to reason that IRF2, a transcriptional factor itself, may regulate inflammatory responses at the transcriptional level. Therefore, we performed RNA sequencing analyses on macrophages with or without IRF2 knockdown. Among genes that were upregulated in the IRF2 knockdown cells, we noticed a clear presence of a number of glycolytic mediators, such as HK2 and PFKFB3 (Fig. 4A). As it had been increasingly recognized that augmented aerobic glycolysis was essential to the development of inflamed phenotype in LPS-treated macrophages (12, 13, 18), we hypothesized that IRF2 promoted inflammation by boosting glycolysis in these cells. At first, we validated the RNA sequencing results by real-time PCR and showed that the expression of HK2 and PFKFB3 was increased in the naive macrophages with IRF2 knockdown (Fig. 4B). We also found that LPS induction of these genes was further augmented in IRF2 knockdown cells (Fig. 4B–D). To determine the metabolic consequence of the elevated glycolytic mediators, we examined ECAR, a surrogate of lactate production rate and intracellular glycolytic level, using the Seahorse bioanalyzer. We found that ECAR in naive IRF2 knockdown macrophages was markedly increased compared with that in control siRNA-transfected cells (Fig. 4E). Furthermore, LPS-augmented ECAR was again enhanced in the macrophages with IRF2 knockdown compared with that in the control cells (Fig. 4E). These data were in line with the greater levels of lactate productions and glucose consumptions found in the naive and LPS-treated macrophages with IRF2 knockdown compared with those in control cells (Fig. 4F, 4G). Collectively, our findings suggest that IRF2 suppresses glycolysis in macrophages by inhibiting the expression of glycolytic mediators. To further determine if the augmented glycolysis observed in IRF2 knockdown macrophages accounted for the aggravated LPS-induced inflammatory responses in these cells, we blocked cellular glycolysis with hexokinase inhibitor 2-DG and found that the increases in the expression of proinflammatory mediators in IRF2 knockdown macrophages were abolished (Fig. 4H). In summary, our data suggest that IRF2 suppresses inflammation by inhibiting the expression of glycolytic genes and thereby cellular glycolysis.

As hypoxia-inducible factor 1α (HIF-1α) is the master regulator of glycolytic gene expression and glycolysis (12, 13, 35), we asked if HIF-1α played a role in the IRF2-modulated events. First, we found that the augmentation in the LPS-induced glycolytic genes HK2 and PFKFB3 in IRF2 knockdown macrophages was abrogated when the cells were treated with specific HIF-1α inhibitor chetomin (Fig. 5A). In line with this observation, the elevated expression of the proinflammatory mediators IL-1β and iNOS after IRF2 was knocked down was also abolished by chetomin in LPS-treated macrophages (Fig. 5B). Together, these data suggest that IRF2-regulated glycolytic gene expression and inflammation is mediated by HIF-1α. To investigate the mechanism by which IRF2 regulated these HIF-1α–mediated events, we examined the levels of HIF-1α and found unexpectedly that HIF-1α expression in both naive and LPS-treated macrophages remained largely unaffected by IRF2 knockdown (Fig. 5C), suggesting that IRF2 may be involved in the transcriptional activity of HIF-1α. To test this hypothesis, we determined HIF-1α binding by ChIP assays to the promoters of its targets, one of the critical steps that decide the transactivation of these genes. As shown in Fig. 5D, there were significant increases in the bindings of HIF-1α to the promoters of HK2 and PFKFB3 upon LPS stimulation. More importantly, the HIF-1α bindings were further enhanced in LPS-treated IRF2 knockdown cells (Fig. 5D). In all, these results indicate that IRF2 inhibits inflammation by suppressing HIF-1α–dependent expression of glycolytic genes and glycolysis.

As a number of anti-inflammatory regulators that function by the negative feedback mechanism also participate in endotoxin tolerance, itself a continuation of acute immune response (36), we asked if IRF2 had such an activity. To do this, macrophages were treated with LPS for 24 h (first LPS) to induce tolerance. The cells were then cultured in fresh media with or without LPS for 6 h (second LPS). We found that tolerant macrophages were nearly unresponsive to the second LPS stimulation as they expressed minimal proinflammatory mediators IL-1β and iNOS (Fig. 6A). However, the LPS tolerant state was partially relieved in macrophages when IRF2 was knocked down as these cells produced markedly more IL-1β and iNOS (Fig. 6B). These data suggest that IRF2 is required for the establishment of endotoxin tolerance in macrophages. To delineate the underlying mechanism, we examined IRF2 expression in the tolerant macrophages. In contrast to the LPS-induced early degradation of the IRF2 protein in the naive macrophages, IRF2 levels were largely unchanged at the same early time point of LPS treatment of the tolerant cells (Fig. 6C). Given that IRF2 inhibited glycolytic gene expression, these data suggest that the failure of IRF2 downregulation may hinder the upregulation of these genes and the ensuing augmentation of glycolysis in LPS-treated tolerant macrophages. As anticipated, we did find that LPS induction of glycolytic genes HK2 and PFKFB3 was literally eliminated in tolerant macrophages compared with naive cells (Fig. 6D). However, the expression of HK2 and PFKFB3 was significantly increased in tolerant macrophages when IRF2 was knocked down (Fig. 6E). Consistent with this finding, glycolysis, as reflected by ECAR and lactate production with and without the second LPS treatment, was markedly augmented in tolerant macrophages with IRF2 knockdown (Fig. 6F, 6G). Furthermore, we found that blocking glycolysis by 2-DG completely abolished the partial recovery of the LPS sensitivity in tolerant macrophages with IRF2 knockdown (Fig. 6H). In summary, these data indicate that IRF2 suppression of glycolysis is required for endotoxin tolerance in macrophages.

Many negative regulators of inflammation participated in alternative macrophage activation, which, together with the proinflammatory one, occupied the two opposite spectrums of macrophage phenotype (37). As we had demonstrated that IRF2 was a negative regulator of proinflammatory response, we next asked if this transcription factor also played a role in alternative activation of macrophages. First, we investigated IRF2 levels in IL-4–treated macrophages (M(IL-4)). As shown in Fig. 7A–C, IL-4 increased the expression of IRF2 at both the mRNA and protein levels. These data suggest that IRF2 may regulate IL-4–induced macrophage activation. Additionally, there were elevated levels of IRF2 in the nuclei of M(IL-4) (Fig. 7D), indicative of its location of action in the regulation of M(IL-4) activation.

We then determined if IRF2 participated in M(IL-4) activation. As shown in Fig. 7E–H, there were significant decreases in IL-4–induced markers of alternatively activated macrophages at both mRNA and protein levels when IRF2 was knocked down. These data suggest that, in contrast to being anti-inflammatory, IRF2 promotes macrophage alternative activation.

To delineate the mechanism by which IRF2 promoted IL-4–induced alternative activation of macrophages, we again interrogated the RNA sequencing results and found that KLF4, a crucial regulator of M(IL-4) activation (38), was downregulated in macrophages with IRF2 knockdown. We validated this finding by real-time PCR. As shown in Fig. 8A, KLF4 transcription was decreased when IRF2 was knocked down. Furthermore, IL-4–induced KLF4 was also diminished in IRF2 knockdown macrophages (Fig. 8A). These data were consistent with the attenuated induction of the KLF4 protein by IL-4 in macrophages with IRF2 knockdown, although KLF4 at basal levels was not detectable by the Ab (Fig. 8B). We further determined how IRF2 transcriptionally regulated KLF4. We found that there were three typical IRF2 binding sites within 2 kb upstream of the KLF4 transcription start site (Fig. 8C). We performed ChIP assays with anti-IRF2 Ab pull-down and examined IRF2 bindings to the predicted sites. As shown in Fig. 8C, we found that IRF2 binding to these sites was significantly increased in IL-4–activated macrophages, suggesting that IRF2 directly mediates IL-4 induction of KLF4. We next confirmed the pro-M(IL-4) activity of KLF4 in that we found that the expression of alternatively activated macrophage markers was markedly reduced when KLF4 was knocked down (Fig. 8D, 8E). In light of these findings, we asked if the promotion of M(IL-4) activation by IRF2 was dependent on its induction of KLF4. To test this hypothesis, we determined the effect of IRF2 in macrophages with or without KLF4 knockdown. As shown in Fig. 8F, the attenuation of M(IL-4) activation by IRF2 siRNAs was largely abolished in macrophages with prior KLF4 knockdown. In addition, consistent with the previous finding that KLF4 was a key coactivator for IL-4–induced STAT6 transcriptional activation (38), we found that STAT6 bindings to the promoters of alternative macrophage markers, but not its phosphorylation or nuclear translocation, were diminished in IRF2 knockdown macrophages upon IL-4 stimulation (Supplemental Figs. 3A–C). In all, these data suggest that IRF2 promotes alternative activation of macrophages by transactivation of KLF4.

The IRF family members have been shown to demonstrate a versatility of activities in regulating immune responses (20–25). However, these investigations were generally focused on mechanisms at the transcriptional and epigenetic levels. In this study, we not only found that IRF2 participated in the differential activation of macrophages, but also established the causal link between IRF2-regulated glycolysis and its anti-inflammatory activity. Our study has thus reinforced the increasingly recognized role of cellular metabolism in the regulation of immunological response, which has been manifested in a number of immune dysregulation-associated diseases. Among this ample evidence is the necessity of glycolytic homeostasis in the maintenance of a balanced immune system. Either hyper- or hypoglycolytic responses in various immune cells, including macrophages, monocytes, dendritic cells, and T lymphocytes, have been implicated in clinically relevant immunological pathologies, such as sepsis and diabetes (39–45). This has led to the concept of modulating glycolysis to experimentally treat immune disorders (46, 47). Given our findings revealing the intersection of IRF2 and glycolysis, this study has also indicated potential novel therapeutic targets in dealing with this group of diseases.

We have shown that IRF2 differentially impacts the recruitments of HIF-1α and KLF4 to the promoters of their respective target genes in that IRF2 inhibited the bindings of HIF-1α, but enhanced those of KLF4, in activated macrophages. However, the mechanism by which IRF2 achieves such specificity remains to be elucidated. One appealing hypothesis is that IRF2 may undergo distinct posttranslational modifications in differentially activated macrophages. These unique modifications may dictate the interaction of IRF2 with either transcriptional coactivators or corepressors, which in turn may determine the accessibility of the promoters of individual target genes to these specific transcription factors and the ensuing transcriptional robustness. Therefore, further studies focusing on identification of IRF2 modifications appear to be warranted.

Additionally, posttranslational modifications are also likely involved in the early degradation of IRF2 in LPS-treated macrophages, similar to that observed with IκBα (2, 30). These findings continuingly advance our understanding that this type of maneuver is a commonly harnessed mechanism by which immune cells overcome early hurdles to launch a robust inflammation that is needed to combat pathological microbial invasion. Furthermore, the observation that the recovery of the IRF2 expression via elevated transcription in the later stage of inflammatory response seems to place IRF2 squarely into the category of inflammation rheostats. This group of molecules has been frequently found to be critical to immune homeostasis, dysregulations of which are known causes for inflammatory disorders (5, 6, 48). Given the anti-inflammatory role of IRF2 and its proalternative macrophage activation, it is surely enticing to delineate its role in vivo in both type I and type II inflammation animal models. Besides, we will also gain mechanistic insight through investigating IRF2 levels and genetic variants in human immune diseases, such as acute lung injury and asthma.

Also importantly, the participation of IRF2 in endotoxin tolerance, a state of immunoparalysis that is a primary cause for secondary infections in sepsis patients, seems to justify the comparison of the dynamics of IRF2 expression in those subjects at the time of admission and recovery. Although recent studies showed that macrophages in tolerant state were characteristic of metabolic catastrophe, including defects in glycolysis, the causes behind these dysregulations remained unclear (17, 39, 49). Our study not only confirmed the contribution of glycolytic hyporesponsiveness to endotoxin tolerance, but it also shed light on the underlying mechanism, for example, the absence of IRF2 degradation leading to failure of glycolytic gene induction and substantial glycolytic augmentation in tolerant macrophages, which is metabolically required for robust inflammatory response. Therefore, our findings identify a new therapeutic target for treating disorders associated with immunoparalysis.

It should be noted that early studies also showed that IRF2-deficient cells or mice were protective from LPS challenge (50–52). These findings, although lacking mechanistic details, seem inconsistent with our results and some other reports that IRF2 was antagonistic of type I IFN response (22, 53, 54). However, these discrepancies may arise from a series of scenarios that could occur in global knockout animals, including differences in genetic background between the knockout mice and the wild-type counterparts, potential phenotypic alternations caused by constitutional ablation of IRF2, and likelihood of unintended deletion of functional noncoding sequences, such as microRNAs, which are located at the same locus as IRF2. Regardless, our conclusion was firmly based on multiple genetic and pharmaceutical approaches, such as replicating experiments using siRNAs with independent sequences to target the same gene expression and confirming the hierarchy of each regulatory event with specific inhibitors.

Finally, the current study is meticulous in delineating IRF2 regulation of glycolysis and glycolysis-dependent immune response in macrophages. The rationale for such an emphasis on one type of metabolic pathway was not a bias, but was rather informed by our RNA sequencing analysis that revealed the predominant targets of IRF2 are genes associated with glycolysis. Nevertheless, given the rapidly accumulating evidence clearly showing that other core metabolisms, such as fatty acid oxidation and glutaminolysis (13, 15–19), are also involved in the activation of a variety of immune cells, there is a good likelihood that IRF2 may interact with those metabolic elements, either directly or indirectly, to control the response of macrophages in pathological settings. This hypothesis remains an evidently interesting point in future studies.

The authors have no financial conflicts of interest.