The STAT4/MLL1 Epigenetic Axis Regulates the Antimicrobial Functions of Murine Macrophages

Macrophage polarization helps to guide inflammatory responses through regulating the kinetics of antimicrobial responses and reparative processes that are necessary for the restoration of healthy physiology (1). Following the trafficking of immature peripheral blood monocytes through the endothelial blood vessel wall, monocyte-derived macrophages can polarize to various effector lineages based on the cytokine milieu within the inflamed tissue. Classically activated (M1) macrophages are generated in response to IFN-γ, produce proinflammatory cytokines, and upregulate the cellular machinery necessary for the phagocytosis and killing of microbes. In contrast, alternatively activated (M2) macrophages are generated in response to IL-4 and/or IL-13 and promote tissue repair in areas damaged by previous inflammatory responses (2). M1 and M2 macrophages play key roles in the initiation and resolution of inflammation, and the gradual switch in generalized macrophage responses from M1 to M2 is a key component of many reparative processes, such as effective wound healing (3, 4).

Aberrant macrophage responses can also drive immunopathologies through the effects of chemokines, cytokines, and antimicrobial enzymes. For example, M1 macrophage activation in response to pathogen-associated or damage-associated molecular patterns is a key component of the “cytokine storm” observed during life-threatening inflammatory responses (5, 6). Physiological stress caused by overexpression of these proinflammatory mediators in response to microbial products or other danger signals can lead to significant morbidity and mortality in patients (7). Improper M2 activation can also cause harm; as an example, M2 macrophages can promote vascularization of tumors and suppression of cytotoxic immune cell functions, supporting cancer growth (8). Therefore, to develop more efficacious treatments aimed at modulating macrophage phenotype/function to improve patient outcomes, it is critical to develop a better understanding of the molecular mechanisms governing macrophage polarization, plasticity, and stability of functional lineage-commitment decisions.

Chromatin dynamics plays a central role in guiding macrophage polarization decisions through regulating the transcriptional accessibility of key lineage-specific genes. Regulation of chromatin-modifying enzyme (CME) expression and function results in changes to the posttranslational modifications on the unstructured tails of histone proteins used to package DNA within the nucleus. The summation of the histone modifications within the promoter region of a given gene, the “histone code,” participates in governing the expression level of the gene in question. Chromatin dynamics governed by CMEs help to guide many aspects of leukocyte development and effector function, and many inflammatory responses are characterized by changes in CME expression and histone modifications. For example, expression of the histone demethylase KDM6B/JMJD3 is essential for the development of M2 macrophages because of its ability to remove repressive histone modifications from the promoter regions of key M2 gene loci (9, 10). In the case of immunopathology, postseptic macrophages have been shown to exhibit decreased activating histone modifications (including acetylation and methylation) in the promoter regions of key M1 proinflammatory genes, resulting in susceptibility to secondary microbial infections following recovery from sepsis (11). Therefore, CMEs can serve as an epigenetic link between extracellular signals and immune cell activation.

The histone methyltransferase KMT2A/mixed-lineage leukemia 1 (MLL1) is a CME that has been shown to play a critical role in regulating leukocyte activation through its ability to add activating methylation marks (i.e., H3K4me3) in gene promoter regions. In the case of adaptive immunity, MLL1 regulates T_H1 cell proliferation through regulation of IL-12 responsiveness (12); it can also regulate the development of T_H2 memory by stabilizing expression of GATA3 through H3K4me3-dependent chromatin remodeling (13). MLL1 activity can also be driven by NF-κB–dependent signals, such as TNF-α and LPS, both of which are present in tissue microenvironments that promote classical activation of macrophages (14). These reports suggest a key role for MLL1-dependent chromatin remodeling in regulating the quality of cellular inflammatory processes. Supporting this concept is the recent description of MLL1 expression in human monocyte-derived macrophages as being strongly correlated with classical activation in these cells (15).

Based on previous reports linking MLL1 expression and function with lineage commitment decisions in immune cells, we sought to investigate its role in regulating macrophage immune responses. By utilizing an inducible Cre-loxp system to specifically target MLL1 gene deletion in phagocytic cells, including macrophages (Lys2Cre^+/− MLL1^fx/fx), we attempted to dissect the contribution of MLL1 and H3K4me3 chromatin remodeling to macrophage effector responses to microbial pathogens.

Wild-type C57BL/6 mice were obtained from Taconic (Hudson, NY). Myeloid-specific MLL1-knockout animals were generated through breeding of an MLL1^fx/fx line (16) with commercially available mice that express Cre under the control of the Lys2 promoter (The Jackson Laboratory, Bar Harbor, ME) to generate the Lys2-Cre MLL1^fx/fx line. Mice were housed under specific pathogen–free conditions at the Unit for Laboratory Animal Medicine of the University of Michigan and treated in accordance with the guidelines of the animal ethical committee. All animal experiments were approved by the University of Michigan Institutional Animal Care and Use committee.

Bone marrow cells from indicated animals were isolated and cultured in L cell–conditioned media (RPMI 1640, 20% v/v FBS, 30% v/v L929-conditioned supernatant) for 7 d. At the end of the differentiation culture, adherent cells were harvested, replated, and cultured in RPMI 1640 with 10% v/v FBS and penicillin/streptomycin. IFN-γ, IL-4, IL-12 (all from Shenandoah Biotechnology, Warwick, PA), IL-23, IL-27, IL-28β (all from R&D Systems, Minneapolis, MN), IFN-α3, and IFN-β (both from PeproTech, Rocky Hill, NJ) were used at 10 ng/ml. LPS (Sigma-Aldrich, St. Louis, MO) was used at 100 ng/ml, and CpG and pI:C (InvivoGen, San Diego, CA) were used at 10 μg/ml.

RNA was isolated from cultured cells using TRIzol Reagent (Thermo Fisher Scientific, Waltham, MA) and converted to cDNA using iScript Supermix (Bio-Rad, Hercules, CA), both according to the manufacturer’s specifications. Quantitative PCR (qPCR) was performed on an ABI 770 sequence detector system using SYBR Green Master Mix or Universal PCR Master Mix (all from Thermo Fisher Scientific). Primers for the indicated mRNAs were obtained from Thermo Fisher Scientific, including primers specifically spanning the disrupted exons 8/9 in the Lys2-Cre MLL1^fx/fx–transgenic animal model (catalog no. Mm01179245_m1). Primers for chromatin immunoprecipitation (ChIP) were as follows: Nos2 proximal (−206 bp from exon 1): 5′-GTCCCAGTTTTGAAGTGACTACG-3′ and 5′-GTTGTGACCCTGGCAGCAG-3′, Nos2 distal (−1059 bp from exon 1): 5′-CCAACTATTGAGGCCACACAC-3′ and 5′-GCTTCCAATAAAGCATTCACA-3′, Tnfa proximal (−261 bp from exon 1): 5′-TCCTGATTGGCCCCAGATTG-3′ and 5′-TAGTGGCCCTACACCTCTGT-3′, Tnfa distal (−560 bp from exon 1): 5′-CTCTCAAGCTGCTCTGCCTT-3′ and 5′-GGACATCCATGGGGGAGAAC-3′, and Stat4 (−100 bp from exon 1): 5′-CCAGGTCTGTGATTGGCTCT-3′ and 5′-ACATCCAGAGGACCCCTTCC-3′.

Cultured cells were processed with cell lysis buffer (Cell Signaling Technology, Danvers, MA) and LDS sample buffer (Thermo Fisher Scientific). Samples were run on precast 4–12% Bis-Tris gels and transferred to nitrocellulose membranes (iBlot; both from Thermo Fisher Scientific) prior to incubation with the indicated Abs: H3K4me1 (Abcam, Cambridge, MA), H3K4me2 (Cayman Chemical, Ann Arbor, MI), H3K4me3 (Abcam), GAPDH (Santa Cruz Biotechnology, Dallas, TX) and phospho- and total STAT1, STAT2, and STAT4 (Cell Signaling Technology). Blots were visualized using SuperSignal West Pico chemiluminescent reagent (Thermo Fisher Scientific), and images were captured on a ChemiDoc Imaging System (Bio-Rad).

Phagocytosis was assayed using fluorescent Escherichia coli bioparticles, according to the manufacturer’s protocol (Vybrant Phagocytosis Assay; Thermo Fisher Scientific) (17). For bacterial killing assays (18), macrophages were seeded in 96-well plates, with a set of replicate conditions within each plate, and treated with cytochalasin D for 30 min. Cells were then challenged with a 10:1 multiplicity of infection of live group A Streptococcus. The plates were centrifuged to synchronize bacterial contact with the monolayer and incubated for 30 min. Thereafter, all media were removed, and plates were washed gently to remove nonphagocytosed bacteria. Following washing, 100 μl of PBS/0.5% saponin was added, and plates were incubated for 10 min. Serial dilutions of lysates were plated onto trypticase soy agar–blood plates, and CFU were counted 24 h later. Phagosome acidification was assayed in cultured cells using dextran particles conjugated with a pH-sensitive fluorescent reporter molecule, according to the manufacturer’s protocol (pHrodo Red Dextran; Thermo Fisher Scientific).

Cells were fixed with paraformaldehyde, and cell pellets were lysed and sonicated following a standard protocol (19). Abs to H3K4me3 (Abcam) were used for immunoprecipitation along with isotype-control Abs in parallel samples. DNA/Ab complexes were precipitated with Protein A Sepharose/Salmon Sperm DNA slurries, and enriched genomic DNA was purified using phenol/chloroform phase separation. qPCR was performed on enriched DNA samples, as described above.

Isolated RNA from bone marrow–derived macrophages (BMDMs) stimulated with the indicated cytokines/TLR ligands was assayed using a Mouse Gene ST 2.1 GeneChip system (Thermo Fisher Scientific). The DNA Sequencing Core at the University of Michigan performed quality control analysis, Illumina operation, data collection, and statistical analysis. The microarray data discussed in this article were submitted to the National Center for Biotechnology Information Gene Expression Omnibus (20) under accession number GSE82109 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE82109).

Analysis of gene chip data was performed by the DNA Sequencing Core at the University of Michigan Medical School. Linear models designed specifically for microarray analysis (21) were fit to the data using the limma package of Bioconductor (22). The models were weighted based on a gene-by-gene update algorithm designed to downweight chips that were determined to be less reproducible (23). The p values were adjusted for multiple comparisons using false discovery rate (24). For all other experiments, significance was calculated using repeated-measures ANOVA, followed by a post hoc Bonferroni test for significance between experimental groups. For single-group analysis, a two-tailed Student t test was used to determine significance. In all cases, p < 0.05 was considered statistically significant. Data analysis was performed using GraphPad Prism v6.0 for Macintosh (GraphPad, San Diego, CA).

Previous studies suggest that MLL1 expression and function can be regulated by NF-κB–dependent signals (14, 25). Because NF-κB signal transduction plays a central role in classical activation, we sought to determine whether MLL1 expression and/or global H3K4me1/2/3 modifications were increased in macrophages in response to classical activation stimuli. BMDMs from C57BL/6 mice were treated with classical activation stimuli, including cytokine (IFN-γ) and TLR ligands (LPS), and expression of Mll1 mRNA was analyzed via qPCR. Increases in Mll1 mRNA were observed in BMDMs in response to classical activation, with the most significant increase observed in response to LPS alone, supporting previous studies correlating NF-κB signal transduction with Mll1 expression (Fig. 1A). Interestingly, there was no significant difference in Mll1 between BMDMs treated with IFN-γ or IFN-γ+LPS, indicating that, at this specific time poststimulation (6 h), there does not appear to be an additive effect of the combined treatment on Mll1 expression.

We next sought to determine whether this increase in MLL1 expression correlated with increases in MLL1-dependent chromatin modifications (i.e., H3K4me3). To accomplish this, we generated a macrophage-specific conditional knockout of MLL1 using a previously generated MLL1^fx/fx mouse line mated to a commercially available Cre-expressing mouse line using the Lys2 promoter. Analysis of Mll1 mRNA expression in BMDMs from wild-type littermate (Cre⁻) and MLL1-deficient (Cre⁺) mice indicated a significant reduction in baseline Mll1 expression in targeted knockout mice (Fig. 1B). We next sought to determine whether this disruption in MLL1 expression would affect global histone methylation patterns in macrophages following classical activation. BMDMs were treated with classical activation stimuli, and levels of global H3K4me1/2/3 were analyzed via Western blot. High levels of global H3K4me1 were observed in both Cre⁻ and Cre⁺ BMDMs; however, Cre⁺ BMDMs exhibited decreases in H3K4me1 compared with Cre⁻ cells for IFN-γ and LPS stimulation alone (Fig. 1C). Interestingly, Cre⁺ BMDMs exhibited similar levels of global H3K4me1 compared with Cre⁻ BMDMs, suggesting that histone methyltransferases other than MLL1 may be more critical for setting global H3K4me1 marks in IFN-γ+LPS–treated BMDMs (Fig. 1C). In contrast, global levels of H3K4me2 were low and were only readily observable via Western blot following combined IFN-γ+LPS treatment (Fig. 1C). However, unlike H3K4me1, Cre⁺ BMDMs exhibited decreased H3K4me2 compared with Cre⁻ BMDMs in response to IFN-γ+LPS treatment (Fig. 1C). Global H3K4me3 levels were increased in activated Cre^−/− BMDMs in response to classical activation, especially in response to LPS stimulation alone or LPS in concert with IFN-γ (Fig. 1C). In contrast, MLL1-deficient BMDMs (Cre^+/−) were unable to increase global H3K4me3 levels in response to IFN-γ or LPS stimulation alone (Fig. 1C). As was observed with H3K4me1, Cre⁺ BMDMS exhibited similar levels of global H3K4me3 in response to IFN-γ+LPS treatment, further implicating compensatory histone methyltransferases in setting global H3K4me3 marks in IFN-γ+LPS–treated BMDMs. (Fig. 1C).

Because MLL1 is an activating histone methyltransferase that is thought to regulate gene expression via trimethylation of histones in gene promoter regions, we next sought to determine whether MLL1 deficiency resulted in decreased expression of proinflammatory genes and gene promoter chromatin modifications in BMDMs. Lys2Cre^+/− MLL1^fxfx BMDMs were cultured in vitro along with control BMDMs (Cre^−/−) and were treated with classical activation stimuli (i.e., IFN-γ, LPS, or the combination). Proinflammatory gene expression was then assayed via qPCR. In response to IFN-γ or LPS alone, MLL1-deficient BMDMs exhibited no significant differences in the expression of Nos2 (Fig. 2A) or Tnfa (Fig. 2B) compared with wild-type controls. However, when the stimuli were added together (IFN-γ+LPS), MLL1-deficient macrophages exhibited significant deficiencies in their ability to produce either proinflammatory mediator (Fig. 2A, 2B). These results suggested that MLL1 deficiency has an inhibitory effect on classical activation of macrophages in response to the combination of TLR ligand and cytokine stimulation.

To determine whether this decrease in gene expression was due to a decrease in MLL1-mediated chromatin modifications, formalin-fixed DNA from classically activated MLL1-deficient BMDMs and wild-type controls was assayed by ChIP for the presence of H3K4me3 in the promoter regions of Nos2 and Tnfa. For analysis of ChIP DNA, primer sets flanking putative NF-κB binding sites were used. Reductions in promoter H3K4me3 were observed in Nos2 and Tnfa promoters in unstimulated Cre⁺ BMDMs compared with Cre⁻ BMDMs (Fig. 2C–F); however, these differences were not statistically significant (p > 0.05). Following stimulation with IFN-γ, promoter regions proximal to exon 1 did not exhibit differential H3K4me3 (Fig. 2C, 2E), whereas more distal promoter sites retained an apparent decrease in H3K4me3 in Cre⁺ BMDMs (Fig. 2D, 2F), although, again, these differences were not statistically significant (p > 0.05). Interestingly, differential H3K4me3 was lost in the Nos2 proximal promoter region following stimulation with LPS alone (Fig. 2C), although this apparent decrease was maintained in distal Nos2 sites, as well as in both Tnfa promoter sites (Fig. 2D–2F). Importantly, following restimulation with IFN-γ+LPS, there was a statistically significant decrease in the level of H3K4me3 observed in Cre⁺ BMDMs compared with Cre⁻ BMDMs, specifically in the proximal Tnfa promoter (Fig. 2E), correlating with the significant reduction in Tnfa mRNA expression by IFN-γ+LPS–stimulated Cre⁺ BMDMs (Fig. 2B). Reductions in promoter H3K4me3 were also observed in Nos2, in particular the distal promoter site (Fig. 2D), as well as the Tnfa distal promoter (Fig. 2F); however, these differences were not statistically significant (p > 0.05).

Because the observed decrease in proinflammatory gene expression in MLL1-deficient BMDMs (6 h poststimulation) did not appear to match with the timing of modulated H3K4me3 chromatin modifications in these cells (24 h poststimulation), we next sought to characterize the kinetics of Mll1 mRNA expression in M1 BMDMs. As observed previously, classical activation resulted in increased Mll1 expression compared with unstimulated cells, with increases observable as early as 2 h poststimulation (Fig. 2G). The peak Mll1 expression was observed at 24 h in M1 wild-type BMDMs, correlating with the timing of H3K4me3 observed in our ChIP assay (Fig. 2G). Interestingly, when ChIP was performed on M1 BMDMs at 2 h poststimulation, there were no readily observable changes in chromatin modifications in the Nos2 or Tnfa promoters, suggesting that any early chromatin-remodeling events in these cells may occur below the limit of detection for the ChIP assay (data not shown). However, in studies of macrophage contributions to inflammatory processes in wound healing, significant changes in H3K4me3 were observed in MLL-deficient macrophages in response to LPS stimulation at 6 h, indicating that MLL-dependent modulations in histone methylation are observable at time points proximal to observed changes in mRNA expression (26).

The decreased expression of proinflammatory mediators by MLL1-deficient macrophages suggested that MLL1 may negatively regulate the ability of these cells to perform antimicrobial functions. To investigate this possibility, BMDMs from Lys2Cre^+/− MLL1^fx/fx mice were cultured with fluorescently labeled E. coli bioparticles, and phagocytosis of bioparticles was measured by a fluorescence plate reader. In comparison with wild-type control BMDMs (Lys2Cre^−/−), MLL1-deficient macrophages exhibited increased phagocytic ability (Fig. 3A). This increase in phagocytosis was observed in unstimulated BMDMs and in BMDMs that were pretreated with IFN-γ; however, no significant differences in phagocytosis were observed when BMDMs were pretreated with IL-4 (Fig. 3A). To determine whether this increased phagocytosis also correlated with increased killing, we challenged littermate control and MLL1-deficient BMDMs with group A Streptococcus and assayed viable recoverable bacterial CFU in these cocultures at the indicated time points (Fig. 3B). MLL1-knockout BMDMs (Lys2Cre^+/− MLL1^fx/fx) exhibited increased kinetics of bacterial killing, with significant differences in coculture CFU observed at 15 min of incubation (Fig. 3B). This increased killing ability of Cre⁺ BMDMs correlated with enhanced phagosome acidification by these cells, as measured by the increased fluorescence of a pH-sensitive reporter dye (pHrodo Red Dextran) endocytosed by BMDMs in culture (Fig. 3C). Enhanced phagosome acidification (as indicated by increased fluorescence of the reporter dye) was readily apparent at early time points posttreatment with reporter dye (5–60 min), correlating with the rapid enhancement of bacterial killing observed in these cells in earlier assays (Fig. 3B). These results indicate that MLL1-deficient BMDMs exhibit rapid increases in bacterial phagocytosis and killing functions.

Experimental data to this point suggested an important role for MLL1 in regulating macrophage antimicrobial functions, especially with regard to phagocytosis of microbes and production of proinflammatory mediators. However, single-analyte gene-expression profiling provided results that appeared contradictory to this hypothesis. For example, the observation of the decreased expression of the antimicrobial enzyme inducible NO synthase (Nos2) (Fig. 2A) appears in contrast to the increased ability of MLL1-deficient BMDMs to effectively kill bacteria in coculture (Fig. 3B). Additionally, although the decreased production of Nos2 mRNA correlated with decreased promoter levels of H3K4me3, suggesting a functional link between MLL1 and inducible NO synthase expression, conflicting results were observed for NO production by MLL1-deficient BMDMs (data not shown).

Therefore, to observe the effect of MLL1 deficiency on global gene expression in macrophages in a more unbiased manner, mRNA from wild-type (Cre⁻) and MLL1-deficient (Cre⁺) BMDMs were analyzed via gene chip (Mouse Gene ST 2.1; Thermo Fisher Scientific) for relative expression changes. Isolated mRNA from Cre⁻ and Cre⁺ BMDMs stimulated with classical activation signals (IFN-γ, LPS, or IFN-γ+LPS) was analyzed using a gene chip panel of >40,000 RefSeq transcripts, and resulting fold expression was determined by analyzing quality-controlled expression values for validated probe sets. Comparisons were then made between Cre⁻ and Cre⁺ BMDMs across each stimulation condition for the determination of differentially expressed genes between MLL1-competent and MLL1-deficient BMDMs. Genes that were identified as being differentially regulated (fold change ≥ 1.5, with a false discovery rate adjusted p value ≤ 0.1) were identified (Fig. 4A). Numerous differentially regulated genes were identified between Cre⁻ and Cre⁺ BMDMs in all culture conditions, with the largest number of differentially regulated genes manifesting in response to stimulation with IFN-γ (Fig. 4B). Interestingly, loss of MLL1 expression resulted in decreased gene expression, as would have been predicted based on MLL1’s function as an epigenetic activation of gene expression, as well as increased gene expression in every culture condition tested (Fig. 4B). The full dataset is available through the National Center for Biotechnology Information Gene Expression Omnibus under GEO accession number GSE82109.

Interestingly, many gene families involved in type I IFN responses were included in the genes that were differentially regulated at baseline (“unstimulated”) between Cre⁻ and Cre⁺ BMDMs (Table I). This included a significant downregulation of Stat4, which encodes for the STAT4 transcription factor that can participate in signal transduction through the type I IFN receptor (IFNAR). Concurrently, many gene products associated with type I IFN responses were upregulated, including mRNAs encoding Mx proteins (Mx1 and Mx2) and 2′-5′-oligoadenylate synthetase proteins (Oas1g, Oasl1, and Oas2) (Table I). These results indicated that BMDMs deficient in MLL1 expression had dysregulated type I IFN responses, possibly driven by decreased STAT4 expression and/or signal transduction.

To confirm the gene expression results obtained through the global mRNA expression–profiling system, Cre⁻ and Cre⁺ BMDMs were cultured in vitro, and gene expression was analyzed via single-analyte qPCR. Initially, the expression of a panel of proinflammatory genes identified as being differentially regulated in resting BMDMs in the gene chip analysis was confirmed (Fig. 5A). As described previously, expression of Mx1, Ms2, and Oasl1 was upregulated in Cre⁺ BMDMs, whereas expression of Stat4 was decreased (Fig. 5A). Decreases were also observed in expression levels of Il1b, Lbp, and Vcam1, although no difference was observed in Ahr expression, as was identified previously in the gene chip analysis (Fig. 5A). Additionally, decreased expression of Mll1 mRNA was observed in resting Cre⁺ BMDMs compared with baseline, as was expected with the use of the conditional-knockout mouse model (Fig. 5A).

We then treated Cre⁻ and Cre⁺ BMDMs with classical activation stimuli and assayed subsequent gene expression by qPCR. Stat4 mRNA expression was decreased in Cre⁺ BMDMs that were treated with IFN-γ+LPS (Fig. 5B). This decrease in Mll1 and Stat4 expression correlated with increases in Mx1 and Mx2 expression, in particular with IFN-γ+LPS treatment (Fig. 5C, 5D). Expression of Mx2 was also significantly increased in Cre⁺ BMDMs treated with LPS (Fig. 5D), although no differences were observed in Stat4 or Mx1 expression between Cre⁻ and Cre⁺ BMDMs under these conditions (Fig. 5B, 5C).

The results of the gene chip studies and single-analyte qPCR indicated that MLL1 may be involved with the regulation of STAT4 signal transduction. To test this, we first sought to determine which STAT4-dependent cytokine signals could support MLL1 expression in macrophages. BMDMs from C57BL/6 mice were treated with a panel of STAT4-activating recombinant cytokines, and Mll1 mRNA expression was assayed by qPCR. At equivalent concentrations, the type I cytokines IL-12, IL-23, and IL-27 were unable to modulate Mll1 expression in BMDMs (Fig. 6A). Additionally, the type III IFN IL-28β (IFN-λ3) was unable to increase expression of Mll1 in BMDMs (Fig. 6A). In contrast, the type I IFNs IFN-α and IFN-β were able to drive Mll1 expression, with the most significant increase observed with IFN-α treatment (Fig. 6A).

Although STAT4 has been shown to participate in type I IFN signal transduction in a noncanonical fashion (27), the primary pathways for these IFNs are via STAT1 and STAT2 (28). Therefore, we first attempted to observe whether MLL1 deficiency had any effect on the ability of macrophages to initiate IFN-α signal transduction through canonical and noncanonical STAT pathways. In response to IFN-α stimulation, Cre⁻ and Cre⁺ BMDMs were able to rapidly phosphorylate STAT1, with no overt differences between Cre⁻ and Cre⁺ BMDMs with regard to kinetics or strength of phosphorylation (Fig. 6B). Low levels of STAT2 phosphorylation were also observed in IFN-α–treated cells; however, these levels were reduced significantly compared with p-STAT1 levels (Fig. 6B). Total protein levels of STAT1 and STAT2 appeared similar between Cre⁻ and Cre⁺ BMDMs, indicating that MLL1 deficiency did not affect the expression levels of these two STAT proteins (Fig. 6B). Interestingly, STAT4 phosphorylation was largely absent in Cre⁻ and Cre⁺ BMDMs, and protein expression levels of total STAT4 were low in both cell types (Fig. 6B). These results indicate that MLL1 deficiency does not affect canonical IFN-α signal transduction and suggest that STAT4 expression in MLL1-deficient BMDMs is not directly driven by type I IFN signaling.

Because our previous mRNA analysis indicated that MLL1-deficient BMDMs exhibit decreases in Stat4 expression in response to type II IFN and TLR ligand stimulation, we sought to determine whether STAT4 protein expression would be modulated in MLL1-deficient BMDMs in response to inflammatory stimuli other than type I IFNs. Therefore, to confirm the reduction in Stat4 expression observed in Cre⁺ BMDMs at the protein level, cultured BMDMs from Cre⁻ and Cre⁺ mice were stimulated with a panel of TLR ligands, and total STAT4 protein was assayed via Western blot. In response to TLR stimulation, Cre⁻ BMDMs exhibited increased STAT4 expression, with the strongest upregulation observed in response to the synthetic dsRNA pI:C (Fig. 6C). In comparison, Cre⁺ BMDMs were able to induce expression of STAT4 protein in response to TLR stimulation, but the total amount of STAT4 protein was decreased in all conditions compared with Cre⁻ BMDMs, with the most striking difference observed in response to pI:C stimulation (Fig. 6C).

We next sought to determine the kinetics of STAT4 expression in MLL1-deficient BMDMs in response to classical activation stimuli. Cultured BMDMs from Cre⁻ and Cre⁺ mice were stimulated with IFN-γ, LPS, or both, and mRNA was isolated from these cells at 2, 6, and 24 h poststimulation for the analysis of Stat4 mRNA expression via qPCR. As observed previously in the microarray analysis (Table I) and single-analyte qPCR of cultured BMDMs (Fig. 5B), significant reductions in Stat4 expression were observed primarily in response to IFN-γ+LPS treatment (Fig. 6D). This significant reduction in Stat4 mRNA was readily observable at 6 h poststimulation, indicating that modulations in STAT4 expression in MLL1-deficient BMDMs arise secondary to stimulation with exogenous cytokines (i.e., IFN-γ) and TLR ligands (Fig. 6D).

We next sought to determine whether MLL1 exerts direct epigenetic control over STAT4 expression by measuring H3K4me3 levels in the promoter of the STAT4 gene locus. Cre⁻ and Cre⁺ BMDMs were fixed in formalin and processed for H3K4me3 ChIP, and precipitated DNA was amplified using primers specific for the STAT4 promoter. At baseline, Cre⁺ BMDMs had decreased H3K4me3 in the STAT4 promoter compared with Cre⁻ BMDMs (Fig. 6E), suggesting that the loss of MLL1 expression had an effect on permissive histone modifications in the STAT4 promoter in BMDMs. Interestingly, there was no observable difference between H3K4me3 levels in the STAT4 promoter in Cre⁻ and Cre⁺ BMDMs stimulated with IFN-γ alone; however, reductions were observed in LPS-treated cultures, and these differences reached significance with IFN-γ+LPS stimulation (Fig. 6E). These results indicate that MLL1 deficiency has a significant effect on permissive chromatin remodeling in the STAT4 promoter, especially in response to classical activation stimuli. Importantly, ChIP assays were unable to detect direction interactions between MLL1 and the STAT4 promoter above background (nonspecific Ig) levels (data not shown), possibly as a result of weak interactions between MLL1 and DNA being lost during formalin fixation and sonication or binding of MLL1 to sequences outside of the regions defined by our primer sets, among other possibilities.

The expression and function of CMEs play a central role in governing macrophage activation and differentiation to various effector lineages. Studies of macrophage molecular biology have defined important relationships between specific extracellular signals (e.g., cytokines, pathogen-associated molecular patterns) and unique histone modifications catalyzed by specific CMEs. For example, expression of the lysine demethylase KDM6B/JMJD3 is increased in macrophages in response to IL-4, leading to removal of repressive histone modifications in the promoter regions of M2 marker genes, supporting alternative activation (10). In contrast, type I IFN stimulation leads to upregulation of the lysine methyltransferase KMT1F/SETDB2, leading to epigenetic suppression of antiviral genes and proinflammatory cytokines in activated macrophages (29). As with many other immune cell lineages, it is becoming apparent that macrophages exhibit distinct epigenetic signatures based on the signals available in the local microenvironment.

In this study, we have described a unique role for the lysine methyltransferase KMT2A/MLL1 in regulating macrophage classical activation through its ability to be driven by cytokine and TLR signals. The resulting histone modifications mediated by the function of MLL1 (i.e., H3K4me3) have a dual role in regulating macrophage classical activation. Activating chromatin remodeling events catalyzed by MLL1 help to drive expression of M1 marker genes, such as Nos2 and Tnfa; in this fashion, MLL1 appears to be important for epigenetic regulation of proinflammatory genes during classical activation. Importantly, the strongest suppression of proinflammatory gene expression and histone methylation occurred in response to treatment with IFN-γ+LPS, rather than each individual stimulus alone, arguing that both cytokine and TLR ligand stimulus are necessary for MLL1-dependent proinflammatory gene expression. In contrast, ablation of MLL1 results in the promotion of phagocytosis and killing of bacteria, as well as the upregulation of genes involved with antiviral responses (e.g., Mx1 and Mx2). These results indicate that epigenetic regulation of macrophage function by MLL1 may bias macrophage classical activation functions toward the promotion of a local proinflammatory state at the expense of cell-intrinsic functions, such as pathogen clearance and antiviral responses.

These results suggest a role for MLL1 in exerting regulatory control on the proinflammatory versus phagocytic/cytotoxic functions of macrophages. To place these results in context with previously published reports regarding the function of MLL1 in immune cells, it appears that MLL1 is preferentially involved in epigenetic regulation of cytokine production in leukocytes. For example, deletion of MLL1 in T cells results in impaired production of lineage-specific cytokines (e.g., IFN-γ and IL-4) in cells skewed to various effector lineages (12, 13). Similar phenotypes are observed in this study: deletion of MLL1 results in reductions in IL-1β and TNF-α production, concurrent with the loss of permissive H3K4me3 in gene promoters. Interestingly, gene chip analysis of mRNA expression in MLL1-deficient versus MLL1-competent macrophages did not identify any differentially regulated genes that are directly involved in phagocytosis or microbial killing (e.g., oxidases), suggesting that the loss of MLL1 may have a more indirect role in the promotion of phagocytosis/killing phenotypes. Further studies are required to fully elucidate the molecular mechanisms governing MLL1’s apparent negative regulation of microbial phagocytosis and killing in wild-type macrophages.

Previous reports have identified an important role for NF-κB signal transduction in response to TNF-α for the regulation of MLL1 expression and function in macrophages (14, 25). Our studies identified potent upregulation of Mll1 mRNA in response to LPS stimulation, which primarily signals through TLR4 and NF-κB. Interestingly, our results also identify a novel pathway for MLL1 expression via STAT4 signal transduction. STAT4 is classically associated with IL-12 signaling and the promotion of T_H1 responses (30); however, STAT4 can also participate as a secondary transduction pathway for type I IFN signaling (27). Our results identify a novel STAT4-dependent positive-feedback loop for MLL1 expression in macrophages, whereby MLL1 epigenetically regulates STAT4 expression, and STAT4-dependent signals drive expression of MLL1. Interestingly, it appears that baseline expression of MLL1 is required for epigenetic regulation of STAT4 expression in macrophages, because the STAT4 promoter is enriched with H3K4me3, and macrophages deficient in MLL1 exhibit decreased STAT4 expression and decreased H3K4me3 in the STAT4 promoter. These results correlate with findings for MLL1 function in CD4⁺ T cells, whereby IL-12 stimulation leads to MLL1 expression and epigenetic regulation of T_H1 phenotypes (12). Therefore, it is expected that a similar positive-feedback loop for STAT4 expression would exist in other leukocyte subsets, especially those that rely on STAT4 signals for guiding lineage-commitment decisions.

Of the putative STAT4-activating cytokines used in this study, only type I IFNs were able to drive expression of Mll1 mRNA, with IFN-α being the most potent stimulator of gene expression. Therefore, it is surprising to observe that, in MLL1-deficient macrophages that exhibit decreased STAT4 expression, there was a concurrent increase in IFN-inducible antiviral genes, such as Mx1 and Mx2. This suggests a disconnect between STAT4 activation in macrophages and the development of IFN-inducible cell-intrinsic antiviral responses. One possibility for this observed phenomenon would be compensatory regulation of other type I IFN signaling components in response to the decrease in STAT4 expression in MLL1-deficient macrophages. However, there was no observed change in Stat1 or type 1 IFN receptor components (Ifnar1 or Ifnar2) in the gene chip analysis performed on in vitro–stimulated macrophages. In addition, there was no observable decrease in STAT1 protein expression or cell surface expression of the type 1 IFN receptor in MLL1-deficient macrophages (data not shown), providing additional evidence that type 1 IFN–induced expression of MLL1 is distinctly regulated by STAT4. The surprising observation of decreased inducible STAT4 expression with concurrent enhanced IFN-inducible gene expression in MLL1-deficient macrophages remains an important unresolved aspect of these studies that requires more in-depth investigation.

Despite the limited understanding of the molecular mechanisms governing the link among MLL1, STAT4, and genes, such as Mx1/Mx2, these results suggest that targeting of MLL1 in IFN-inducing viral responses may be a promising therapy for the promotion of macrophage antiviral responses. For example, lung macrophages are known to play a central role in the defense against influenza through the production of type 1 IFN during infection (31). Inhibition of MLL1 function during influenza infection may help to promote the expression of antiviral genes while limiting the suppressive effects of type 1 IFN signaling. These results also suggest that differential expression of histone methyltransferases has drastic effects on macrophage antiviral responses. For example, type I IFN induction of the histone methyltransferase Setdb2 appears to have an opposite effect from MLL1, whereby macrophage antiviral responses are blunted as a result of enhanced Setdb2 and histone 3 lysine 9 (H3K9) methylation (29). The disparate effects of MLL1 versus Setdb2-mediated epigenetic regulation of macrophage antiviral responses is surprising, given that both enzymes are upregulated in response to type I IFN. Therefore, further investigations into the relative contribution of MLL1 versus Setdb2 in macrophage-mediated antiviral responses is of critical importance.

The observed changes in macrophage function observed in our MLL1-deficient model system are suggestive of an important role for this CME for antimicrobial functions; however, many questions remain regarding the specific molecular mechanisms governing the interplay between MLL1 and functional outcomes. For example, the results of our phagocytosis and killing assays strongly suggest that MLL1-deficient macrophages exhibit enhanced intracellular bacterial killing functions, but other possibilities may explain the results of our studies. For example, extracellular killing of group A Streptococcus in our coculture assays could have occurred as a result of secretion of reactive oxygen or reactive nitrogen species (ROx/NOx), or secretion of antimicrobial enzymes (such as lysosome). With regard to ROx/NOx, extracellular release of NO appeared to be decreased in MLL1-deficient BMDMs, correlating with decreased Nos2 mRNA expression, but this was only in response to relatively high doses of LPS and only at early time points (24 h) following initial stimulation (26). These differences were not observed following 48 h of LPS stimulation (26). Additionally, although extracellular production of NO was decreased in response to high doses of LPS, intracellular NO levels were unchanged at early time points and, surprisingly, were increased at later time points of in vitro stimulation (26), suggesting that intracellular killing due to NOx might be enhanced in MLL1-deficient macrophages under certain conditions. Importantly, lysozyme was one of the genes shown to be significantly increased in MLL1-deficient macrophages (Fig. 5A); secretion of high levels of lysozyme by Cre⁺ BMDMs may have driven the rapid killing of group A Streptococcus in our coculture system separate from enhanced phagocytosis or phagosome acidification. It is also critical for future studies to clarify the specific functions of MLL1 as opposed to other MLL family methyltransferases in governing histone modifications and gene expression patterns, especially with regard to STAT4 regulation. Multiple histone methyltransferases can mediate H3K4me3, and the interplay between MLL1 and STAT4 expression may be due to an indirect effect by MLL1 regulation of these other CMEs. The results of this study can only draw a correlation between MLL1 and STAT4; although this relationship appears to be reciprocal in nature, a direct link between MLL1-dependent chromatin remodeling and inducible STAT4 expression cannot be confirmed without direct evidence of MLL1 binding to the STAT4 promoter. Future studies will endeavor to identify the landscape of MLL1 binding to key proinflammatory gene promoters in macrophages following classical activation.

The epigenetic regulation of macrophage functions by MLL1 appears to govern the balance between proinflammatory function and antimicrobial clearance responses during classical activation. These results provide a novel role for MLL1-dependent H3K4me3 chromatin remodeling in macrophage proinflammatory responses, as well as suggest differential epigenetic programs for distinct effector mechanisms of classical activation (i.e., cytokine production versus microbial killing). Therefore, MLL1 may prove to be an effective therapeutic target for infectious diseases that exhibit severe immunopathology as a major driver of morbidity and mortality. With the first generation of MLL1 inhibitors now becoming available (32), it may be beneficial to consider their future use for restricting pathogen-induced immunopathology, such as in severe sepsis responses, viral infections (e.g., influenza), or other pathogen-mediated inflammatory disorders.

We thank J. Michelle Kahlenberg for advice on scientific approach and Ronald Allen, Sarah Salter-Green, Mackenzie Zierau, and Yesen Zhou for technical assistance.

The authors have no financial conflicts of interest.