Monocytes were isolated from the EDTA anticoagulated blood of healthy volunteers. Written informed consent was given under National Research Ethics Service approval from Frenchay Research Ethics Committee reference 09/H0107/22 and South West 4 Research Ethics Committee reference 10/HO102/72, respectively. Mononuclear cells were isolated using Ficoll-Paque Plus (GE Healthcare Life Sciences), red blood cells were lysed with 150 mM ammonium chloride/0.1% BSA, and monocytes were allowed to adhere to plastic in RPMI 1640/1% human serum AB (SigmaAldrich) for 1 h. Non-adherent cells were removed with warm Dulbecco’s phosphate buffered saline (D-PBS; Gibco), and the medium was replaced with RPMI 1640/10% fetal bovine serum (FBS; SigmaAldrich) for 1 h. Based on staining with Rabbit anti-(human CD14) antibody, Ab78313, adhered cells were at least 85% monocytes. Monocytes were differentiated into macrophages in RPMI 1640 medium containing 10% FBS and 20 ng/mL of colony-stimulating factor-1 (human recombinant CSF-1, R & D systems), which was replenished on day 3. Approximately 80% of the resulting cells were macrophages based on positive staining with mouse monoclonal anti-CD68 (M0876, Dako). Differentiated macrophages were treated for 6, 18, 32 and 48 h in the same medium with either 100 ng/mL of recombinant human interferon-γ (IFN-γ) (R & D systems) or 10 ng/mL of recombinant human interleukin-4 (IL-4) (R & D systems). AZD1152 was purchased from Selleck.

Total RNA was isolated from the macrophages prepared from three different healthy donors using the PureLink™ RNA Mini Kit (Ambion). Total RNA was quantified using an ND1000 NaNo Drop spectrophotometer, and 100 ng were reversed transcribed using the QuantiTect Reverse Transcription Kit (Qiagen) with additional genomic DNA elimination step indicated in the manufacturer’s instructions. For quantitative polymerase chain reaction (qPCR), the cDNA samples were diluted 1:3 in 10 mM Tris–HCl, pH 8.0 and amplified using the LightCycler 480 SYBR Green I Master mix (Roche) in an Eco Real-Time PCR System (Illumina), using primer sets shown in Table 1. Data were normalized to total RNA in each reaction. For qPCR array, RNA from three different donors was analyzed using the Human Epigenetic Chromatin Modification Enzymes RT² Profiler PCR Array (Qiagen) according to the manufacturer’s protocol. Briefly, total RNA (400 ng, genomic DNA eliminated) was reverse transcribed using RT² First Strand Kit (Qiagen) and diluted in RNase-free water. The amplification reaction was conducted in 384 well format in a Roche LightCycler 480 (95°C 10 min for one cycle followed by 95°C 15 s and 60°C 1 min for 45 cycles). Threshold cycle (C_T) values were exported and analyzed using web based SABiosciences PCR Array Data Analysis Software.¹ A panel of five housekeeping genes integral to the array were used to calculate for each probe ΔC_T = (C_T probe − C_T average of housekeeping genes). The values of ΔΔC_T = ΔC_T experimental sample − ΔC_T control were calculated for each probe and converted to fold changes = (2^(−ΔΔCT)). For transcriptomic analysis, purified RNA samples from four different donors were submitted to the Illumina Gene Expression ServiceXS (Leiden, Netherlands) and were processed for analysis on the Illumina HumanHT-12 v4 microarray as described in detail.² The results were deposited under number GSE83957. Fold changes and statistics (multiple testing corrections) of generated raw data were performed using GeneSpring (Agilent Technologies). Venn Diagrams were generated by using web-based software.³ Gene ontology enrichment (GOE) and KEGG pathway analysis were performed using the DAVID public database.⁴

Macrophages were lysed in SDS lysis buffer [2% SDS (w/v)/16% glycerol (v/v) in 50 mM Tris, pH 6.8]. Protein was measured (Micro BCA kit, Thermo Scientific Pierce). Equal amounts of reduced protein were fractionated by SDS-polyacrylamide gel electrophoresis, transferred to PVDF membranes (Merck Millipore), blocked in TBST/5% skimmed milk followed by incubation in primary antibody. Proteins were detected using appropriate HRP-conjugated secondary antibodies (SigmaAldrich) and enhanced chemiluminescence (Immobilon, Merck Millipore) and Hyperfilm™ ECL (GE Healthcare Life Sciences). The antibodies used were Phospho-Rb-Ser807/811, E2F1, KDM6A, and SMYD3 (Cell Signaling), Histone H3 and Histone H3-Phospho S10 (Abcam), GAPDH (Millipore), and p27^kip1 (BD Biosciences).

To measure S-phase entry to the cell cycle, macrophages were labeled with 10 µM BrdU (SigmaAldrich) for 24 h. Cell proliferation was quantitated by immunohistochemistry as previously described (18). The percentage of BrdU positive nuclei was counted using NIH ImageJ software.

Recombinant adenovirus encoding for human p27^kip1 was a kind gift from Professor Betsy Nabel (NIH, MD, USA). This virus and a control adenovirus expressing destabilized, enhanced green fluorescent protein (dsEGFP) were used as described previously (19). To generate the adenovirus expressing E2F1, the plasmid E2F1 wt-pGex2TK containing the coding sequence for human E2F1 from William Kaelin (20), was purchased from Addgene (Addgene plasmid # 21668). It was amplified using KOD DNA polymerase (Merck-Millipore, UK) to include EcoRI and BamH1 flanking sites and subcloned into the shuttle vector pDC515io from Microbix (ON, Canada). Recombination (Flp/FRT mediated) was performed in 293IQ cells to inhibit transgene protein expression (a gift from Dr. D. Matthews, University of Bristol) (21). For gene silencing, short hair pin (Sh) sequences were predicted using http://cancan.cshl.edu/RNAi_central/RNAi.cgi?type=shRNA. The sequences used to silence KDM6A, 5′CTGCCATTAAATGCTACTTAAATAGTGAAGCCACAGATGTATTTAAGTAGCATTTAATGGCAT3′, KDM6B, 5′CGCCCAGTCTGTGAAACCGAAGTAGTGAAGCCACAGATGTACTTCGGTTTCACAGACTGGGCA3′ and firefly luciferase, 5′CGCCTGAAGTCTCTGATTAATAGTGAAGCCACAGATGTATTAATCAGAGACTTCAGGCGGT3′, as a control, were embedded in the backbone of the primary microRNA-30, as described previously (22) and were synthesized by Eurofins. DNA sequences for modified microRNA-30 were synthesized by Eurofins, cloned into the Nhe1-BamH1 sites of the shuttle vector pDC515 and adenoviruses were generated as described above. Virus stocks were purified by CsCl banding and titrated by plaque assay. Monocytes differentiated for 4 days were infected with Ad-p27^kip1 or Rad66 at 10⁸ plaque forming units (pfu)/ml for 24 h. Alternatively, cells were infected with Ad-E2F1 or Ad-dsEGFP at 5 × 10⁷ pfu/mL for 20 h followed by further 18 h treatment in the presence of IFN-γ or IL-4. For gene silencing, cells were infected with shKDM6A, shKDM6B, or shLuciferase (shLUC) viruses at 2 × 10⁸ pfu/ml for 72 h before stimulation with IFN-γ or IL-4 for 6 h.

Normality of data sets was analyzed by the method of Kolmogorov and Smironov. Differences between means of normally distributed variables with similar variances were analyzed using a paired Student’s t-test or, for multiple comparisons, ANOVA followed by a Dunnett or Student–Newman–Keuls post-test, as appropriate. For the array, data from the Benjamini–Hochberg method were used for multiple testing correction. All data are from independent experiments on cells from different donors and are presented as mean ± SE. *indicates p < 0.05, **indicates p < 0.01, ***indicates p < 0.001.

To identify optimal time points for transcriptomic analysis, we measured changes in the mRNA expression of phenotypic markers SOCS3 (23) and CD206 (12). As expected, IFN-γ but not IL-4 induced SOCS3 mRNA expression (Figure 1A), which was significant by 6 h and remained maximal at 18 and 32 h: it then declined but was still significantly elevated above untreated controls at 48 h (Figure 1A). Also as expected, IL-4 stimulated but IFN-γ inhibited CD206 mRNA expression after 48 h (Figure 1B). Interestingly, however, IFN-γ and IL-4 both induced CD206 to a similar extent at 6 h but the effect of IL-4 increased further from 18-48 h, whereas that of IFN-γ waned such that CD206 expression had declined below control levels after 32 and 48 h (Figure 1B). We, therefore, chose the 18 h and 48 h samples from these experiments to capture the differential effects of IFN-γ and IL-4 on steady-state mRNA levels of 84 epigenetic regulators using a commercially available RT-qPCR array. The combination of IFN-γ and IL-4 was also investigated because antagonistic interactions might enhance phenotypic differences (see Introduction). The results were normalized against a panel of five housekeeping genes integral to the profiler and significant changes were identified by using the array manufacturer’s software, which yielded values of mean fold change and p values after false discovery rate correction (Table 2). Eight significant differences and three non-significant trends that were subsequently found significant by conventional RT-qPCR are highlighted in bold. There were more significant responses to IFN-γ (10 gene changes at 18 h, of which 7 persisted at 48 h) than IL-4 (7 gene changes, only 2 of which were significant at both 18 h and 48 h). Contrary to our expectation of antagonistic responses, IFN-γ and IL-4 appeared to produce additive effects on the expression of epigenetic regulators (Table 2). The significant changes from the RT-qPCR screen were validated and extended by a detailed time course study using the full set of samples (Figures 2A–C and 3A–H). All the significant changes observed in the array were confirmed; and also the trends toward increase of KDM6B by IL-4 and decrease of KAT2A and SETD6 by IFN-γ were shown to be significant by standard RT-qPCR normalized against total RNA, which meets current recommendations (24). Only three genes, namely CIITA, KDM6B, and NCOA1 showed increased expression (Figures 2A–C), whereas the other 8, namely AURKB, ESCO2, KAT2A, PRMT7, SETD6, SMYD3, SUV39H1, and WHSC1 showed decreased expression (Figures 3A–H). The IFN-γ-induced increases in CIITA, KDM6B, and NCOA1 showed a similar pattern to SOCS3, with significant induction by 6 h and a tendency to decline thereafter (Figures 1A and 2A–C). The IL-4-induced increase in KDM6B was also significant by 6 h of treatment but then declined, unlike CD206, which remained elevated (Figures 1B and 2B). The decreased expression of AURKB, ESCO2, KAT2A, PRMT7, SETD6, SMYD3, SUV39H1, and WHSC1 was delayed to 18 h and beyond (Figures 3A–H). In the few cases, where it was possible, we sought to confirm the mRNA data by protein or activity measurements. In the case of SMYD3, IFN-γ significantly reduced protein levels after 48 h (Figure 4A). Either IFN-γ or IL-4 inhibited AURKB activity measured by the phosphorylation of histone3 on serine-10 (H3pS-10) (Figure 4B) (25). Consistent with this, we demonstrated that AZD1152, a pharmacological inhibitor of AURKB (26) also inhibited H3pS-10 to the same extent as either IFN-γ or IL-4 (Figure 4C).

Interestingly, phosphorylation of H3pS-10 by AURKB is necessary for chromatin reorganization during mitosis (27–29), ESCO2 promotes sister chromatid cohesion (30), SUV39H1 also has a role in chromosome segregation (31, 32) and WHSC1 has been ascribed a role in DNA repair during replication (33). Hence all these enzymes have established roles in cell division. Consistent with other previous literature (34, 35), we found that IFN-γ or IL-4 inhibited phosphorylation of retinoblastoma protein (Rb, Figure 4B), confirming that they arrested cells at the G1/S checkpoint in the cell cycle (36). Treatment with either IFN-γ or IL-4 also profoundly decreased BrdU incorporation as a marker of DNA replication in our macrophages (Figure 4D). These observations led us to question whether decreased expression of histone-modifying genes by IFN-γ or IL-4 might be the incidental consequence of inhibiting proliferation.

To investigate the relationship between proliferation and mRNA levels of epigenetic regulators further, we cultured macrophages differentiated for 6 days for four more days either with or without growth factors, which arrests cells in the early G1 phase of the cell cycle. Consistent with this, prolonged culture decreased phosphorylated retinoblastoma protein (pRb) and H3pS-10 levels and these effects were even greater without growth factors (Figure 5A). Furthermore, BrdU incorporation after 10 days of differentiation declined by a further 76% (n = 4, p = 0.002) when growth factors were omitted. Similarly, expression of AURKB and ESCO2 mRNA each declined significantly between 6 and 10 days of culture and the decrease was greater after growth factor depletion (Figures 5B,C), suggesting that their expression was at least partly dependent on proliferation. Growth factor depletion did not affect IFN-γ or IL-4 upregulated genes, CIITA, KDM6B, or NCOA1, or the downregulated genes KAT2A, PRMT7, SETD6, or SMYD3 (Figure 5D). SUV39H1 and WHSC1 expression showed non-significant trends toward reduction (Figure 5D).

IFN-γ or IL-4 arrest cell proliferation at the G1/S checkpoint thanks to elevation of the cyclin-dependent kinase inhibitor p21^Cip1 (34, 35). We sought to inhibit proliferation at the G1/S checkpoint by an alternative mechanism. Given that gene silencing is inefficient in macrophages, we overexpressed the cyclin-dependent kinase inhibitor p27^kip1 (37) from an adenovirus. Infection with the p27^kip1 expressing virus increased p27^kip1 protein, as expected (Figure 6A), and dramatically decreased BrdU incorporation (Figure 6B), pRb (Figure 6C) and H3pS-10 (Figure 6D). Hyperphosphorylation of Rb releases the S-phase transcription factor, E2F, which induced multiple genes that include Cyclin E and proliferating cell nuclear antigen (PCNA) (38). As expected, therefore, overexpression of p27^kip1 profoundly decreased mRNA levels of cyclin E (84 ± 2%) and PCNA (86 ± 6%, both n = 3, p < 0.01), confirming G1/S blockade. Overexpression of p27^kip1 decreased mRNA levels of AURKB, ESCO2, SUV39H1, and WHSC1 (Figure 6E), to a similar extent as IFN-γ or IL-4. By contrast, levels of the other genes downregulated by IFN-γ, namely KAT2A, PRMT7, SETD6, and SMYD3 were not significantly reduced by p27^kip1 overexpression (Figure 6E) and were, therefore, clearly independent of inhibition of proliferation by either growth factor depletion or p27^kip1 overexpression. Steady-state mRNA levels of the genes upregulated by IFN-γ or IL-4, namely CIITA, KDM6B, and NCOA1, were not decreased or even increased by overexpression of p27^kip1 (Figure 6E).

It has been suggested that AURKB is a direct target of transcription factor E2F (29). Furthermore, when we interrogated the ENCODE database, we found chromatin immunoprecipitation evidence for binding of E2F transcription factors to the proximal promoters of the AURKB, ESCO2, SUV39H1, and WHSC1 genes, whereas CIITA, KDM6B, and NCOA1 had no such sites. We, therefore, investigated whether the effects of IFN-γ or IL-4 on AURKB, ESCO2, SUV39H1, and WHSC1 could be reversed by adenovirus-mediated overexpression of E2F1. Infection with the E2F1 expressing virus increased E2F1 protein (Figure 7A) and, as expected, the mRNA levels of the known E2F responsive genes, Cyclin E and PCNA (38) (Figures 7B,C). E2F1 gene transfer completely reversed the inhibitory effect of IFN-γ and IL-4 on AURKB, ESCO2, SUV39H1, and WHSC1 mRNA levels (Figures 7D–G), except in the case of AURKB for which the effect of IL-4 was only partly reversed (Figure 7D). These data provided further support for the conclusion that the effects of IFN-γ or IL-4 on these genes were, at least partly, mediated indirectly through inhibition of proliferation.

From the above results we concluded that increased expression of CIITA, KDM6B, and NCOA1 and decreased expression of KAT2A, PRMT7, SETD6, and SMYD3 was independent of any effects on proliferation. To begin to understand the impact of these changes on macrophage phenotype, we chose to further investigate KDM6B, which was the only transcriptional enhancer that was increased at the mRNA level by both IFN-γ and IL-4. Given that the action of KDM6B is demethylation of lysine 27 on histone3, which is associated with increased gene transcription (13), this upregulation might be expected to promote transcriptional responses to both cytokines. Conversely, inhibition and silencing of KDM6B might decrease gene expression associated with IFN-γ and IL-4. To narrow down the search for those genes regulated by KDM6B, either on its own or in combination with KDM6A, we first performed a microarray study of transcripts upregulated by IFN-γ or IL-4 in the presence and absence of the combined KDM6A and B inhibitor GSK-J4 (39). From preliminary time course and dose-response studies (results not shown), 6 h exposure was sufficient and 60 µM GSK-J4 was chosen because it significantly suppressed TNFα induction by lipopolysaccharide (LPS) by approximately 70% [confirming previous results (39)]. CD206 induction by IL-4 was also inhibited, albeit by only 30% (both p < 0.05), whereas a housekeeping gene, 36B4, was not affected. From these pre-validated samples, the transcriptomic analysis showed that IFN-γ significantly upregulated 906 and IL-4 upregulated 271 transcripts after 6 h (a complete gene list is deposited under GSE83957). Only 62 of these (6%) were upregulated by both IFN-γ and IL-4, which confirms the differential phenotypes stimulated by these two cytokines, as demonstrated more extensively previously (1). This data also emphasizes how unusual KDM6B is in being upregulated by both IFN-γ and IL-4. Of the 831 IFN-γ upregulated transcripts recognized by the genevenn program used to generate Venn diagrams, 181 (22%) were significantly decreased by the additional presence of GSK-J4 (Figure 8A). Using GOE and KEGG pathway analysis, these genes were associated with several aspects of immune cell function and transcriptional activation (Tables 3 and 4). Of the 254 IL-4 upregulated transcripts only 28 (11%) were significantly reduced by the additional presence of GSK-J4 (Figure 8B). There were insufficient genes in this cluster to perform GOE or KEGG pathway analysis. Only two transcripts (C17orf87 and LOC650919), neither associated with a known function, were common to both lists. The genes most inhibited by GSK-J4 in the presence of either IFN-γ or IL-4 are illustrated by heat maps in Figures 8C,D, respectively, with the details of these and further genes listed in Table 5. Heading the list of IFN-γ upregulated, GSK-J4 downregulated genes were CCL7 and CCL8, which are known genes associated with activation by IFN-γ (40). Heading the list of IL-4 upregulated, GSK-J4 downregulated gene was CD209, which is an established IL-4 responsive gene (41). A selection of the more abundant transcripts that were inhibited at least twofold by GSK-J4, were chosen for further analysis. Because there were so few abundant IL-4 stimulated, GSK-J4 inhibited transcripts in the array, we also included CD206, which was used for initial sample validation but just failed to reach significance in the array experiment (i.e., a false negative). First, upregulation by IFN-γ or IL-4 and its reversal by GSK-J4 was confirmed by RT-qPCR (Figures 8E,F, respectively). GSK-J4 is non-selective for KDM6A and KDM6B (39). Hence, to distinguish the roles of KDM6A and KDM6B, the effects of silencing one, the other or both was investigated by using shRNA. Given that silencing is difficult in primary macrophages, we used adenovirus-mediated delivery of shRNAs selective for KDM6A or KDM6B and compared these to delivery of a control adenovirus that expressed shRNA against firefly luciferase (shLUC). The housekeeping gene, 36B4, was also studied as a further control. Based on mRNA levels, silencing of KDM6A and B was highly selective, albeit incomplete at the maximum tolerable adenovirus dose (Figure 9A). Specificity and efficacy was confirmed at the protein level for KDM6A (Figure 9B) but no suitable antibody is available for KDM6B. The effects of KDM6A and B silencing were measured in the presence of IFN-γ or IL-4. From these results, the induction of CCL7 by IFN-γ depended selectively on KDM6A, whereas that of VAMP5 required KDM6A and B redundantly (Figure 9C). The induction of CD206 by IL-4 also required KDM6A and B redundantly, whereas that of PALLD depended selectively on KDM6B (Figure 9C). We concluded that a subgroup of transcripts upregulated by IFN-γ or IL-4 depended on KDM6B, either on its own or redundantly with KDM6A.

Using a focused array, we demonstrated that IFN-γ or IL-4 modulate the mRNA expression of at least seven epigenetic regulators in human blood monocyte-derived macrophages, independently of any effects on cell proliferation. CIITA, KDM6B, and NCOA1 showed increased mRNA expression within 6 h of stimulation, whereas KAT2A, PRMT7, SETD6, and SMYD3 showed decreased expression that required at least 18 h of treatment. The effects were confirmed at the level of protein for SMYD3. Based on similar effects of growth factor depletion or overexpression of p27^kip1, we concluded that IFN-γ or IL-4 decreased expression of AURKB, ESCO2, SUV39H1 and WHSC1 mRNA and AUKB activity, at least partly, as a consequence of cell cycle arrest at the G1/S checkpoint. Furthermore, this decreased expression could be reversed by overexpression of E2F1, which is known from published chromatin immunoprecipitation (ENCODE) studies to bind directly to the relevant promoter regions. These data expand the list of epigenetic regulators the expression of which is regulated by IFN-γ and IL-4. As a first step to establishing whether these changes impact on macrophage phenotype, we demonstrated by pharmacological inhibition and shRNA silencing that KDM6B participates in a subset of the divergent gene expression changes in response to IFN-γ and IL-4.

In general, several families of enzymes that can alter the phosphorylation, acetylation, and methylation status of specific histone residues play a major role in epigenetic regulation (42–44). Histone H3S-10 phosphorylation catalyzed by AURKB has been implicated in chromatin condensation during mitosis (45, 46). Moreover, the finding that AURKB is an E2F target (29) led to the conclusion that it is a useful marker of cell proliferation, similar to PCNA. Our data showing that AURKB is downregulated by IFN-γ and IL-4-induced cell cycle arrest and is restored by the E2F1 extend these conclusions to primary macrophages. Previous work showing that the AURKB inhibitor, AZD1152, abrogates growth of human acute myeloid leukemia cells (47) and that growth arrest of mouse Raw264.7 macrophages by H. Pylori is associated with downregulation of AURKB (48) are also consistent with our conclusions.

Histone acetyl transferases (HATs) promote opening of the chromatin and enhance transcription, whereas histone de-acetylases (HDACs) have the opposite effect (49–51). The ability of CIITA to recruit HATs, including KAT2A (also known as pCAF-B) and NCOA1 (also known as SRC-1), to the promoter of the major histocompatibility complex-II (MHC-II) gene, has been extensively studied in macrophages (52, 53). Given the previous literature, CIITA upregulation by IFN-γ could be seen as a positive control for our array study. However, we also found CIITA to be upregulated to a lesser extent by IL-4, and both effects were independent of cell proliferation. Upregulation of NCOA1 selectively by IFN-γ most likely enhances the effects CIITA (52, 53) but downregulation of KAT2A seems paradoxical. However, this might also enhance the action of IFN-γ by decreasing acetylation and potentiating the functions of IRFs (54). The HAT, ESCO2, was also downregulated by IFN-γ or IL-4, although from our data this appeared to be mainly the consequence of the inhibition of cell proliferation. As its full name “Establishment of Sister Chromatid Cohesion N-Acetyltransferase” implies, ESCO2 has a known role in mitosis (30). It functions as part of the cohesion complex and its mutation leads to the cohesinopathy, Roberts syndrome (55). However, cohesin (and perhaps therefore ESCO2) has also been ascribed a wider role in gene transcription (56–58) and ESCO2 participates in Notch signaling (59), observations that might have implications for proliferating macrophages, although this remains to be investigated. Overall, our results imply that treatment with IFN-γ has the ability to both increase and decrease activity of specific HATs, thereby increasing or decreasing expression of different genes. In future experiments, beyond the present scope, it will be interesting to investigate the effects of manipulating levels of the HATs we have identified as up- or downregulated on both positive and negative transcriptional responses to IFN-γ. Recent data from the group of de Winther and colleagues showed, for example, that IFN-γ treatment specifically altered the acetylation status of the promoters of two downregulated genes, Il1b and Il6, in mouse macrophages (60), although the role of specific HATs and HDACs in these changes was not defined. We did not detect effects of IFN-γ or IL-4 on mRNA expression of any of the HDACs-1 to -11 that were included in our RT-qPCR screen. However, other mechanisms including changes in recruitment and activation of HATs and HDACs at specific promoters also contributes to acetylation status (61).

Histone methyl transferases (HMTs) and lysine de-methylases (KDMs) can be stimulatory or inhibitory to transcription depending on the site and degree of methylation that is optimal (62). For example, H3K4 methylation has important consequences for both enhancer and promoter activity of macrophage specific genes (63). Mono and di-methylation appear permissive for enhancers but tri-methylation for promoters of LPS-sensitive genes (61). A previous study on human macrophages stimulated with LPS and IFN-γ (64) showed increased H3K4 methylation associated with increased expression of the HMT, myeloid lymphoid leukemia (MLL). Given that we did not see any change in MLL expression with IFN-γ alone (absent from Table 2), it is possible that this is an effect of LPS, although this requires verification. In our study, the HMTs, PRMT7, SETD6, and SMYD3, were downregulated after priming by IFN-γ alone, independently of inhibition of proliferation. PRMT7 is a member of the PRMT histone arginine methylases, whereas SETD6 and SMYD3 are lysine methyl transferases. PRMT7 upregulates expression of metalloproteinase-9 (MMP-9) (65) in breast carcinoma cells but its role in macrophages is unknown. The related PRMT4 promotes major histocompatibility II (MHCII) gene expression (52). SETD6 activity has been linked to repression of the nuclear factor κB (NF-κB) system (66, 67) and upregulation of estrogen-responsive genes (68) in other cell types but there appears to have been very little previous work in mouse or human macrophages beyond demonstrating its presence preferentially in alternatively activated human macrophages, consistent with our results (64). SMYD3 di- and tri-methylates H3K4 residues (62). Although not an S-phase gene, SMYD3 has been identified as essential for cancer cell proliferation (69). It also plays a role in rescue from senescence (70), estrogen response (71), and MMP-9 induction (72) in various cancer cell lines. SMYD3 is also little studied in macrophages, although a previous study demonstrated its downregulation by a combination of LPS and IFN-γ, which is consistent with our findings (64).

In addition, the HMTs, SUV39H1, and WHSC1, were downregulated by IFN-γ or IL-4, at least in part, as a consequence of cell cycle arrest. Since SUV39H1 methylates H3K9 and places a repressive mark (62), it is predicted to reduce transcription of susceptible genes. Interestingly, one of the genes decreased by SUV39H1 in macrophages is p21^waf1 (73), the cyclin-dependent kinase inhibitor that is responsible for inhibition of proliferation by IFN-γ or IL-4 (34, 35). Conversely, expression of p21^waf1 indirectly downregulates SUV39H1, which implies a mutual feedback mechanism that presumably fine tunes the rate of proliferation. WHSC1 is a candidate gene implicated in Wolf-Hirschhorn syndrome, which is caused by deletions within the chromosome 4p16.3 region (74, 75). It has the ability to methylate several lysine residues in H3 and H4 (76) and could, therefore, act as a transcriptional activator or repressor. It has been ascribed a variety of functions, including in replicative DNA repair (33), which implies a role in S phase, but also in sustaining NF-κB pathway activity in tumors (77), which suggests activity may be present in other phases of the cell cycle. So far, there appears to be no knowledge regarding its role in macrophages.

We chose to prioritize the lysine demethylase, KDM6B, for study in greater detail because it removes repressive H3K27Me3 marks and is, therefore, a putative transcriptional activator. It is also amenable to selective pharmacological inhibition, which would lead to downregulation of target gene expression. Furthermore, KDM6B has been previously implicated in macrophage polarization by either bacterial LPS or IL-4, depending on the source of macrophages investigated. For example, KDM6B is upregulated in response to bacterial LPS in both mouse (78) and human (39) macrophages; and as many as 70% of LPS responsive genes in mouse macrophages recruit KDM6B to their promoters (15). This does not always lead to H3K27 demethylation (15) but KDM6A and B nevertheless act redundantly to potentiate responses to LPS in human macrophages (39). Other work in mouse macrophages showed that KDM6B can be upregulated by IL-4 in a STAT-6-dependent manner and that it is essential for IL-4 induced polarization in vitro and in response to certain kinds of parasitic infection in vivo (14, 17). However, no previous study has investigated the role of KDM6B on both pro-inflammatory and anti-inflammatory polarization in the same preparation of macrophages. Our transcriptomic study demonstrated for the first time that KDM6A and B play a part in polarization by IFN-γ, although a smaller proportion of genes (approximately 20%) appear to be affected than for responses to LPS (39). Likewise, KDM6B modulates some IL-4 polarization genes alone or redundantly with KDM6A in human macrophages but this seems to be a much small proportion (11%) than in mouse macrophages (14, 17). The finding that KDM6A and B acted redundantly for some processes is consistent with the previous study on LPS (39).

Our studies significantly expand knowledge of the expression changes in epigenetic regulators during polarization of human macrophages. Upregulation or downregulation of genes does not necessarily imply that these will be the only enzymes that play a major role in responses to IFN-γ or IL-4. However, the previous literature on CIITA and NCOA1 together with our new results with KDM6B illustrates the importance of upregulated genes. Importantly, we identify several changes that are independent of inhibition of proliferation, a complicating factor that appears to have been overlooked in previous studies. However, those changes partly dependent of proliferation may also be of significance. Indeed, recent work has highlighted the importance of proliferation in replenishing populations of resident macrophages (79). There have also been interesting findings in models of inflammation, especially atherosclerotic plaque formation (80), suggesting that proliferation rather than recruitment may play the major role in sustaining macrophage numbers. From our data, proliferation has a major impact on epigenetic programing, and this undoubtedly influences macrophage behavior. Future work, beyond the present scope, should probe into the downregulated genes we have identified, many of which are virtually unstudied in macrophages. However, we recognize that mechanisms other than changes in expression level regulate the function of epigenetic writers, readers and erasers at the promoters and enhancers of pro- and anti-inflammatory genes (61, 81). Except in the case of KDM6B, the contribution that the expression changes we observed make to epigenetic regulation in macrophages remains to be established.