Differentiation from MOs to DCs and from MOs to MACs results in cell-type-specific demethylation of thousands of genes
To dissect the downstream contribution of IL-4 in the acquisition of cell-type-specific DNA methylation changes in MO-to-DC differentiation, we generated three sets of matching samples corresponding to MOs from human peripheral blood, immature DCs (iDCs) and immature MACs (iMACs), following incubation of MOs with GM-CSF/IL-4 and GM-CSF only, respectively. Mature DCs (mDCs) and MACs (mMACs) were then created by exposing iDCs and iMACs to LPS (Fig. 1a). The comparison of MO-to-DC and MO-to-MAC differentiation allowed us to isolate the specific effect of IL-4, which is the differential factor in these two processes. We monitored these processes by testing different markers using RT-PCR (Additional file 1A) and fluorescence-activated cell sorting (FACS; Additional file 1B). For instance, quantitative RT-PCR demonstrated upregulation of DC markers (CD209) and mature DC markers (CD83), and that the level of expression of CD14 receptor was high in MOs, intermediate in MACs and low/negative in DCs. FACS analyses revealed that MOs were efficiently differentiated to iDCs (87–93 %, according to CD209 and CD206) and iMACs (88 %, according to CD206), and indicated a shift in the CD83 and CD86 markers that support efficient maturation of these cells, generating 79 % mDCs and 81 % mMACs (Additional file 1B).
We then performed DNA methylation profiling using bead arrays that interrogated the DNA methylation status of >450,000 CpG sites across the entire genome, covering 99 % of RefSeq genes. Statistical analysis of the combined data from the three biological replicates of MO-to-DC and MO-to-MAC revealed large changes in DNA methylation during the differentiation step (1,780 and 2,644 CpG sites, respectively). In contrast, only a few genes displayed differential DNA methylation during the maturation step (75 and 27 CpG sites for DC and MAC maturation, respectively) (Fig. 1b and Additional file 2). In all cases, demethylation prevailed over gains in DNA methylation, consistent with the results reported by others [6, 17, 18]. Specifically, demethylated CpG sites represented 92.9 % of total differentially methylated CpGs in MO-to-iDC (1,654 CpGs) and 97.8 % of differentially methylated CpG in MO-to-iMAC (2,586 CpGs). This contrasts with the findings in MO-to-OC differentiation, in which de novo deposition of the DNA methylation occurs to a similar extent as demethylation [9]. Changes corresponding to the average of three sample sets were almost identical to the pattern obtained for each individual sample, highlighting the specificity of the differences observed (Fig. 1c and Additional file 3A). The results for DC differentiation were similar to those reported by Zhang and colleagues [17] (78 % of the demethylated genes in their study were present in our own data, Additional file 3B).
Although most CpGs displaying a loss of methylation were common to the two differentiation processes (Fig. 1d), a significant fraction of demethylated CpGs were specific to each process: 14.2 % in MO-to-iDC differentiation (235 genes) and 45.1 % in MO-to-iMAC differentiation (1,167 genes). This implies that DNA demethylation may be important in determining the differences between the lineages. Given that IL-4 is the only cytokine to differ between these two processes, our findings suggest that events downstream of IL-4 may not only be responsible for the set of genes specifically demethylated in DCs, but also may directly block demethylation of those that are specific to MACs.
An analysis of the distribution of CpGs with a significant decrease in DNA methylation (Fig. 1e) revealed that most of them map to gene bodies (789 CpGs in MO-to-iDCs and 1,242 in MO-to-iMACs). Over 22 % were located at intergenic regions in both differentiation processes. Only about 15 % of the changes occurred near the transcription start site (TSS) (Fig. 1e). This reinforces the notion that a high proportion of the changes occur in regulatory regions outside promoters, such as enhancers located in the body of genes. Indeed, using the Illumina annotation tool we determined that 41 % and 43 % of all demethylated CpG sites in DC and MAC differentiation, respectively, are located in enhancers. The proportion of enhancers was particularly high in gene bodies and intergenic regions, as expected (Fig. 1e).
Gene ontology (GO) analysis of hypomethylated CpG revealed significant enrichment [false discovery rate (FDR) < 0.05] of a variety of functional categories important in iDC and iMAC differentiation and function, including inflammatory and innate immune response (Fig. 1f). These data suggest that DNA demethylation is targeted to genomic regions that are activated during DC and MAC differentiation.
As expected, we identified changes in several genes involved in DC and MAC function among the group of demethylated genes in both iDCs and iMACs (Additional file 4). For example, CSF1R, which codes for the receptor of the cytokine CSF1 involved in MAC differentiation, and CCL22, a cytokine that is released by DCs and MACs, were dramatically demethylated (Additional file 4). iMACs displayed specific demethylation on CCL20, an inflammation chemokine, and IL1B, a cytokine involved in immune and inflammatory response. We observed very specific demethylation in DC differentiation at a CpG site in the gene bodies of DUOX1, an oxidase involved in the antimicrobial-mediated response, and the signalling receptor SLAMF1 (Additional file 4).
We then confirmed the robustness of the DNA methylation data in MO-to-iDC and MO-to-iMAC differentiation by bisulfite genomic pyrosequencing of CpG sites. The selection included genes that were demethylated in both differentiation processes (CSF1R, CCL22, DUSP5), and some that were only demethylated in MAC (IL1B, ARSB, CCL20) or DC differentiation (SLAMF1, DUOX1, PFAS). In all cases, bisulfite pyrosequencing confirmed the results of the beadchip array (Fig. 1g and Additional file 3C) and the demethylation at the aforementioned genes. It is well established that terminal differentiation of MOs into DCs/MACs occurs in the absence of cell division, indicating the occurrence of active DNA demethylation mechanisms. To further confirm this, we measured the extent of DNA replication in our DC and MAC differentiation experiments by treating cells with BrdU pulses. Consistent with previous observations [18], we found no significant differences between the negative control and the BrdU pulses, implying that DNA methylation changes observed during this period were independent of DNA replication (Additional file 1C). The participation of active DNA demethylation events in this process is reinforced by the previous findings of our and other groups [6, 9]. In fact, we observed changes in 5hmC, which is an intermediate oxidized base, resulting from TET2 activity and leading to active demethylation (Additional file 3D).
To test the implication and functionality of the methylcytosine dioxygenase TET2 in demethylation during DC and MAC differentiation, we downregulated TET2 levels using siRNA transfection against various TET2 sites and compared it to transfection with a control siRNA before DC/MAC differentiation was induced. TET2 downregulation partially impaired demethylation of both common and DC/MAC-specific genes. The impairment was partial because of a technical aspect related to the inability to achieve the maximum downregulation of TET2 before the differentiation processes had already started. In addition to the reduced demethylation, TET2 downregulation also resulted in a decrease in surface CD209 and CD83 markers; together with an increase in CD14 (which is higher in MOs than in DCs and MACs) (Additional file 5), demonstrating the functionality of DNA demethylation during these two processes.
Expression changes and their relationship with DNA demethylation in MAC and DC differentiation and maturation
To further investigate the functionality of DNA methylation changes, we generated expression profiles for the same cell types (MOs and derived iDCs, iMACs, mDCs and mMACs). We noted large changes in expression in both processes. Specifically, we observed upregulation of 2,920 and 3,095 genes and downregulation of 1,513 and 1,476 genes during the differentiation of MOs to iDCs and to iMACs, respectively (>2-fold change or <0.5-fold change; p-value < 0.01; FDR < 0.05) (Fig. 2a). We also identified large changes in the maturation process, whereby 927 and 1,461 genes were upregulated, and 1,961 and 2,829 were downregulated in the maturation from iDCs and iMACs to mDCs and mMACs, respectively, after LPS-mediated activation (Fig. 2a). Unlike changes in DNA methylation, which occurred primarily in the direction of demethylation and were concentrated in the differentiation of MOs to iDCs and iMACs, expression changes occurred in the direction of upregulation and downregulation, and large changes were observed during differentiation and maturation. A high proportion of expression changes were common to the processes of differentiation into DCs and MACs (Fig. 2b). Specifically, 73.12 % and 68.98 % of the upregulated genes and 72.24 % and 74.05 % of the downregulated genes were common to MO-to-iDC and MO-to-iMAC differentiation, respectively, whereas 54.37 % and 34.49 % of the upregulated genes and 61.09 % and 42.88 % of the downregulated genes were common to the two maturation processes.
To investigate the relationship between DNA methylation and expression changes, we compared the two data sets, focusing on genes that underwent significant demethylation. We found that DNA demethylation events were associated with both gene upregulation and downregulation (Fig. 2c), although most genes that became demethylated were overexpressed (70.4 % for MO-to-iDC and 67.1 % for MO-to-iMAC). We also examined whether the location of a given CpG site was related to the effects on expression. CpGs located in the TSS200 and the first exon had the strongest association between demethylation and overexpression (Fig. 2d) for both MO-to-iDC and MO-to-iMAC. Analysing the list of genes that were both demethylated and overexpressed during the differentiation step revealed the enrichment of categories of genes that are functionally relevant to DC and MAC biology (Additional file 6). For instance, we observed that genes in the inflammasome pathway that leads to IL1-mediated inflammation (including PYCARD, IL1B and IL1A, which act together during the MAC innate response [19]) were demethylated and overexpressed during MAC differentiation. The inflammasome sensor protein gene AIM2 [20] was also demethylated during MAC differentiation and overexpressed in the MAC maturation step, strongly suggesting the need for additive signals to trigger this supramolecular inflammatory system.
As mentioned above, most DNA methylation changes occur at the differentiation level, both for DC and MAC differentiation, whereas large expression changes occur at the activation step, suggesting that a proportion of genes may undergo DNA methylation changes before their expression levels change. Indeed, we identified a set of genes for DC and MAC differentiation/maturation that became demethylated during differentiation but were only overexpressed at the maturation level (Fig. 2e), as if demethylation were priming these genes for upregulation for when they need to be expressed, that is, for when DCs or MACs encounter a compound such as LPS. Some of these genes were common to DCs and MACs, but others were specific to each cell type (Fig. 2f). Among these genes we identified some like IL1B and CCL20 that undergo DNA demethylation during MAC differentiation, but only achieve overexpression in MACs following LPS treatment (Fig. 2g) (Additional file 6). In such cases, time-course analysis of histone modifications like H3K27me3 and H3K9me3 revealed that changes in these marks also precede LPS-mediated stimulation (Fig. 2h and Additional file 7), suggesting that other regulatory elements are directly responsible for activation of these genes once the chromatin context is suitable. Interestingly, the increase in these two heterochromatic marks took place in DCs, and not in MACs, where expression does not increase upon LPS-mediated stimulation.
Other genes had different relationships with DNA methylation changes, suggesting a variety of functional consequences associated with DNA demethylation observed at the differentiation step (Additional file 7).
Inhibition of the JAK3-STAT6 pathway impairs DNA methylation and expression changes of DC-specific genes and is a positive switch for changes at MAC-specific genes
IL-4 signalling is crucial and indispensable to the development of human MO-derived DCs. One of the most important outcomes of our DNA methylation analysis was the identification of a subset of genes that are specifically demethylated in DC differentiation in response to IL-4. To address the role of IL-4 in driving these DC-specific DNA methylation changes, we studied the contribution of signalling mediators downstream of IL-4R. Membrane-bound type I IL-4R activates the tyrosine kinase JAK3, which phosphorylates STAT6 at Tyr641, leading to its translocation to the nucleus and binding to target genes [21–23] (Fig. 3a). To examine the role of the IL-4-JAK3-STAT6 pathway in the acquisition of DC-specific DNA methylation and expression changes, we first tested the impact of JAK3 inhibition on the regulation of the aforementioned genes. To this end, we first used a JAK3-selective inhibitor, PF-956980 [24]. We differentiated MOs to DCs and MACs with two different concentrations of PF-956980 to select the conditions under which it is active. STAT6 phosphorylation, which renders STAT6 into its active form, is only present under the conditions for DC differentiation and not for MAC differentiation (when IL-4 is absent). As expected, STAT6 phosphorylation disappeared following JAK3 inhibition with 400 nM and 1,000 nM PF-956980 (Fig. 3b). In the case of MACs, we did not observe STAT6 phosphorylation, given the lack of stimulation of JAK3, and therefore the addition of PF-956980 did not make any difference (Fig. 3b). Treatment with PF-956980 affected the presence of the surface markers CD209 and CD83 during GM-CSF/IL-4-mediated differentiation to DCs (Fig. 3c and Additional file 8), resulting in the generation of profiles closer to those displayed by MACs. This demonstrates the functional effects of PF-956980 in inhibiting DC differentiation.
The effect of PF-956980 was very specific to the impairment of demethylation of DC-specific genes in DC differentiation (Fig. 3d), and had little effect on the demethylation of MAC-specific genes in MAC differentiation or in genes that are commonly demethylated in both DC and MAC differentiation (Fig. 3d and Additional file 8B). Interestingly, in the presence of JAK3 inhibitors and under the conditions required for DC differentiation, DNA methylation levels of genes that were specifically demethylated in MO-to-iMAC differentiation resembled those observed in the absence of IL-4, indicating that inhibition of the pathway downstream of IL-4 removed the constraints on this set of genes towards their DNA demethylation under the standard conditions for DC differentiation (Fig. 3d).
In general, the effects at the expression level were as expected, and impaired DNA demethylation was associated with diminished overexpression of DC-specific genes during differentiation. Most notably, we observed impaired overexpression of genes that only underwent expression changes in the maturation step, once DNA demethylation had been inhibited through the action of JAK3 inhibitors (Additional file 8C).
To explore the extent of the role of the IL-4-JAK3-STAT6 pathway in the acquisition of the DC-specific methylation signature, we performed a new methylation profiling to test the effects of inhibiting JAK3. A comparison of MOs with MOs differentiated to iDCs and iMACs both in the presence and absence of the JAK3 inhibitor PF-956980 revealed that the DNA methylation patterns of iDCs incubated with PF-956980 cluster together with iMACs (Additional file 8D). In other words, treatment with PF-956980 erases the DC-specific signature, and renders a DNA demethylation pattern indistinguishable to that of iMACs (Fig. 3e). The specific analysis of some of the previously studied genes confirmed this effect (Additional file 8E).
To unequivocally test the potential causal relationship between JAK3 and STAT6 in the demethylation of DC-specific genes, we investigated the consequences of ablating JAK3 and STAT6 expression in MOs. We downregulated JAK3 and STAT6 levels in MOs using transient transfection experiments with siRNA cocktails that target different sites for each of these two proteins in comparison with a control siRNA. Twenty-four hours after transfection, we induced DC differentiation with GM-CSF/IL-4. Under these conditions, we used a western blot to check the effects on JAK3 and STAT6 levels 4 days after GM-CSF/IL-4 stimulation of MOs. This method enabled us to confirm that the STAT6 and JAK3 were downregulated by close to 50 % and 20 % (Fig. 3f), respectively. As a result, we observed a noticeable shift of the surface DC markers CD209 and CD83 (Additional file 9A).
We then checked the effects of JAK3 and STAT6 depletion on the demethylation of DC-specific, MAC-specific and DC/MAC common genes. Similar to the results obtained from the pharmacological inhibition of JAK3, siRNA-mediated depletion of JAK3 and STAT6 very specifically impaired the demethylation of DC-specific genes in DC differentiation (Fig. 3g) and had little effect on the demethylation of MAC-specific genes in MAC differentiation (Additional file 9B) or in genes that are commonly demethylated in both DC and MAC differentiation. These results not only confirmed the participation of JAK3, downstream to IL-4, in the demethylation of DC-specific genes, but also the participation of STAT6, the target of JAK3.
Constitutively activated STAT6 induces demethylation of DC-specific genes during GM-CSF-only differentiation
To further investigate the potential direct involvement of STAT6 in the demethylation of DC-specific genes, we performed chromatin immunoprecipitation (ChIP) assays with STAT6. We found that STAT6 did interact specifically with DC-specific genes like DUOX1 and SLAMF1 in DC differentiation (Fig. 4a), whilst there was no binding of these genes during MAC differentiation. Interestingly, pharmacological inhibition of JAK3 led to impaired binding of STAT6 in DC-specific genes (Fig. 4a), reinforcing the notion of the dependence on IL-4 and JAK3 for this interaction. We then investigated whether STAT6 interacts with TET2, either directly or through other intermediates such as PU.1. However, we were unable to identify any direct interaction between STAT6 and TET2 (not shown). It should be noted that these experiments are technically challenging, and such an interaction cannot be fully discounted. An alternative mechanism may be provided if STAT6 recruits PU.1, which has in turn been proven to recruit TET2 [9] in a related MO differentiation process. In fact, synergism between STAT6 and PU.1 has been previously shown [25]. To test whether PU.1 participates in demethylating these genes, we also performed siRNA experiments against PU.1. We determined that PU.1 downregulation also impairs demethylation of some of these genes, although in a less specific manner than STAT6. In addition, we also observed an effect on the surface markers of both DCs and MACs (Additional file 10).
To conclusively establish the role of IL-4/JAK3-dependent demethylation of DC-specific genes via STAT6, we performed gain-of-function experiments in MOs stimulated exclusively with GM-CSF and transfected with a constitutively activated form of STAT6. STAT6VT carries two amino acid changes in the SH2 domain that affect the overall structure and stability of the monomeric and dimeric protein [26] (Fig. 4b). When overexpressed in mammalian cells, STAT6VT undergoes tyrosine phosphorylation; is translocated to the nucleus (Fig. 4c), where it binds DNA; and activates transcription in an IL-4-independent manner. We infected MOs with a green fluorescent protein (GFP)-expressing lentiviral MIG vector (pCDH-MIG) containing STAT6VT and, in parallel, an empty GFP-expressing MIG vector as a negative control. Following infection, we stimulated cells with GM-CSF in the absence of IL-4, that is, under our conditions for MAC differentiation. Infection of MOs with the GFP-expressing empty vector achieved higher levels than those with STAT6VT GFP vector, probably due to the lower titre of lentiviruses containing a larger construct (Fig. 4d). In any case, we were able to isolate GFP+ cells in both conditions, following 9 days after GM-CSF stimulation. Not surprisingly, the ectopic expression of STAT6VT resulted in increased levels of the DC-specific marker DC-sign, following GM-CSF stimulation and in the absence of IL-4 (Fig. 4e). We then performed bisulfite pyrosequencing of DC-specific and MAC-specific genes and found that STAT6VT overexpression was able to induce demethylation of DC-specific genes, such as DUOX1 and SLAMF1, by-passing IL-4R upstream signalling (Fig. 4f). In addition, the MAC-specific genes CCL20 and ARSB, which are normally demethylated in the presence of GM-CSF, did not become demethylated under the presence of STAT6VT (Fig. 4f), strongly indicating that STAT6 prevents their demethylation under the conditions of DC differentiation. In contrast, genes that were demethylated under our standard DC and MAC differentiation conditions (like CCL22 and CSF1R) were not affected by the overexpression of STAT6VT. In summary, STAT6 is not only responsible for demethylating DC-specific genes but also for preventing demethylation of MAC-specific genes.
Altogether, our results demonstrate a direct relationship between the extracellular stimulation through IL-4 leading to MO-to-DC differentiation and the acquisition of DC-specific DNA methylation and expression patterns, together with the inhibition of MAC-specific genes. Moreover, we prove the role of the IL-4-JAK3-STAT6 pathway in instructing the cell epigenome to engage a specific differentiation state towards DCs, at the expense of MAC differentiation.