Genome-wide identification and analysis of transcribed enhancers during macrophage polarization

Background Macrophages are sentinel cells that play essential role in tissue homeostasis and host defence. Owing to their plasticity, macrophages acquire a range of functional phenotypes in response to microenvironmental stimuli, of which M1 (IFNγ) and M2 (IL-4/IL-13) phenotypes are well-known for their opposing proand antiinflammatory roles. Enhancers have emerged as regulatory DNA elements crucial for transcriptional activation of gene expression, with recent studies highlighting their importance in macrophages. Results Using cap analysis of gene expression (CAGE) and epigenetic data, we identify on a large-scale transcribed enhancers in mouse macrophages, their time kinetics and target protein-coding genes. We observe an increase in target gene expression, concomitant with increasing numbers of associated enhancers and find that genes associated to many enhancers show a shift towards stronger enrichment for macrophage-specific biological not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/163519 doi: bioRxiv preprint first posted online Jul. 14, 2017;


Background
Macrophages are innate immune system sentinel cells that mediate homeostatic and protective functions, including host defence against invading pathogens [1].Macrophages respond to a wide range of external stimuli by acquiring heterogeneous activation states that exert functional programs tailored for specific microenvironments [2].Recently, a classification system for activated or polarized macrophages has been developed, but still remains controversial [2,3].Two functionally extreme subpopulations of activated macrophages are traditionally highlighted [4].M1 or classically activated macrophages are pro-inflammatory macrophages that are polarized in response to particular stimuli including IFNγ, LPS and TNF [3].M1 macrophages are characterized by efficient antigen presentation, high bactericidal activity and production of pro-inflammatory cytokines, reactive oxygen and nitrogen intermediates [3,5].M2 or alternatively activated macrophages are induced by stimuli such as IL-4 and IL-13 and are predominantly regulatory macrophages involved in homeostasis, angiogenesis, wound healing, tissue remodelling and parasitic and bacterial infection [1,2,[6][7][8].M2 macrophages release anti-inflammatory cytokines and show less efficient antigen presentation and decreased production of pro-inflammatory cytokines [2,6].Macrophage polarization towards M1 or M2 phenotypes is driven by specific transcriptional changes and is controlled by complex cellular mechanisms [3,9].
Imbalance in M1 and M2 populations of macrophages with opposing pro-and anti-inflammatory roles has been implicated in disease progression [1].Intracellular pathogen Mycobacterium tuberculosis, the causative agent of tuberculosis, interferes with M1 polarization of macrophages and promotes M2 polarization, thereby avoiding anti-bacterial action of M1 macrophages [10,11].Tumour microenvironments promote phenotypic switches from M1 to M2 macrophages, which might contribute to the tumour progression by inhibiting immune responses to tumour antigens [1,2].Conversely, the phenotypic switch from a M2 population of macrophages to M1 might contribute to obesity and metabolic syndrome [1,2,12].Therefore, the development of techniques for manipulation and specific targeting of macrophage populations could ultimately improve diagnosis and treatment of inflammatory diseases.[2].To advance this area of research, the cellular mechanisms responsible for macrophage polarization need to be further deciphered.
not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint (which was .http://dx.doi.org/10.1101/163519doi: bioRxiv preprint first posted online Jul.14, 2017; Gene expression in eukaryotic cells is a complex process guided by a multitude of mechanisms [13].Precise regulation is required to ensure dynamic control of tissue-specific gene expression and to fine tune the responses to external stimuli [14].One such level of control is facilitated via regulation of RNA transcription.This process is mediated by a complex transcriptional machinery with its components recognising specific regulatory regions of DNA.Promoters represent a better-characterized class of such regions from which RNA transcription is initiated [15,16].They act in concert with other cis-regulatory DNA elements, including enhancers, which are believed to play key roles in transcriptional regulation [17]. Enhancers are defined as regulatory DNA regions that activate transcription of target genes in a distance-and orientation-independent manner [17].According to the dominant model, transcriptional regulation by enhancers is exerted via direct physical interaction between enhancer and target gene promoter mediated by DNA looping [17,18].Recent identification of distinct properties of enhancer regions enabled novel approaches to enhancer profiling [17].Enhancer regions are often distinguished by a specific combination of chromatin marks present at these locations, such as H3K4me1 and H3K27ac [19,20].Enhancer sequences contain transcription factor binding sites (TFBS) that recruit transcription factors (TFs) to regulate target genes [21,22].In addition, enhancers are frequently bound by proteins such as histone acetyltransferase p300 and insulator-binding protein CTCF [20,[23][24][25].Large-scale profiling of these enhancer-associated signatures by chromatin immunoprecipitation followed by sequencing (ChIP-seq) [25,26] has greatly advanced enhancer identification and enabled systematic and genome-wide enhancer mapping [27,28].Another group of methods such as chromosome conformation capture (3C) [29] and its variant Hi-C [30] has been employed to profile physical DNA contacts, including those between promoters and enhancers [31,32].However, none of these methods has become a gold standard of enhancer detection, and the field is still actively developing.
Recent studies have led to the unexpected finding that most active enhancers recruit RNA polymerase II and are bi-directionally and divergently transcribed to produce RNA transcripts, referred to as eRNAs [33,34].While the functionality of eRNA remains controversial, a recent study by Hon et al. showed that many enhancers are transcribed into potentially functional long-noncoding RNAs (lncRNAs) playing a role in inflammation and immunity [35,36].Recently, quantification of eRNA transcription laid the foundation for a novel method of large-scale enhancer profiling [37].In their study utilizing cap analysis of gene expression (CAGE) [38], Andersson et al. performed genome-wide mapping of transcriptional events followed by identification of enhancers based on co-occurrence of closely located divergent transcripts representing eRNAs [37].The capacity of CAGE to simultaneously profile the expression of eRNAs and genes became an additional advantage, since eRNA production was shown to positively correlate with the production of mRNAs of target genes [33,39].
These and other studies unravelled the fundamental importance of enhancer regions as DNA regulatory elements in multiple cell types, including macrophages [27,28,34,39,40].Enhancers are extremely widespread, with an estimation of up to one million enhancers in mammalian genomes [19,22,23,41].They are major determinants of gene expression programs required for establishing cell type specificity and mediating response to extracellular signals [22,42,43].Our current understanding of these elements, however, remains incomplete.High tissue-specificity of enhancers is a major hurdle towards establishing a comprehensive catalogue of the full enhancer population [22,42].Moreover, emerging evidence indicate that enhancers enhancers and promoters in the correlation-based E-P pairs was significantly smaller at 191,033nt as compared to 278,735nt for all TAD-based pairs (Additional file 1: Figure S1b).We further investigated associations between enhancers and target protein-coding genes (Additional file 2).Of all 10,767 protein-coding genes with CAGE expression (see Methods), 4,149 genes (38.5%) were not associated to any enhancer (Fig. 1b, upper panel).Given previous evidence of additive action of enhancers [17,49], we asked whether genes regulated by different numbers of enhancers have different gene expression levels.Genes without associated enhancers were overall lower expressed than genes associated to one (two-sided Wilcoxon signed-rank test p-value < 2.2*10 -16 ) or more enhancers.A steady increase in gene expression concomitant with higher numbers of associated enhancers (Fig. 1b, Kruskal-Wallis rank sum test p-value < 2.2*10 -16 ) was observed, supporting the model of additive enhancer action.
We further asked whether genes associated to different numbers of enhancers within the enhancer-gene interactome show functional differences.Gene set enrichment analysis (GSEA) was performed for gene sets of similar size to avoid a size-related bias (see Methods).The 1,416 genes associated to a single enhancer were enriched for general cellular pathways including "Aminoacyl-tRNA biosynthesis" and "Ubiquitin mediated proteolysis", as well as a few inflammation-related pathways (Fig. 1c).However, in contrast, the 1,358 genes associated to more than eight enhancers showed stronger enrichment for macrophage-related terms, such as "Jak-STAT signaling pathway" and "Chemokine signaling pathway" (Fig. 1d).GSEA for 1,306 genes associated to three or four enhancers showed enrichment for a combination of general cellular and macrophagerelated pathways (Additional file 3).Finally, the larger set of 4,149 genes not associated to any enhancer showed the strongest enrichment for general cellular pathways (Additional file 4).Hence, a shift towards stronger enrichment for macrophage-related pathways was a concomitant of higher numbers of associated enhancers.

Macrophage-specific expression
We opted for a similar strategy as Yao et al. [50] (see Methods) to uncover eRNAs and genes with higher expression in macrophages as compared to other FANTOM5 mouse tissues (further referred to as macrophagespecific).We identified 1,844 macrophage-specific and 8,923 non-macrophage-specific genes (Additional file 1: Figure S2).These two sets showed differences in numbers of associated enhancers, with 65.6% of macrophagespecific genes being associated to more than one enhancer, whereas this proportion dropped to 44.7% for nonmacrophage-specific genes (Fig. 2a).These results were in agreement with our observation of stronger enrichment for macrophage-related functions in genes associated to many enhancers.Similar to the trend observed above, both macrophage-specific and non-macrophage-specific genes showed higher gene expression concomitant with higher numbers of associated enhancers, with non-macrophage-specific genes showing lower expression levels than macrophage-specific ones (Additional file 1: Figure S3).Among 8,667 active enhancers, 54.7% were deemed macrophage-specific (see Methods), in agreement with known tissue-specificity of enhancers [22,41,42].Interestingly, non-macrophage-specific enhancers still showed higher eRNA expression in macrophages as compared to the non-macrophage samples (Fig. 2c, left not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint (which was .http://dx.doi.org/10.1101/163519doi: bioRxiv preprint first posted online Jul.14, 2017; panel).This may be explained by the fact that for this analysis we excluded all enhancers that showed zero eRNA expression in the majority of our macrophage samples (see Methods).
Next, we asked whether these two enhancer sets could regulate genes with different functions.Genes associated exclusively to macrophage-specific enhancers, as well as genes associated exclusively to non-macrophagespecific enhancers were sub-selected.As expected, genes in the former set showed overall higher expression in macrophage samples as compared to the non-macrophage samples (Fig. 2b, right panel).In contrast, expression of genes associated exclusively to non-macrophage-specific enhancers was lower in macrophage samples.Interestingly, the opposite was observed for non-macrophage-specific enhancers (Fig. 2c).Genes associated to macrophage-specific enhancers were enriched for both general and macrophage-related processes (Fig. 2d).This observation reflects the fact that production of macrophage-specific factors and activation of housekeeping processes that facilitate it might be both regulated by the same set of enhancers.Genes associated to nonmacrophage-specific enhancers were enriched for only four KEGG pathway maps with FDR < 0.05 (Fig. 2e), none of which can be considered a typical macrophage pathway.We obtained consistent results when we repeated the analysis for a subset of 500 genes with the highest expression in macrophages (Additional files [5][6].Taken together, these findings demonstrate that most of the identified active enhancers in macrophages show macrophage-specific eRNA expression and regulate genes with macrophage-specific as well as general cellular functions.

Stimuli-induced transcriptional changes
We set out to determine transcriptional changes that were dynamically induced by IFNγ (M1) and IL-4/IL-13 (M2) in mouse macrophages, and to infer enhancers important in these processes (Fig. 3a).M1-and M2responsive enhancers and genes were identified as those up-regulated upon stimulation; regulatory associations were retained for pairs with a positive correlation of expression in the corresponding polarization state (see Methods).In this manner, we discovered 115 M1-responsive enhancers regulating 105 M1-responsive genes (further referred to as sets E1 and G1), as well as 131 M2-responsive enhancers regulating 98 M2-responsive genes (sets E2 and G2) (Fig. 3b and Additional files 8-9).Notably, 77% of E1 and 71% of E2 enhancers were deemed macrophage-specific in our settings.GSEA of G1 and G2 gene sets showed significant enrichment for GO and KEGG terms relevant to immune system and macrophage functions (Fig. 3c and Additional file 1: Figure S4).These results highlight the importance of enhancer regulatory control during macrophage polarization and suggest a striking influence of cytokine stimulation on activation of enhancers, which, in turn, drive some of the transcriptional responses seen during M1 and M2 polarization.
M1 and M2 macrophages are known to possess different phenotypes and functions [2].As expected, G1 and G2 sets had only 19 genes in common.Similarly, a small overlap of only 14 enhancers was observed for E1 and E2 sets.Moreover, enhancers and genes selected as stimuli-responsive for a single activation state showed significant differences in time-course expression in M1 and M2 macrophages (Fig. 3d).These data indicate that M1 and M2 macrophages not only differ in their gene expression profiles, but also differ in their active enhancer repertoire that likely drives observed gene expression changes.
Previous studies reported and exploited positive expression correlation of eRNA and target genes [33,37,50].
Hence, we compared expression correlation of E1-G1 and E2-G2 pairs in M1 and M2 macrophages (Fig. 3e) to not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.

Marker genes of macrophage activation are regulated by stimuli-responsive enhancers
We further asked which known marker genes of macrophage activation [1-3, 6, 51] were identified in our setting (Table 1).Out of 20 M1 and 26 M2 marker genes, we found eight M1 marker genes in the G1 set and eight M2 marker genes in the G2 set, that were associated to responsive enhancers in the respective stimulation (significant overlap with hypergeometric test p-value < 10 -10 ) (Table 1).The G1' set contained an additional four M1 marker genes (Gpr18, Il12b, Il6, Inhba) and the G2' set an additional three M2 marker genes (Il27ra, Klf4, Myc), which, although stimuli-responsive themselves, were not associated to stimuli-responsive enhancers.Next we inferred potential marker enhancers that regulate marker genes specifically during M1 or M2 polarization.Each of the 16 M1 and M2 marker genes in G1 and G2 was associated to a minimum of one and maximum of nine enhancers in the E1 and E2 stimuli-responsive sets, respectively (Table 1).Of those, we identified enhancers that were selectively responsive in a single polarization state and showed higher expression in this state as compared to the other polarization (see Methods).A total of 13 M1 and 22 M2 enhancers were inferred as potential polarization markers (Table 1).
Interestingly, three of the M1 marker genes, Cxcl9, Cxcl10, and Cxcl11 are located within one TAD and are coregulated by a group of three marker enhancers (Fig. 4a-c).These enhancers, along with the two marker enhancers regulating Cxcl10 or Cxcl11 but not Cxcl9 (Table 1) are located in close proximity, in the intronic regions of the Art3 gene (Fig. 4c).These enhancer regions were previously reported to show induced RNA polymerase II binding in macrophages upon stimulation with LPS, one of the known M1 activators [34].In addition, these marker enhancer regions were shown to carry H3K4me1 enhancer histone marks in untreated macrophages [47].Moreover, H3K27ac modification, associated with active enhancers, is stronger enriched in these regions in M1 as compared to M2 and untreated macrophages [47] (Fig. 4c), providing further evidence of their functionality in macrophage M1 polarization.Among M2 marker genes, Arg1 as expected, is substantially expressed in M2 polarized macrophages but has extremely low expression in M1 and untreated macrophages (Additional file 1: Figure S5a).We found a single M2-responsive enhancer that might drive expression of Arg1 in M2 macrophages and might serve as a marker enhancer (Table 1 and Additional file 1: Figure S5b).On the contrary, M2 marker gene Egr2, a TF that activates macrophage-specific genes [52], is associated to as many as nine M2-specific enhancers (Table 1).Egr2 showed immediate up-regulation in response to both IFNγ and IL-4/IL-13 stimulation, however, in M2 macrophages the up-regulation sustained for up to 24 hours, whereas in M1 macrophages expression dropped rapidly after 2 hours (Fig. 4d, upper panel).Time-course eRNA expression for two Egr2 marker enhancers with the highest expression at 2 and 4 hours is shown in Fig. 4d.The distribution of all nine Egr2 marker enhancers within a TAD (Fig. 4e) may suggest that the regions identified as nine individual enhancers potentially demarcate fewer regions of stretch enhancers [53,54].We observed a similar distribution for enhancers of M2 marker gene Igf1, which is known to shape the M2 macrophage phenotype and regulate immune metabolism [55] (Additional file 1: Figure S6).Importantly, in both Egr2 and Igf1, marker enhancer regions carried H3K4me1 in untreated macrophages and showed the strongest enrichment with H3K27ac in M2 as compared to M1 and untreated macrophages [47] (Fig. 4e and Additional file 1: Figure S6c).

Transcription factor binding sites are enriched in enhancer regions
To investigate whether our enhancer sets are enriched for known TFBS, we performed an over-representation analysis of experimentally determined protein DNA binding sites established through ChIP-seq [56,57] (see Methods).The sets of macrophage-specific and non-macrophage-specific enhancers are both enriched for binding sites of general factors (p300, Tbp), as well as a range of TFs with well-established roles in macrophages, such as macrophage lineage-determining factor Spi1 (PU.1) [40,58], Cebpb, required for macrophage polarization [59], and Rela, regulating inflammatory genes [60] (Additional file 7).Interestingly, transcription binding sites (TFBS) for Spi1 overlap 54.1% of macrophage-specific enhancers, but only 38% of non-macrophage-specific enhancers (overlap ratio of 1.4 for macrophage-specific/non-macrophage-specific enhancers).We observed similar and higher overlap ratios for other functionally important TFs in macrophages, including Stat1, Rela, Irf1, Junb, and Cebpb [3,[59][60][61] (Table 2).Similarly, the E1 and E2 stimuli-responsive enhancer sets are enriched for TFBS of known macrophage TFs including Spi1, Cebpb, Rela, and Irf and Stat families [3,62,63] (Table 3, Additional file 10).Interestingly, TFBS of Stat1, Rela and Irf1, involved in M1 polarization [3,64], overlap a higher percentage of E1 enhancers as compared to E2 and macrophage-specific enhancers (Tables 2-3, and Additional files 7 and 10).For instance, Irf1 TFBS overlap 21.7% of macrophage-specific enhancers, 26.7% of E2 but 44.3% of E1 enhancers.In addition, the expression of genes encoding these TFs is higher in M1 as compared to M2 macrophages.Taken together, these results provide an additional layer of support for our regions as functionally important macrophage enhancers and implicate key macrophage TFs in modulating their activity.These findings further reflect that enhancers are selectively activated depending on the transcriptional machinery involved in the cellular response.
not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.

Discussion
In this study we investigated the enhancer landscape in mouse macrophages and its dynamic changes during M1 and M2 polarization.Using CAGE data combined with ChIP-seq, we identified 8,667 active enhancers forming 64,891 regulatory associations with protein-coding gene promoters in mouse macrophages.We highlighted tissue-and stimuli-specificity of both enhancers and their regulatory interactions.The enhancer-gene interactome established here supports a model of additive action of enhancers [17,49], with higher gene expression concomitant with higher numbers of associated enhancers.Moreover, we observed a shift towards stronger enrichment for macrophage-specific biological processes in genes associated to many enhancers.
Cytokine stimulation in macrophages had a striking influence on enhancer activation, which highlights the importance of enhancers in polarization towards M1 and M2 phenotypes.In addition, we inferred potential stimuli-specific marker enhancers.Finally, we find that binding sites of inflammatory TFs are enriched in enhancer regions, proposing a link between the response to stimuli and enhancer transcriptional activation.
We used a combination of two complementary data types, transcriptomic and epigenomic data, to infer more reliable active enhancer regions in mouse macrophages.Instead of a widely used linear proximity-based approach [34,37,44], we employed TAD data for the inference of enhancer-gene associations.Accumulating evidence suggests that linear proximity might not be an accurate predictor of enhancer-gene interactions, as many enhancers regulate distal genes, bypassing the nearest promoter [65,66].At the same time, TADs have emerged as units of chromatin organisation that favour internal DNA contacts [48], and the majority of characterized interactions between enhancers and target promoters occur within the same TAD [48,66,67].
TADs are well conserved across cell types, whereas active enhancers and enhancer-promoter regulatory interactions within TADs vary between tissues and conditions [42,65].Different populations of tissue macrophages were shown to be highly heterogeneous and to possess distinct sets of active enhancers, as defined not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint (which was .http://dx.doi.org/10.1101/163519doi: bioRxiv preprint first posted online Jul.14, 2017; by ChIP-seq profiling of histone modifications [27,28].Kaikkonen et al. used H3K4me2 ChIP-seq to identify a set of enhancers that is established in macrophages de novo in response to the stimulation with TLR4 agonist, highlighting stimuli-specific action of enhancers in macrophages [39].These and other studies have stressed the complexity of enhancer action in macrophages and other cell types, which yet remains to be fully characterized.
In our study, we inferred 8,667 active enhancers in mouse macrophages.Of these, 70% overlap RNA polymerase II ChIP-seq peaks in untreated mouse macrophages [47].Our enhancer regions show significant enrichment for binding sites of histone acetyltransferase p300, an enhancer-associated marker [25], and known inflammatory TFs.Hence, the regions identified here show a range of known enhancer properties, generally supporting our approach.Most of the active enhancers show macrophage-specific eRNA expression, in line with known tissue-specificity of enhancers [22,41,42].
Comparing our enhancer set to previous studies of mouse enhancers, we found for example that 39.8% of our enhancers overlap a set of cis-regulatory elements from 19 non-macrophage mouse tissues identified by Shen et al. [41].In another recent study, Schoenfelder et al. employed a Capture Hi-C approach to identify enhancers and their target promoters in mouse fetal liver cells and embryonic stem cells [32].These data support only 24.8% of our 64,891 E-P pairs, highlighting again tissue-specificity of enhancers and enhancer-promoter interactions.
Recent reports suggested that genes regulated by multiple enhancers were higher expressed than those regulated by a single enhancer, proposing that enhancers might contribute additively to the expression of their target genes [17,49].In support of this, we observed a steady increase of gene expression concomitant with increasing numbers of associated enhancers, with the genes not associated to any enhancers showing the lowest overall expression.A study of 12 mouse tissues has reported the enrichment for tissue-specific functions in genes associated to enhancers that transcribe eRNAs as compared to genes associated to non-transcribed enhancers [68].Jin et al. recently showed that genes that did not interact with distal enhancers were enriched for housekeeping genes and also suggested that cell-specific genes were extensively controlled by cis-regulators [69].We showed in this study that genes associated to many enhancers were more enriched for macrophagerelated functions as compared to genes associated to only few or no enhancers.This finding might reflect a more fundamental principle of genome organisation and evolution, such as the importance of multiple enhancers for fine-tuned and redundant control of cell specialization and cell-specific responses.We focused on enhancers and genes that responded to the polarization stimuli with increased expression in order to study transcriptional changes during macrophage activation.Notably, many stimuli-responsive genes were associated to stimuli-responsive enhancers, highlighting the importance of enhancer regulation in macrophage polarization.As expected for such a cell-type-specific process as macrophage polarization, most enhancers showed macrophage-specific eRNA expression, and genes were enriched for macrophage-specific functions.As an important example, we assessed 20 M1 and 26 M2 marker genes to demonstrate enhancer regulation for known macrophage polarization markers [1-3, 6, 51].Three M1 markers (Ccl20, Fpr2, Ido1) and four M2 markers (Chi3l3, Chi3l4, Alox12e, Chia) were not expressed in our data.For a total of 16 M1 or M2 macrophage marker genes, we identified associated enhancers.Moreover, for 11 of these marker genes, we found enhancers that might regulate these genes specifically in M1 or M2 stimulation (Table 1).Hence, these enhancers present new potential markers for a particular macrophage activation status.Seven additional marker not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.
Moreover, we observed a particular genomic distribution of potential marker enhancers associated to Egr2 and Igf1 M2 marker genes, which suggested that these regulatory DNA regions might represent stretch enhancers.
Parker et al., in a recent study, investigated stretch enhancers in human cells and proposed that such extended regions could serve as molecular runways to attract tissue-specific TFs and focus their activity [54].Similarly to Parker et al., potential stretch enhancer regions identified here were associated to cell-type specific genes and were demarcated by broad H3K27ac signals, specifically higher enriched in M2 as compared to M1 and untreated macrophages (Fig. 4e, Additional file 1: Figure S6c).Therefore, we propose that stretch enhancers might be involved in the regulation of macrophage polarization.However, further studies are required to investigate this phenomenon in more detail.
Our approach inferred M1-and M2-responsive enhancers that were strongly enriched for TFBS of known inflammatory TFs.These results are in line with previous reports in mouse macrophages.For example, Spi1 (PU.1) has been extensively studied as a crucial TF involved in macrophage differentiation and transcriptional regulation [40].Moreover, Spi1 was deemed a pioneering or lineage-determining TF in macrophages, which defines enhancer regions and occupies many enhancers in macrophages [27,34,40,58].Furthermore, Heinz et al. suggested that collaborative action of Spi1 with Cebpb was required for the deposition of enhancerassociated chromatin marks [58].Ghisletti et al. reported enrichment for NF-kB (Rel) and Irf TFs in enhancers induced by LPS in mouse macrophages [40].Likewise, transcribed enhancers induced by LPS and IFNγ stimulation showed enrichment for NF-kB/Rel, Irf, and Stat1 binding motifs [34].In addition, we previously showed that TFs including Rela and Irf1 drive expression of protein-coding and lncRNA genes during macrophage polarization [9].Taken together, our results link enhancer activation to the transcriptional program induced by M1 and M2 stimuli.

Conclusions
In this study, we have established a genome-wide catalogue of enhancers and enhancer-promoter regulatory interactions in mouse macrophages.We identified 8,667 active enhancers forming 64,891 regulatory associations with protein-coding gene promoters and highlight tissue-and stimuli-specificity.We observed increased gene expression concomitant with increasing numbers of associated enhancers, which is in support of a model of additive enhancer action.In addition, we noticed a shift towards stronger enrichment for macrophage-specific biological processes in genes associated to many enhancers.We found that cytokine stimulation in macrophages had a striking influence on enhancer activation and we used our enhancer-gene interactome to infer potential stimuli-specific marker enhancers.Our study represents the most comprehensive analysis of transcribed enhancer activities in mouse macrophages to date and our results extend current not peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint (which was .http://dx.doi.org/10.1101/163519doi: bioRxiv preprint first posted online Jul.14, 2017; knowledge of transcriptional regulation and highlight the importance of enhancers in macrophages in general and during polarization in particular.

Fig. 1
Fig. 1 Macrophage enhancer-gene interactome.a Pipeline for identification of enhancers and enhancer-promoter associations.b Number and expression of genes associated to different number of enhancers.Dashed line shows median expression of genes not associated to any enhancer.c KEGG pathway maps significantly enriched for genes associated to a single enhancer, FDR < 0.05.d Top 15 KEGG pathway maps with the lowest FDR enriched for genes associated to more than 8 enhancers.In c and d next to the bars are the numbers of genes in the KEGG pathway covered by our gene list; dashed lines indicate FDR = 0.05.

Fig. 2 16 .d
Fig. 2 Macrophage-specific enhancer and gene expression.a Percentage of genes associated to different number of enhancers.b Expression of 4,739 macrophage-specific enhancer eRNAs and 1,481 associated genes.c Expression of 3,928 non-macrophage-specific enhancer eRNAs and 1,207 associated genes.In b and c expression is shown in 184 macrophage and 774 non-macrophage samples, asterisks denote Wilcoxon rank sum test p-value < 2.2*10 -16 .d Top 15 KEGG pathway maps significantly enriched for genes associated exclusively to macrophage-specific enhancers.e KEGG pathway maps enriched for genes associated exclusively to non-macrophage-specific enhancers with FDR < 0.05.In d and e next to the bars are the numbers of genes in the KEGG pathway covered by our gene list; dashed lines indicate FDR = 0.05.

Fig. 3 5 .e
Fig. 3 Stimuli-responsive genes and enhancers.a Time-course data used in this study.b Enhancer and gene sets.E1 and E2: M1-and M2-responsive enhancers regulating M1-and M2-responsive genes (G1 and G2), respectively; E1' and E2': M1-and M2-responsive enhancers regulating non-stimuli-responsive genes; G1' and G2': M1-and M2-responsive genes not regulated by stimuli-responsive enhancers.Black arrows denote regulatory associations between stimuli-responsive enhancers and genes.c GO biological process terms enriched for G1 and G2 genes (all terms with FDR < 0.05 for G1; six terms with the lowest FDR for G2 are shown); dashed lines indicate FDR = 0.05.d Expression of stimuli-responsive enhancer eRNAs (upper panel) and genes (lower panel) unique to M1 and M2.Statistical significance was determined by Wilcoxon signed-rank test, asterisks indicate p-value < 10 -5 .e Correlation of time-course expression of M1-responsive (upper panel) and M2-responsive (lower panel) enhancers and genes.Vertical dashed lines show median values.

Fig. 4
Fig. 4 Examples of M1 and M2 marker genes and enhancers.a Expression of M1 marker genes Cxcl9, Cxcl10 and Cxcl11.b eRNA expression of three potential marker enhancers that co-regulate Cxcl9, Cxcl10 and Cxcl11.c Genomic region of a TAD containing Cxcl9, Cxcl10, Cxcl11, and associated enhancers.Black links connect the marker genes with the three potential marker enhancers.Grey links denote other enhancer-gene interactions that we identified in macrophages.d Expression of M2 marker gene Egr2 and two of M2 marker enhancers associated to Egr2. e Genomic region of a TAD containing Egr2 and associated enhancers.Black links connect Egr2 with the nine M2 marker enhancers.Grey links denote other enhancer-gene interactions that we identified in macrophages.In a, b and d data were averaged over replicates and log-transformed.Error bars are the SEM.In c and e genes are split into two tracks based on the strand, wide orange marks denote gene promoters; histone mark tracks show ChIP-seq peaks with the height of -10*log10(p-value) (data from [47]).

Table 1
M1 and M2 macrophage polarization markers.peer-reviewed) is the author/funder.All rights reserved.No reuse allowed without permission. not

Table 2
TFs regulating more macrophage-specific than non-macrophage-specific enhancers.

Table 3
TFs with binding sites enriched in both E1 and E2 enhancer sets.Columns show TF name, average expression in M1 and M2 macrophages (in TPM, see Methods), percentage of enhancers overlapping corresponding binding sites, and a M1/M2 percentage ratio.