The Lineage-Specific Transcription Factor PU.1 Prevents Polycomb-Mediated Heterochromatin Formation at Macrophage-Specific Genes

Abstract

Lineage-specific transcription factors (TFs) are important determinants of cellular identity, but their exact mode of action has remained unclear. Here we show using a macrophage differentiation system that the lineage-specific TF PU.1 keeps macrophage-specific genes accessible during differentiation by preventing Polycomb repressive complex 2 (PRC2) binding to transcriptional regulatory elements. We demonstrate that the distal enhancer of a gene becomes bound by PRC2 as cells differentiate in the absence of PU.1 binding and that the gene is wrapped into heterochromatin, which is characterized by increased nucleosome occupancy and H3K27 trimethylation. This renders the gene inaccessible to the transcriptional machinery and prevents induction of the gene in response to an external signal in mature cells. In contrast, if PU.1 is bound at the transcriptional regulatory region of a gene during differentiation, PRC2 is not recruited, nucleosome occupancy is kept low, and the gene can be induced in mature macrophages. Similar results were obtained at the enhancers of other macrophage-specific genes that fail to bind PU.1 as an estrogen receptor fusion (PUER) in this system. These results show that one role of PU.1 is to exclude PRC2 and to prevent heterochromatin formation at macrophage-specific genes.

INTRODUCTION

We previously analyzed chromatin architecture at three inducible genes (IL12B, IL1A, and IFNB1) in mouse macrophages and found that the enhancers of these genes are associated with intermediate levels of nucleosome occupancy ranging from 40% to 70% in primary macrophages derived from mouse bone marrow-derived macrophages (BMDMs) (1). Furthermore, we showed that, upon induction by bacterial lipopolysaccharides (LPS), 2 to 3 nucleosomes were removed from each enhancer. This clearance of enhancer nucleosomes correlated with binding of signal-induced transcription factors (TFs) to the enhancers and recruitment of the transcriptional machinery to both the distal enhancers and promoters of the genes, suggesting that nucleosome clearance from enhancers is required for binding of TFs. At two of the genes (IL12B and IL1A), we found no significant binding of sequence-specific TFs to sites close to the transcriptional start sites (TSS), indicating that regulation of transcription by sequence-specific factors occurs mostly from distal sites, and we hypothesized that looping brings the enhancers close to the promoters and allows transcription to ensue. The distal enhancers of these genes are bound by the macrophage lineage-specific TFs PU.1 and CCAAT/enhancer-binding protein beta (C/EBPβ) in macrophages before LPS induction, and it has been suggested that binding of lineage-specific TFs facilitates subsequent binding of signal-induced TFs and transcription of the associated genes (2, 3). Lineage-specific TFs have also been termed master regulators or pioneer TFs, and it has been suggested that they can access their sites in chromatin and keep enhancers accessible (for a review, see reference 4). To determine whether PU.1 plays a role in keeping macrophage-specific enhancers minimally occupied by nucleosomes during macrophage differentiation and accessible to the transcriptional machinery at a later stage, we used a previously developed macrophage differentiation system (5). PU.1 deletion leads to embryonic lethality and severe defects in hematopoiesis, but a hematopoietic progenitor cell line derived from the fetal liver of a PU.1^−/− mouse can be propagated by growth in the presence of interleukin-3 (IL-3) (6). More importantly, reintroduction of PU.1 as an estrogen receptor fusion (PUER) allows differentiation of these progenitors into macrophage-like cells and has been used to study the role of PU.1 during macrophage differentiation. One recent study determined the effects of PU.1 binding on nucleosome occupancy in this PUER-expressing cell line and showed that, on average, PU.1 sites in the genome were more highly occupied by nucleosomes in the absence of PU.1 (3). Another study that used short hairpin RNA (shRNA)-mediated knockdown of PU.1 in bone marrow-derived macrophages came to a similar conclusion, indicating that PU.1 keeps regulatory regions depleted of nucleosomes (7).

Transcriptionally silent genes are often associated with specific repressive histone modifications, including H3K27me3, a mark placed by Polycomb repressive complex 2 (PRC2) (8–10). PRC2 binding and H3K27me3 are hallmarks of facultative heterochromatin (for a review, see reference 11), and these markers are enriched at CpG islands in promoters (12–14), but H3K27me3 has also been found at enhancers (15–17). Polycomb genes were originally identified in Drosophila through mutations that led to developmental defects (18) and were later shown to be involved in repression of genes during the development of mammals (for recent reviews, see references 19 and 20). An involvement of PRC2 in early development was demonstrated by its role in embryonic stem cells (ESCs) (21, 22), where it was found that genes associated with both “repressive” (H3K27me3) and “active” (H3K4me1) histone modifications are poised for activation (23). What regulates the balance between these apparently opposing histone modifications and what mediates PRC2 recruitment in mammals have remained unclear.

Using a quantitative nucleosome occupancy assay, we have analyzed the role of PU.1 in keeping the enhancers of IL1A and IL12B accessible during macrophage differentiation by determining the changes in nucleosome occupancy at these sites when PUER was reexpressed in PU.1^−/− hematopoietic progenitors or when primary macrophages (BMDMs) were differentiated in the presence of shRNAs targeting PU.1. Significantly, when PUER-expressing cells were differentiated into macrophage-like cells by growth in the presence of tamoxifen, we found that IL1A, but not IL12B, was capable of induction by LPS. Furthermore, we showed that nucleosome occupancy at PU.1 sites in the IL1A enhancer was high in PU.1^−/− cells and decreased as PUER bound to these sites. In contrast, we showed that the IL12B enhancer was less occupied by nucleosomes in PU.1^−/− cells than in mature macrophages but that, nevertheless, PUER was unable to bind when cells were grown for prolonged times even in the presence of tamoxifen. Instead, we demonstrated that differentiation of these cells into macrophage-like cells by growth in the presence of tamoxifen led to heterochromatin formation at the IL12B gene. We found that nucleosome occupancy at the IL12B enhancer increased and that the whole IL12B locus became associated with PRC2 and H3K27me3. Analysis of existing genome-wide data indicated that other enhancers that cannot bind PUER in this system are often associated with heterochromatin markers (i.e., H3K27me3) in other cell types, and we confirmed that a selection of these enhancers is trimethylated on H3K27 in the absence of PUER binding. Our results indicate that PU.1 binding to enhancers prevents heterochromatin formation at macrophage-specific genes and keeps these regions accessible to TF binding and recruitment of the transcriptional machinery in macrophages.

MATERIALS AND METHODS

Primary cell isolation, cell lines, and growth conditions.

BMDMs and splenic B cells were isolated from 8-week-old female C57BL/6 mice (NCI) with oversight from the Michigan State University IACUC, and BMDMs were grown and induced with LPS as described previously (1) for the times indicated in the figures. The PU.1^−/− and PUER-expressing cell lines were obtained from Peter Laslo and grown as described previously (6). In brief, PU.1^−/− cells or PUER-expressing cells were cultured in Iscove's modified Dulbecco's medium (IMDM) without phenol red (Gibco) and with 10% fetal bovine serum (FBS), 2 mM l-glutamine, 50 μM β-mercaptoethanol, 1× penicillin-streptomycin, and 5 ng/ml recombinant mouse IL-3 (Life Technologies). Where indicated, PUER-expressing cells were resuspended in complete medium with 100 nM 4-hydroxytamoxifen (4-OHT) (Sigma).

shRNA-mediated knockdown of SFPI1 in mouse bone marrow cells.

Lentiviral particles containing shRNAs targeting SFPI1 that had been prevalidated by the Broad Consortium (TRC Collection Mission shRNA library; Sigma) or control shRNAs targeting firefly luciferase were produced in HEK293T cells. Briefly, HEK293T cells were seeded at a density of 1 × 10⁷ cells per 150-mm-diameter plate and grown for 24 h in IMDM supplemented with 10% FBS, 50 μM β-mercaptoethanol, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, and 1× penicillin-streptomycin. A calcium phosphate-DNA suspension was prepared for each 150-mm-diameter plate by dropwise addition of 2 ml HEPES-buffered saline (50 mM HEPES, 280 mM NaCl, 10 mM KCl, 1.5 mM Na₂HPO₄, 12 mM d-glucose, pH 7.05) to 2 ml of a solution containing 0.3 M CaCl₂, 25 μg of target shRNA, 15 μg of packaging vector (psPAX2; Addgene), and 5 μg of envelope vector (pMD2.G; Addgene), and the mixture was allowed to stand for 3 to 4 min. After replacement of the IMDM, the calcium phosphate-DNA suspension was added dropwise to HEK293T cells with gentle mixing and cells were incubated at 37°C for 16 h. Medium was replaced again, cells were grown for 24 h, and virus-containing supernatant was collected. Concentrated virus supernatant was either used directly or frozen at −80°C and stored for later use. For lentiviral transductions, bone marrow cells from the femur and tibia from 8-week-old C57BL/6 female mice were collected as described previously and grown for 48 h in BMDM medium containing L929 cell supernatant as a source of macrophage colony-stimulating factor (M-CSF) (1). Cells were infected with lentivirus in the presence of 8 μg/ml Polybrene (Sigma) and incubated for 4 h at 37°C. After replacement of the medium, cells were grown for 48 h and transduced cells were selected by growth in the presence of 5 μg/ml puromycin for 5 days. Cells were harvested for various experiments as described previously (1).

Chromatin isolation and Western blotting.

Chromatin isolation was performed essentially as described previously (24). Briefly, 2 × 10⁶ cells that had been transduced with lentivirus bearing shRNAs targeting SFPI1 or untreated control cells were resuspended in 400 μl extraction buffer which contained 0.2% NP-40 but no sodium butyrate. The final chromatin-bound fraction was obtained by extraction with 160 μl high-salt solubilization buffer, and DNA-bound proteins were obtained by trichloroacetic acid (TCA) precipitation. The resulting pellet was washed with acetone and then resuspended in 50 μl of lithium dodecyl sulfate (LDS) sample buffer (Novex; Life Technologies) containing 2 μl 1 M dithiothreitol (DTT), and samples were heated to 75°C for 10 min. SDS-PAGE was performed on a 4% to 12% Bis-Tris Plus gel (Novex; Life Technologies). Western analysis was performed after protein transfer onto a nitrocellulose membrane and quantification of total protein levels by Ponceau red staining. An antibody detecting PU.1 (sc-352X; Santa Cruz) was used to determine knockdown efficiency. Chemiluminescent signal levels after incubation with appropriate secondary antibodies were quantified on a ChemiDoc MP Imager (Bio-Rad).

mRNA determination.

RNA was extracted and cDNA prepared and analyzed by quantitative reverse transcription-PCR (qRT-PCR) with specific primer pairs as described before (1). Specific primer pairs were designed to quantify mRNA levels in exon 4 for IL12B, exon 7 for IL1A, exon 18 for colony stimulating factor 1 receptor (CSFR1), exon 15 for epidermal growth factor (EGF)-like module receptor 1 (EMR1), exon 1 for Toll-like receptor 4 (TLR4), exon 12 for KIT, exon 2 for SFPI1, and exon 1 for C/EBPβ. Measurements were normalized to ribosomal protein L4 (RPL4) mRNA levels. Primer pair sequences can be provided upon request.

ChIP.

Chromatin immunoprecipitation (ChIP) experiments were performed as described previously (1) except that for H3K27me3 and Suz12 we diluted isolated chromatin in low-salt ChIP buffer (20 mM Tris-HCl, 200 mM NaCl, 0.5% Triton X-100, 2 mM EDTA, Halt protease inhibitor cocktail without EDTA [Thermo Scientific], pH 8) and incubated the reaction mixtures with 0.5 μg anti-H3K27me3 (C36B11; Cell Signaling) and 0.6 μg anti-Suz12 (D39F6; Cell Signaling), respectively. For other ChIP experiments, we used 3 μg of anti-Med1 (A300-793A; Bethyl Laboratories). The locations corresponding to LPS-inducible enhancers identified by Ghisletti et al. (2) are 6.8 kb upstream of CCDC64B, 7 kb downstream of IL20, 9.8 kb upstream of IL27, 11.8 kb upstream of STFA3, and 63 kb upstream of ASB5. Where indicated, data are displayed as severalfold increases compared to an intergenic region located between the IL1A and IL1B genes. Primer pairs for these and all other amplicons used can be given upon request. For experiments where we compared PU.1 binding in PU.1^−/− cells to binding in cells expressing PUER or to BMDMs, we could not normalize our data in that way, because background levels of PU.1 were also significantly reduced at nonspecific regions in the genome. We therefore display our data as percentages of immunoprecipitation (IP), but we adjusted these values for various levels of input. This allowed us to adjust for slight variations in the total (input) DNA in each ChIP reaction mixture, which we found to affect the percentage of IP obtained in each reaction. While this adjustment did not affect the overall result, it considerably reduced the standard errors of the means (SEM) of the results from our biological replicates. Briefly, for each replicate experiment, we compared the values for total (input) DNA obtained for each cell type to total (input) DNA values obtained for BMDMs (grown in the absence of LPS) at each genomic location that we measured and calculated an average deviation of each input from the levels determined with the BMDMs. Percent IP values obtained with different antibodies were then adjusted by this average deviation.

Quantitative nucleosome occupancy assay and qRT-PCR.

The quantitative MNase assay was performed and DNA was quantified by qRT-PCR as described before (1, 25). Primer pair sequences can be provided upon request. The MNase assay was performed at least twice for each cell type, and an in-depth analysis with overlapping primer pairs covering each region of interest is shown for a representative experiment performed side by side with control cells (BMDMs). Analysis of nucleosome occupancy in PU.1 knockdown cells was performed once after pooling cells obtained from two separate lentiviral transductions with two distinct shRNAs that yielded significant levels of PU.1 mRNA knockdown in replicate experiments. For bar graphs, the data are displayed using the IGB genome browser, and overlays and difference maps were generated by the program.

Data analysis.

Bioinformatic data analysis was performed using HOMER, and Venn diagrams were created with mergePeaks (3). The data sets used for analysis were GSM538004 (PUER peaks in PUER-expressing cells grown for 2 days in the presence of tamoxifen) (3), GSM487450 and GSM487449 (PU.1 peaks in resting BMDMs and p300 peaks in BMDMs grown for 2 h in the presence of LPS, respectively) (2), GSM946531, GSM946537, and GSM946523 (H3K27me3 peaks in erythroid progenitors, erythrocytes, and megakaryocytes, respectively) (16), GSM721294 and GSM721295 (H3K27me3 peaks in myoblasts and myotubes, respectively) (15), and GSM1412513 (H3K27me3 peaks in 3T3-L1 adipocytes) (17). The overlapping and unique sites identified by mergePeaks can be found in Tables S1 to S3 in the supplemental material. Cluster analysis was performed manually, and a table was created in Excel. For gene ontology analysis of the clusters, the PANTHER 9.0 classification system was used, which included a Bonferroni correction for multiple testing (26).

RESULTS

IL1A but not IL12B can be induced by LPS when PUER cells are differentiated into macrophage-like cells.

To investigate the role of lineage-specific TF PU.1 during macrophage differentiation, we used a previously described inducible PUER cell line derived from PU.1^−/− hematopoietic progenitors (kindly provided by Peter Laslo, University of Leeds) (6). Growth of these cells for prolonged times in the presence of tamoxifen and IL-3 has been shown to facilitate differentiation into macrophage-like cells. As shown in Fig. 1, we confirmed that growth for as little as 1 h in the presence of tamoxifen led to a considerable reduction in expression of the KIT hematopoietic progenitor marker (Fig. 1A). We also confirmed that macrophage-specific markers, including TLR4, EMR1, and CSFR, were upregulated under these conditions (Fig. 1B). However, we noted that while some markers were expressed at levels similar to those found in BMDMs after only 1 day of growth in the presence of tamoxifen (e.g., TLR4), the levels of others (e.g., CSF1R) remained considerably lower even after 7 days. Significantly, we found that IL1A was induced in response to LPS on day 1, whereas IL12B was not induced even when cells had been grown for 7 days in the presence of tamoxifen (Fig. 1C).

FIG 1.

Expression of macrophage-specific genes in PUER-expressing cells grown in the presence of tamoxifen. PUER-expressing cells were grown in the presence of tamoxifen for the times indicated, and mRNA was quantified as described in Materials and Methods. mRNA levels found in BMDMs are shown for comparison. Data were normalized to levels of RPL4 mRNA. (A) mRNA levels of the progenitor marker KIT. Levels found in PUER cells grown in the absence of tamoxifen were set to 100%. (B) mRNA levels of the macrophage-specific genes CSF1R, EMR1, and TLR4. Levels found in BMDMs were set to 100%. d, day. (C) mRNA of IL12B and IL1A before LPS addition or 1.5 h after LPS addition. Levels found in BMDMs grown for 1.5 h in the presence of LPS were set to 100%. Bars show SEM of the results from at least three replicate experiments.

Nucleosome binding at the IL1A and IL12B enhancers in PU.1^−/− hematopoietic progenitors.

To determine whether PU.1 plays a role in keeping the enhancers of IL12B and IL1A accessible to signal-induced TFs, we analyzed nucleosome occupancy in PU.1^−/− hematopoietic progenitors using a previously described quantitative assay (25). Panels A and B of Fig. 2 show that, at the IL1A enhancer, nucleosome occupancy around the predicted PU.1 consensus sites was increased in PU.1^−/− hematopoietic progenitors compared to the levels found in BMDMs (compare magenta to blue bars and lines). The greatest increase in nucleosome occupancy was found at a region that becomes completely cleared of nucleosomes when IL1A is expressed, as we had previously demonstrated (1) (region marked by the blue box in Fig. 2A and B; see also Fig. 4D). These results suggest that PU.1 binding displaces nucleosomes in the IL1A enhancer, in agreement with previous results from genome-wide studies (3). PU.1 is also expressed in other hematopoietic cells, and we had previously shown that low levels of PU.1 are bound to the IL1A enhancer in splenic B cells (1). As shown in Fig. 2C and D, we found that the pattern of nucleosome occupancy at the IL1A enhancer in B cells was similar to that found in PU.1^−/− cells and that the patterns differed at only one location (indicated by the arrow in Fig. 2C). The two predicted PU.1 sites flanking this region have the preferred consensus GAGGAA sequence (27), while the first two bases at other predicted sites in the region that becomes depleted of nucleosomes in the IL1A enhancer differ (i.e., XXGGAA). Our results suggest that low levels of PU.1 found in B cells may allow PU.1 binding to the strong PU.1 consensus sites and displace nucleosomes in their vicinity but leave weaker sites unbound and occupied by nucleosomes. High levels of PU.1 found in BMDMs may lead to additional binding of PU.1 to weaker sites and further depletion of nucleosomes.

FIG 2.

Nucleosome occupancy at the IL1A and IL12B enhancers in PU.1^−/− progenitors, BMDMs, and B cells analyzed using a previously described quantitative assay (25). (A to D) Levels of nucleosome occupancy at the IL1A enhancer in BMDMs (blue), PU.1^−/− progenitors (magenta), and splenic B cells (green) are shown as bar graphs (A and C) and line graphs (B and D). Occupancy was analyzed in at least three independent experiments for each cell type, and bars representing the SEM are included in the line graphs of panels B and D. The blue box in panels A to D indicates the region that was cleared of nucleosomes in LPS-stimulated BMDMs (1), and nucleosome occupancy in this region was significantly different in the PU.1^−/− or B cells from that in BMDMs as indicated by the P values (Student's t test) (B and D). Consensus sites for TFs are indicated underneath the panels. IRF, interferon regulatory factor. (E to H) Nucleosome occupancy at the IL12B enhancer is shown as described for the IL1A enhancer in panels A to D. The statistical significance of the differences found over the whole IL12B enhancer region in various cell types compared to BMDMs is indicated by the P values (Student's t test) (F and H).

FIG 4.

Nucleosome removal upon LPS induction determined in cells transduced with shRNAs targeting PU.1. Nucleosome occupancy was analyzed as described in the Fig. 3 legend in cells transduced with shRNAs against SFPI1 (magenta) or luciferase (blue) 1 h after growth of cells in the presence of LPS. (A and B) The levels of nucleosome occupancy at the IL12B enhancer in PU.1^−/− cells and BMDMs were statistically significantly different as indicated by Student's t test (P value shown). (C and D) The levels of nucleosome occupancy at the IL1A enhancer were not significantly different.

In contrast to our findings at IL1A, we found that the absence of PU.1 had the opposite effect on IL12B. Thus, nucleosome occupancy at the IL12B enhancer was lower in PU.1^−/− hematopoietic progenitor cells than in BMDMs (Fig. 2E and F). The levels of nucleosome occupancy at preferred nucleosomal positions were around 45% to 50% in PU.1^−/− cells compared to 60% to 75% in BMDMs (Fig. 2E and F, magenta compared to blue bars and lines). Lower levels of nucleosome occupancy at the IL12B enhancer were also seen in splenic B cells (Fig. 2G and H). Our results at the IL12B enhancer are in direct contrast to results from genome-wide studies that have analyzed average nucleosome occupancy levels in the presence and absence of PU.1 after alignment of all PU.1 binding sites (3, 7). Such a view may obscure distinct effects at individual sites. Taken together, our results show that the absence of PU.1 has opposite effects on nucleosome occupancy at the IL1A and IL12B enhancers.

Nucleosome binding at the enhancers when PU.1 is knocked down in bone marrow-derived hematopoietic progenitors.

To confirm that the differences in nucleosome occupancy that we observed were a consequence of the lack of PU.1 and not due to other, secondary effects in the PU.1^−/− cell line, we also knocked down PU.1 in primary bone marrow-derived hematopoietic progenitors using lentiviral delivery of shRNAs (Fig. 3). For these experiments, cells isolated from bone marrow were grown for 2 days in culture and then transduced with lentivirus bearing shRNAs against SFPI1, the gene encoding PU.1. Cells were then grown in the presence of M-CSF for 7 more days, which favors differentiation of hematopoietic progenitors into macrophages. We depleted SFPI1 mRNA to about 50% of wild-type levels, an effect similar to that achieved by others in primary macrophages (Fig. 3A) (7). The levels of chromatin-bound PU.1 protein were found to be reduced by 33% in the knockdowns (Fig. 3B). At this level of knockdown, we found strong impairment of IL12B induction 1.5 h after LPS addition (10-fold reduction), while the effects on IL1A were less dramatic (20% to 30% loss) (Fig. 3C). While knockdown of SFPI1 had negligible effects on nucleosome occupancy at the IL1A enhancer that were not statistically significant (P value = 0.8) (Fig. 3D and E), nucleosome occupancy at the IL12B enhancer was decreased to levels similar to those found in PU.1^−/− cells (Fig. 3F and G). Our findings at individual enhancers determined using our quantitative occupancy assay are again in contrast to results from genome-wide studies in BMDMs with PU.1 knockdown, which had suggested that, on average, nucleosome occupancy at PU.1 sites increases in the absence of PU.1 (7).

FIG 3.

Gene expression and nucleosome occupancy analyzed in BMDMs differentiated in the presence of specific shRNAs targeting PU.1 or control shRNAs targeting firefly luciferase. Bone marrow progenitors were transduced with lentivirus bearing specific or control shRNAs and differentiated in the presence of M-CSF as described in Materials and Methods. (A) SFPI1 mRNA was quantified in cells grown with or without LPS for 1.5 h, and data were normalized to mRNA levels of RPL4. Levels of SFPI1 found in BMDMs grown without LPS were set to 100%. (B) PU.1 protein levels in the chromatin-bound fraction of untreated cells (UNT) and cells knocked down for PU.1 (shPU.1) are shown. Cells transformed with one of the specific shRNAs identified as described for panel A were pooled, and the chromatin-bound fraction was isolated as described in Materials and Methods (24). A Western blot probed with an antibody against PU.1 is shown, as well as the total protein in the chromatin-bound fraction detected by Ponceau red staining of the nitrocellulose membrane. (C) mRNA levels of IL12B and IL1A in cells grown for 1.5 h in the presence of LPS were quantified, and the levels found in BMDMs were set to 100%. (D to G) Nucleosome occupancy at the IL1A (D and E) and IL12B (F and G) enhancers in the absence of LPS was analyzed in cells transduced with specific shRNAs against SFPI1 (magenta) and against luciferase (blue). For these experiments, cells transduced with either of the two SFPI1-specific shRNAs identified as described for panel A were pooled and analyzed together. The error bars show the confidence intervals of measurements at each genomic location derived from curve fitting of MNase digestion data as described in reference 25. Student's t tests indicate that the differences between PU.1^−/− cells and BMDMs in occupancy at IL12B are statistically significant, while those corresponding to occupancy at IL1A are not. P values are indicated (E and G). Comparisons of untreated cells and cells treated with shRNAs against luciferase (LUC) at IL12B can be found in Fig. S1 in the supplemental material; the results showed no statistically significant differences.

Knockdown of PU.1 impairs nucleosome removal at the IL12B enhancer upon LPS induction.

We also determined whether the reduced levels of PU.1 present in our partial knockdown affected nucleosome removal at the IL12B enhancer in response to LPS (Fig. 4A and B). We found that nucleosome occupancy was somewhat higher at the IL12B enhancer 1 h after LPS addition when PU.1 was knocked down compared to control cell results (Fig. 4A and B). Student's t test indicated that the increase at IL12B was statistically significant. Nucleosome removal at IL1A was not affected by the PU.1 knockdown (Fig. 4C and D). Taken together, our results demonstrate that our partial PU.1 knockdown affects induction and nucleosome removal at some genes but not others. This is in agreement with previous studies in PU.1-heterozygous mice, which have shown that certain cytokines as well as some macrophage markers are expressed at lower levels when PU.1 levels are reduced (28, 29).

PUER binds and facilitates recruitment of other TFs and the transcriptional machinery to the IL1A but not the IL12B enhancer.

We found that PUER bound to the IL1A enhancer when cells were grown for as little as 6 h in the presence of tamoxifen (Fig. 5A, orange bars). After 4 and 7 days of growth in the presence of tamoxifen, PUER binding to the IL1A enhancer was comparable to the levels of PU.1 found in BMDMs at some locations in the IL1A enhancer (−10.1 kb) but remained somewhat lower at others (−10.3 kb) (Fig. 5A, compare red and maroon to blue bars). PUER mRNA levels in PUER-expressing cells were similar to or even somewhat higher than those in BMDMs (Fig. 6A), suggesting that expression of PUER is not limiting in this system. Significantly, and in contrast to our findings at IL1A, we did not detect any binding of PUER to the IL12B enhancer even when cells were grown for 7 days in the presence of tamoxifen (Fig. 5A).

FIG 5.

TF binding and recruitment of the transcriptional machinery. Data represent PUER, C/EBPβ, and NF-κB binding and recruitment of the mediator to the IL12B and IL1A enhancers and promoters analyzed in PU.1^−/− progenitors (magenta), in PUER cells grown in the absence of tamoxifen (yellow) or for 6 h (orange), 4 days (red), or 7 days (maroon) in the presence of tamoxifen, and in BMDMs (blue), grown in the absence or presence of LPS for 1.5 h. Labels underneath each panel indicate the distance from the TSS of IL12B and IL1A. Control locations in the KIT promoter (KIT prom) and the ORF of the RPL4 gene are also shown. ChIP data are displayed as percent IP (%IP) adjusted for inputs as described in the Materials and Methods. (A) PU.1 and PUER binding. (B) C/EBPβ binding. (C) NF-κB binding. (D) Mediator binding as determined with an antibody against Med1. Experiments were performed at least three times, and error bars (SEM) are shown.

FIG 6.

mRNA levels of SFPI1 and C/EBPβ. mRNA was analyzed in PUER-expressing cells grown for the indicated times in the presence of tamoxifen and in BMDMs, grown for 1.5 h with LPS (plus LPS) (yellow) or without LPS (min LPS) (blue) as described in the Fig. 1 legend. mRNA levels found in BMDMs grown in the absence of LPS were set to 100%. (A) Expression analysis of PUER and PU.1. (B) Expression analysis of C/EBPβ.

As shown in Fig. 5B, we also detected high levels of C/EBPβ binding at the IL1A enhancer when cells were grown in the presence of tamoxifen for as little as 6 h, and the levels further increased on days 4 and 7 (Fig. 5B). Our data indicate that C/EBPβ binding to the IL1A enhancer in PUER-expressing cells was higher than in BMDMs. This result is consistent with the higher C/EBPβ mRNA levels found in PU.1^−/− progenitors and PUER cells than in BMDMs (Fig. 6B). Our results further show that C/EBPβ binding to the IL1A enhancer is dependent on PUER binding, which is in agreement with genome-wide studies that had indicated that C/EBPβ binds cooperatively with PU.1 to certain sites (3). Strikingly, we did not detect any C/EBPβ binding to the IL12B enhancer even after prolonged growth of cells in the presence of tamoxifen (Fig. 5B).

Upon LPS induction, NF-κB bound to the IL1A enhancer but not to the IL12B enhancer in PUER cells that were grown in the presence of tamoxifen and significant levels of NF-κB binding were detected after 7 days (Fig. 5C, maroon bars). Similarly, we detected recruitment of the transcriptional machinery as represented by the Med1 mediator subunit to the IL1A enhancer but not to the IL12B enhancer when cells were stimulated with LPS, and significant levels of Med1 were detected 4 and 7 days after growth of PUER-expressing cells in the presence of tamoxifen (Fig. 5D, red and maroon bars). This also led to recruitment of Med1 to the IL1A promoter consistent with DNA looping, which we have previously suggested may bring the enhancer into the proximity of the promoter under conditions that lead to gene expression (1). Furthermore, we detected intermediate levels of Med1 recruitment to the IL1A enhancer in the absence of LPS when cells were grown for 7 days in the presence of tamoxifen consistent with IL1A mRNA production under these conditions (Fig. 5D, maroon bars and Fig. 1C, hatched bars).

Growth of PUER-expressing cells in the presence of tamoxifen and PUER binding leads to lower nucleosome occupancy at the IL1A enhancer.

Our results shown in Fig. 2A and B demonstrated that the IL1A enhancer is more highly occupied by nucleosomes in PU.1^−/− cells than in BMDMs and suggested that PU.1 binding may be sufficient to displace nucleosomes. To determine if growth of PUER-expressing cells in the presence of tamoxifen reduces nucleosome binding at the IL1A enhancer, we analyzed nucleosome occupancy during a time course experiment. As shown in Fig. 7A, we found that as early as 1 h after addition of tamoxifen, nucleosome occupancy at the IL1A enhancer resembled that found in resting BMDMs (compare light orange to blue bars). Our results and those of others indicate that at this time point, significant PUER binding can be detected at the IL1A enhancer (3) (M. Tagore and M. Floer, unpublished data). Significantly, upon prolonged growth of cells in the presence of tamoxifen, we detected further removal of nucleosomes from the IL1A enhancer in the absence of LPS, and after 6 h, the occupancy at the nucleosomal position in the IL1A enhancer (indicated by the arrow in Fig. 7A) was about 20% lower than in BMDMs. This decrease in nucleosome binding can also be seen in the difference maps of Fig. 7B, which show the difference in nucleosome binding at each time point compared to BMDMs. This view shows significant depletion of nucleosomes after 6 h, 4 days, and 7 days of growth in the presence of tamoxifen even in the absence of LPS (1). Student's t tests confirmed that the differences in the levels seen before PUER expression as well as after prolonged growth in the presence of tamoxifen were statistically significant (P values are indicated in Fig. 7B). Our gene expression analysis had detected some level of IL1A mRNA in the absence of LPS induction at day 7 (Fig. 1C) as well as recruitment of the transcriptional machinery (see maroon bars in Fig. 5D). We conclude that, in PUER-expressing cells grown for prolonged times in the presence of tamoxifen, IL1A expression becomes partially uncoupled from LPS induction.

FIG 7.

Changes in nucleosome occupancy at the IL1A enhancer in PUER cells grown in the presence of tamoxifen. Nucleosome occupancy was analyzed in PU.1^−/− progenitors (magenta), in PUER cells grown in the absence of tamoxifen (yellow) or for 1 h (light orange), 6 h (dark orange), 4 days (red), or 7 days (maroon) in the presence of tamoxifen, or in BMDMs (blue). All cells were grown in the absence of LPS. (A) Occupancy is shown as bar graphs, and levels found in BMDMs are shown for comparison. (B) Difference maps show percentages of differences in occupancy (% Diff. in occ.) (compared to BMDMs) as cells were grown for increasing times in the presence of tamoxifen (P values of Student's t tests are shown). The arrow indicates the nucleosomal position that showed that greatest change in occupancy. The analysis of the time course of PUER cell differentiation by growth in the presence of tamoxifen was performed twice, and data are from one representative experiment.

Growth of PUER-expressing cells in the presence of tamoxifen leads to heterochromatin formation at the IL12B locus.

In contrast to our findings at IL1A, we found that the IL12B enhancer was less occupied by nucleosomes in PU.1^−/− cells and in our partial PU.1 knockdowns (see Fig. 2E and F and 3F and G). When PUER-expressing cells were grown for prolonged times in the presence of tamoxifen, we found that the IL12B enhancer became more highly occupied by nucleosomes (Fig. 8A). Binding at peak nucleosomal positions in the IL12B enhancer (indicated by the arrows in Fig. 8A) increased as early as 6 h after tamoxifen addition and reached levels similar to those found in resting BMDMs after 7 days. P values of Student's t tests of comparisons to BMDMs are shown in Fig. 8A. A summary of the changes at the IL12B enhancer indicating that median nucleosome occupancy increased in the region is shown in the box plot of Fig. 8B. Furthermore, a region that included a cluster of predicted PU.1 consensus sites (indicated by the blue box in Fig. 8A) showed slightly higher nucleosome occupancy in PUER-expressing cells grown in the presence of tamoxifen than in BMDMs (the levels of nucleosome occupancy in BMDMs are indicated by the hatched blue line in each panel of Fig. 8A). We conclude that under these conditions, PU.1 was not bound to the IL12B enhancer, leading to the formation of nucleosomes at the PU.1 consensus sites.

FIG 8.

Nucleosome occupancy, Polycomb binding, and H3K27me3 at IL12B. (A) Nucleosome occupancy at the IL12B enhancer was analyzed in PU.1^−/− progenitors (magenta), in PUER cells grown for 6 h (orange), 4 days (red), or 7 days (maroon) in the presence of tamoxifen, or in BMDMs (blue). Cells were grown in the absence of LPS. The blue box highlights the region that was more highly occupied in PUER cells grown for prolonged times in the presence of tamoxifen than in BMDMs, and the hatched blue lines show the levels of nucleosome occupancy found in this region in BMDMs. Arrows indicate the peak positions of the three nucleosomes that are removed when BMDMs are stimulated with LPS (1). The time course experiment was performed twice, and results of a representative experiment are shown. Differences in occupancy at each time point compared to BMDMs were analyzed for statistical significance by Student's t tests (P values are shown). (B) Results shown in panel A are summarized in a box plot. (C) H3K27me3 was analyzed by ChIP in cells grown as described for panel A. Colors are the same those described for panel A, and the results from cells grown for 4 and 7 days in the presence of tamoxifen are shown. (D) Nucleosome occupancy at the promoters of KIT, VPREB2, and IL12B in different cell types was measured as described for panel A. The average changes at the three nucleosomal peaks in the IL12B enhancer are shown for comparison. Bars indicate SEM of the results from two independent experiments. (E) Suz12 binding determined by ChIP in cells as described for panel C. The ChIP experiments described for panels C and E were performed at least three times, and error bars (SEM) are shown. Student's t tests showed that the differences in H3K27me3 and Suz12 binding at IL12B between PUER-expressing cells grown for 4 days or 7 days in the presence of tamoxifen and BMDMs are statistically significant (P < 0.01) but not the differences between PU.1^−/− cells and BMDMs.

To determine whether the nucleosomes formed at IL12B in the absence of PU.1 were qualitatively different from those formed in the presence of PU.1, we analyzed trimethylation at H3K27 (Fig. 8C). H3K27me3 is a modification often found at genes that are transcriptionally silent and is a mark for facultative heterochromatin. Indeed, we detected H3K27me3 at the IL12B enhancer when PUER-expressing cells were grown for 4 and 7 days in the presence of tamoxifen. H3K27me3 was also detected at the IL12B promoter, at the open reading frame (ORF), and in the intervening region (as indicated by a location 7 kb upstream of the promoter), suggesting that the whole IL12B locus becomes modified under these conditions. Levels of H3K27me3 found at the IL12B gene locus on day 7 were similar to those found at the promoter of a gene that is repressed as macrophages differentiate (e.g., the progenitor marker KIT) or at a gene that is expressed only in another hematopoietic lineage (e.g., VPREB2, expressed only in the B cell lineage) (see Fig. 8C) (30). These results suggest that in the absence of PU.1 binding to the IL12B enhancer, the IL12B locus becomes associated with heterochromatin markers during differentiation, a fate normally reserved for genes that are repressed in macrophages.

Our results further showed that the appearance of H3K27me3 correlated with an increase in nucleosome occupancy at IL12B and other sites (Fig. 8D). Thus, we that found the KIT promoter was occupied by nucleosomes in PU.1^−/− progenitors in around 45% of the population and that occupancy increased to about 60% when PUER-expressing cells were grown for 7 days in the presence of tamoxifen. Occupancy at the KIT promoter was around 90% in BMDMs, consistent with the high occupancy of this region found in a previous study in a myeloid cell line where KIT is not expressed (31). Similarly, we found that the VPREB2 promoter was occupied in around 40% of the population in PU.1^−/− progenitors and that occupancy increased to around 65% in PUER-expressing cells grown for 7 days in the presence of tamoxifen. The IL12B promoter was already highly occupied in PU.1^−/− cells, and occupancy did not change significantly when cells were grown in the presence of tamoxifen. Taken together, our results indicate that, during differentiation, regions that are silenced in macrophages become associated with highly occupied nucleosomes that are trimethylated on H3K27. We conclude that high nucleosome occupancy and the presence of H3K27me3 are hallmarks of facultative heterochromatin.

Nucleosomes at the IL12B locus are trimethylated on H3K27 by recruited PRC2.

Trimethylation of H3K27 is mediated by Polycomb repressive complex 2 (PRC2) (8–10, 32). As shown in Fig. 8E, we detected recruitment of the PRC2 subunit Suz12 to the IL12B enhancer, promoter, and ORF when cells were grown for 4 and 7 days in the presence of tamoxifen. Suz12 levels found at IL12B on day 7 were similar to those found at the KIT and VPREB2 promoters and were negligible at IL1A (Fig. 8E, maroon bars). Taken together, our data indicate that in the absence of PU.1 binding to the IL12B enhancer, PRC2 is recruited when cells are differentiated into macrophage-like cells, which leads to H3K27 trimethylation. Our data further suggest that Polycomb binding and H3K27 trimethylation spread over the whole region to encompass the IL12B promoter, the intervening region, and the IL12B ORF (see Fig. 8C and E).

Heterochromatin is formed at other LPS-inducible enhancers.

To determine whether other LPS-inducible enhancers fail to bind PUER, we examined existing genome-wide data. Heinz et al. had previously identified PUER-bound sites when PUER-expressing cells had been grown for 2 days in the presence of tamoxifen (3). We used the HOMER software developed by those investigators to determine the overlap of PUER-bound sites (Fig. 9A, red) with PU.1-bound sites identified in resting BMDMs by Ghisletti et al. (2) (blue). We found that while a large number of sites overlapped, 20,345 sites bound in BMDMs did not bind PUER; this set included the IL12B enhancer. To determine whether the sites not bound by PUER included LPS-inducible enhancers that function in normal macrophages, we compared this data set with the LPS-inducible p300 peaks that had been identified in BMDMs by Ghisletti et al. (2). Those investigators identified 2,350 LPS-induced p300 peaks and concluded that inducible p300 binding was a good marker for LPS-responsive enhancers. We found that, of those 2,350 inducible peaks (Fig. 9B, dark red), 942 were bound by PU.1 in BMDMs before LPS induction (light red); Ghisletti et al. had found a slightly higher number of inducible p300 peaks to coincide with PU.1 binding 2 h after LPS induction (2). Of these 942 PU.1-bound, LPS-inducible p300 peaks, 381 failed to bind PUER when it was reexpressed in PU.1^−/− progenitors (Fig. 9B, blue). Thus, we identified a large (40%) fraction of putative PU.1-dependent, LPS-inducible enhancers that are active in BMDMs but did not bind PUER when reexpressed in PU.1^−/− cells and that resembled the IL12B enhancer.

FIG 9.

H3K27me3 at LPS-inducible enhancers in the absence of PU.1 binding. Venn diagrams comparing published data sets were created with HOMER (3) as described in Materials and Methods. (A) Overlap of PUER-bound sites identified in PUER-expressing cells grown for 2 days in the presence of tamoxifen (red) (3) with PU.1-bound sites identified in resting BMDMs (blue) (2). (B) A Venn diagram shows that a subset of LPS-inducible p300 peaks identified in BMDMs (dark red) (2) was bound by PU.1 in resting BMDMs (light red) (2), but comparison of these sites with the PUER-bound data set from panel A showed that a fraction did not bind PUER (blue). (C) A Venn diagram shows that a fraction of the sites that did not bind PUER that were identified as described for panel B (blue) was associated with H3K27me3 peaks (yellow) identified in the 6 cell types indicated in panel D (15–17). (D) Sites identified as described for panel C were divided into 4 clusters as described in Materials and Methods. The scale from yellow to dark green indicates the fraction of H3K27me3 peaks found in each category and cell type indicated in the figure. Gene ontology (GO) sets that were enriched in clusters I and II or in clusters III and IV were identified using PANTHER (26). The statistical significance of the enrichment is demonstrated by the P values shown. (E) H3K27me3 was analyzed by ChIP as described in the Fig. 8C legend. (F) PU.1 and PUER binding in cells was analyzed by ChIP as described for panel E. The experiments described for panels E and F were performed at least twice, and error bars (SEM) are shown. Student's t tests showed that the differences at all enhancers except that of IL1A were statistically significant between PUER-expressing cells grown for 4 days or 7 days in the presence of tamoxifen and BMDMs (P < 0.01) but not between PU.1^−/− cells and BMDMs.

We then investigated whether any of these macrophage-specific enhancers that fail to bind PUER in PU.1^−/− cells were associated with markers of facultative heterochromatin in other cell types. We compared this set of enhancers to six data sets from H3K27me3 ChIP sequencing (ChIP-seq) experiments generated by the Mouse Encyclopedia of DNA Elements (ENCODE) Consortium in various cell types, including erythroid progenitors, erythrocytes, and megakaryocytes (15), myoblasts and myotubes (16), and 3T3-L1 adipocytes (17) (Fig. 9C). Of the 381 PU.1-dependent LPS-inducible macrophage enhancers (blue), 210 (55%) showed H3K27 trimethylation in at least one other cell type (yellow). We then performed a clustering analysis of these 381 enhancers according to their H3K27 trimethylation patterns in different cell types (Fig. 9D). We divided the 381 peaks into 4 clusters: sites that were trimethylated on H3K27 in 5 to 6 cell types (cluster I), in 2 to 4 cell types (cluster II), or in any 1 cell type (cluster III) or that were not trimethylated (cluster IV). The color intensity from yellow to dark green indicates the fraction of sites that were trimethylated in each cell type and cluster. Gene ontology analysis using PANTHER (26) revealed that clusters I and II were overrepresented for macrophage activation and immune response genes, while clusters III and IV were overrepresented for genes involved in constitutive cellular processes. These results suggest that a large fraction of enhancers that regulate LPS-induced responses associated with immune function in macrophages become associated with facultative heterochromatin in other cell types. A list of these and all the other sites identified in Fig. 9 can be found in the Tables S1 to S3 in the supplemental material.

H3K27me3 at enhancers that cannot bind PUER in macrophage-like cells.

We confirmed H3K27me3 binding at a subset of the PU.1-dependent, LPS-inducible enhancers identified in Fig. 9C in PUER-expressing cells that were grown for prolonged times in the presence of tamoxifen (Fig. 9E). The figure shows five distal enhancers that are between 6.8 kb and 63 kb upstream or downstream of the TSS of the nearest LPS-inducible gene and that had been assigned to these genes by proximity (2). None of these enhancers bound PUER even after 7 days of growth in the presence of tamoxifen; instead, the enhancers became associated with H3K27me3 (Fig. 9E and F). We also included in this study the distal and proximal enhancers of IFNB1, which we had previously studied (1). IFNB1 was not induced in PUER-expressing cells (Tagore and Floer, data not shown), and both enhancers became associated with H3K27me3 (Fig. 9E). Levels found at the enhancers of IL12B and IL1A are shown for comparison in the figure. Taken together, our results show that facultative heterochromatin is formed at other enhancers in the absence of PU.1 binding and suggest that this is a widespread mechanism that renders macrophage-specific enhancers unresponsive in other cell types.

DISCUSSION

Heterochromatin formation at macrophage-specific enhancers in the absence of PU.1 binding.

Our results show that a lack of PU.1 binding to the IL12B enhancer, when it was reexpressed as a PUER fusion to induce differentiation of PU.1^−/− hematopoietic progenitors, rendered the IL12B gene unresponsive to LPS stimulation in mature macrophages (Fig. 5A and 1C). In the absence of PUER binding, C/EBPβ and signal-induced TFs failed to bind to the enhancer, and the transcriptional machinery was not recruited to the IL12B promoter in response to LPS (Fig. 5C and D). Instead, the IL12B gene locus became wrapped into facultative heterochromatin, as demonstrated by recruitment of PRC2 and trimethylation of H3K27 (Fig. 8). Furthermore, we found that formation of facultative heterochromatin at IL12B and other genes was associated with an increase in nucleosome occupancy (Fig. 8A, B, and D). Similar changes occurred at other macrophage-specific enhancers that failed to bind PUER in this system, and these enhancers also acquired heterochromatin markers (i.e., H3K27me3) during differentiation (Fig. 9E and F). Taken together, our findings suggest that one role of lineage-specific or pioneer TFs such as PU.1 is to prevent Polycomb binding to regulatory regions of genes that become induced only in fully differentiated macrophages. We propose that PU.1 binding marks these genes for future use and prevents their silencing in macrophages. This idea is consistent with the finding that PU.1 is bound to the regulatory regions of most macrophage-specific genes in macrophages even before they are induced (2, 3, 33). Our survey of existing data showed that these enhancers are often associated with H3K27me3 in other cell types (15–17), indicating that facultative heterochromatin is formed at macrophage-specific enhancers in cell types that do not express the associated genes. Our results strongly suggest that heterochromatin formation at the IL12B locus is triggered by events that occur at the distal enhancer of IL12B, and we hypothesize that heterochromatin formation may start from the distal enhancer and spread to the promoter and gene body. Significantly, we did not detect PRC2 binding or trimethylation of H3K27 at the IL12B locus in PU.1^−/− progenitors. Instead, these heterochromatin markers appeared only when cells were differentiated into macrophage-like cells (Fig. 8C and E). This result indicated that the absence of PU.1 alone was not sufficient to induce recruitment of Polycomb but suggests that additional events during differentiation triggered PRC2 binding at macrophage-specific enhancers that failed to bind PU.1. It has been suggested that short CG-rich DNA sequences may mediate Polycomb binding in mammals and that Polycomb binding is counteracted by TF binding to neighboring sites (34). Whether such PRC2-recruiting sequences exist in the IL12B enhancer remains to be determined.

Differences between the IL12B and IL1A enhancers.

In our partial PU.1 knockdown, the levels of nucleosome occupancy at the IL12B enhancer resembled those found in the PU.1^−/− progenitors (compare Fig. 3F and G to Fig. 2E and F), and IL12B induction by LPS was impaired (Fig. 3C). Under these conditions, nucleosomes were less well positioned in the region, indicating that PU.1 binding to the IL12B enhancer during macrophage differentiation may lead to the typical pattern of positioned nucleosomes found in BMDMs. We also found that nucleosome removal at the IL12B enhancer in response to LPS was somewhat affected in the partial PU.1 knockdown (Fig. 4A and B), suggesting that PU.1 may play a role in enhancer clearance by, for example, recruiting nucleosome remodelers to the IL12B enhancer. It remains to be determined what keeps the IL12B enhancer minimally occupied by nucleosomes in the PU.1^−/− progenitors and under the conditions of our partial knockdown, but we speculate that other factors, including other Ets family TFs, may be able to bind to PU.1 sites when PU.1 levels are limiting. In contrast, our results showed that IL1A was not affected by the partial PU.1 knockdown (Fig. 3C and 4C and D) and that PUER binding could be induced at the IL1A enhancer when cells were grown in the presence of tamoxifen (Fig. 5A). We had previously shown that PU.1 binding to the IL12B enhancer was about 2-fold to 3-fold lower than to the IL1A enhancer as determined by ChIP (also shown in Fig. 5A, blue bars) (1). Taken together, these findings may suggest that a higher affinity of PU.1 for sites in the IL1A enhancer may render the gene less sensitive than IL12B to limiting amounts of PU.1, although additional factors may play a role.

Partial uncoupling of IL1A expression from LPS signaling when PUER cells were grown in the presence of tamoxifen for prolonged times.

Our studies revealed that the tight regulation of IL1A expression seen only in response to an inducing signal was lost in PUER-expressing cells grown for prolonged times in the presence of tamoxifen. Thus, we detected increased basal levels of IL1A mRNA in the absence of LPS after 7 days of growth in the presence of tamoxifen (Fig. 1C). Our findings are consistent with partial nucleosome depletion at the IL1A enhancer under these conditions and recruitment of mediator (Fig. 5D and 7). We speculate that high levels of C/EBPβ in PUER cells may play some role in uncoupling IL1A expression from LPS signaling. Prior to LPS induction, C/EBPβ mRNA levels were about 2-fold higher in PUER cells grown in the absence of tamoxifen than in BMDMs, and C/EBPβ mRNA levels further increased when cells were grown for prolonged times in the presence of tamoxifen, resulting in 10-fold-higher levels after 7 days (Fig. 6B). Under these conditions, the levels of C/EBPβ bound at the IL1A enhancer were significantly higher than the levels found in BMDMs (Fig. 5B), and we suggest that this may have facilitated recruitment of the transcriptional machinery and transcription in the absence of signal-induced transcription factors such as NF-κB (see Fig. 5C and D). In contrast, we hypothesize that in normal macrophages, C/EBPβ levels are kept low to create a signal-dependent switch. Our results are in agreement with the notion that the lineage-specific TFs PU.1 and C/EBPβ may play a role in recruitment of the transcriptional machinery. Furthermore, our results suggest that the IL1A enhancer may not function by assembling an enhanceosome found at other innate immune genes such as IFNB1, which has a very precise structure that is formed only in the presence of all the required TFs (35, 36). Instead, our results show that NF-κB is not absolutely required for expression of IL1A and we hypothesize that binding of TFs may occur in a stepwise fashion and that high-enough levels of any one of several TFs may result in transcription. Such a billboard model has been suggested for some enhancers, in which a number of different combinations of factors can be arrayed, with variable spacing and stoichiometry, to affect gene expression (37).

PU.1 and nucleosome binding at PU.1 sites.

Previous studies have suggested that binding of so-called pioneer TFs such as PU.1 to chromatinized DNA may occur, but whether this leads to displacement of nucleosomes or whether PU.1 can bind to its sites on a nucleosomal surface has remained unclear (4). In support of the nucleosome displacement hypothesis, genome-wide studies have shown that, on average, PU.1 sites are less occupied by nucleosomes when PU.1 is bound (3, 7). However, this conclusion was drawn from an analysis performed on average occupancies after alignment of all PU.1 sites. Such an analysis can amplify effects found at individual sites, and we found that the results were less clear when we inspected individual PU.1 sites in these genome-wide data sets. Our approach allows a quantitative assessment of the changes in nucleosome occupancy in response to PU.1 binding at the IL1A and IL12B enhancers and is consistent with the notion that PU.1 binding and nucleosome binding are mutually exclusive. While PU.1 binding did not completely clear sites of nucleosomes, we detected decreases in nucleosome occupancy of about 25% and 20% when PU.1 was bound to the IL1A and IL12B enhancers, respectively (Fig. 2A and B, 7, and 8A). These findings indicate that PU.1 binding may displace nucleosomes either by merely competing with nucleosomes or by recruiting nucleosome remodelers that disassemble nucleosomes. Alternatively, a PU.1-bound nucleosome may have altered properties such as, for example, less protection against digestion by MNase. We note that our previous studies in Saccharomyces cerevisiae have revealed that an altered, partially unwrapped nucleosome is formed by binding of the nucleosome remodeler RSC to a nucleosome at the UASg of the GAL1 and GAL10 genes and that formation of this structure facilitates binding of the activator Gal4 to nucleosomal sites without displacing the nucleosome (38). Whether similar structures exist in mammalian cells remains to be determined.

Conclusion.

In summary, our results show that reexpression of PU.1 in hematopoietic progenitors restored function to certain enhancers of macrophage-specific genes but not others. These findings indicate that lineage-specific TFs may have to be present during the early stages of differentiation and that their expression at a later stage may not be sufficient to reverse changes in chromatin architecture that occur in their absence. We hypothesize that limited accessibility of gene regulatory regions may be the underlying reason for the low reprogramming efficiencies achieved when induced pluripotent stem cells (iPSCs) are generated by ectopic expression of pluripotency factors (for a review, see reference 39). Thus, it is commonly observed that only a fraction of cells can be reprogrammed into iPSCs by expression of various TFs. Furthermore, it has been suggested that iPSCs are functionally different from embryonic stem cells and that they may retain a memory of their previous state (40–42). We hypothesize that recruitment of PRC2 and heterochromatin formation at some gene regulatory regions may create a barrier that prevents binding of ectopically expressed TFs, and we hope that a better understanding of these processes will lead to improved reprogramming protocols in the future.