Identification of factors mediating the developmental regulation of the early acting –3.9 kb chicken lysozyme enhancer element

Abstract

The chicken lysozyme gene –3.9 kb enhancer forms a DNase I hypersensitive site (DHS) early in macrophage differentiation, but not in more primitive multipotent myeloid precursor cells. A nucleosome becomes precisely positioned across the enhancer in parallel with DHS formation. In transfection assays, the 5′-part of the –3.9 kb element has ubiquitous enhancer activity. The 3′-part has no stimulatory activity, but is necessary for enhancer repression in lysozyme non-expressing cells. Recent studies have shown that the chromatin fine structure of this region is affected by inhibition of histone deacetylase activity after Trichostatin A (TSA) treatment, but only in lysozyme non-expressing cells. These results indicated a developmental modification of chromatin structure from a dynamic, but inactive, to a stabilised, possibly hyperacetylated, active state. Here we have identified positively and negatively acting transcription factors binding to the –3.9 kb enhancer and determined their contribution to enhancer activity. Furthermore, we examined the influence of TSA treatment on enhancer activity in macrophage cells and lysozyme non-expressing cells, including multipotent macrophage precursors. Interestingly, TSA treatment was able to restore enhancer activity fully in macrophage precursor cells, but not in non-macrophage lineage cells. These results suggest (i) that the transcription factor complement of multipotent progenitor cells is similar to that of lysozyme-expressing cells and (ii) that developmental regulation of the –3.9 kb enhancer is mediated by the interplay of repressing and activating factors that respond to or initiate changes in the chromatin acetylation state.

INTRODUCTION

Gene loci typically encompass a variety of different interacting cis-regulatory elements. The activity of these elements is subject to developmental regulation. In recent years, it has become evident that the activation of genes during development is mediated by dynamic changes in the higher order chromatin structure. To gain insight into the molecular details and the dynamics of this process, we need to study the factors mediating developmental regulation and chromatin structure of individual model gene loci.

The chicken lysozyme gene is a well-studied marker gene for the myeloid lineage of the haemopoietic system. During macrophage differentiation, transcription increases and reaches its maximal level in activated macrophages (1,2). Lysozyme gene expression is controlled by a set of at least six cis-regulatory elements. Different combinations of these elements are activated during the course of activation of the locus (reviewed in 3). Lysozyme locus activation occurs early in development in myeloblasts and is mediated by two early acting enhancers at –6.1 and –3.9 kb and the promoter. A third enhancer at –2.7 kb becomes active later in development (4). Experiments in transgenic mice demonstrated that the chromatin remodelling abilities of early enhancers and the late enhancer are different. In the absence of promoter sequences, DNase I hypersensitive site (DHS) formation at the early enhancers cannot take place, indicating that these elements have to interact early in macrophage differentiation in order for chromatin remodelling at the early enhancers to occur. In contrast, the late acting –2.7 kb enhancer is capable of forming a DHS on its own (5).

Recent UV-photofootprinting studies indicated that chromatin reorganisation of the lysozyme locus starts even earlier in development than the appearance of DHSs at the –6.1 and –3.9 kb enhancers. UV-photofootprinting is a method that detects changes in DNA structure caused by interaction with proteins (6,7). This approach allowed us to detect the formation of a chromatin fine structure characteristic of lysozyme-expressing macrophages in multipotent myeloid precursor cells. These changes were not detectable by standard chromatin structure analysis methods, like DNase I digestion to map DHSs or micrococcal nuclease (MNase) digestion to map specifically positioned nucleosomes. The formation of this chromatin pattern was independent of the stable binding of end-stage transcription factors and actual gene expression (8). Transcription factor binding in lysozyme-expressing cells also changed DNA structure, but this pattern was superimposed on a basic pattern already seen in precursor cells. Trichostatin A (TSA) experiments, leading to hyperacetylation of chromatin proteins and certain transcription factors, demonstrated that in lysozyme-non-expressing cells the maintenance of this pattern is dynamic, but once transcription factor complexes have stably bound and gene expression has started, it is fixed. Our interpretation of these results is that early in haemopoietic development, tissue-specific chromatin remodelling events take place that may be mediated by the transient action of transcription factors. This, however, would imply that the functional transcription factor complement of multipotent progenitor cells is highly similar to that of mature cells.

In the study presented here we have addressed this issue by identifying the factors mediating developmental regulation of one of the early chicken lysozyme enhancers, the –3.9 kb enhancer. This element is particularly suited to the examination of the influence of chromatin structure on gene expression, since we have previously shown that it is organised in a specifically positioned nucleosome only when active (9). In addition, this enhancer consists of two parts that together mediate differential expression of the element in macrophages and non-macrophage cells. The 5′-part of the enhancer stimulates reporter gene activity in all cell types. The 3′-part has no enhancer activity, but is required to repress gene activity in lysozyme-non-expressing cells. Up until now, very little was known about the transcription factors recognising this element. In the study presented here, we have identified these factors by in vitro DNA–protein interaction and in vivo footprinting and studied their roles in transient transfection assays. We show here that the repressing activity of the 3′-enhancer part is mediated by a member of the Forkhead (Fox) transcription factor family. Interestingly, TSA treatment relieves repression in lysozyme-non-expressing cells, but this is seen only in multipotent progenitor cells.

MATERIALS AND METHODS

Cell culture and reporter gene plasmids

HD11 (10), HD50 MEP and HD37 cells (11) were grown in Iscove’s medium containing 8% foetal calf serum and 2% chicken serum. The reporter gene plasmid pTATA carries the luciferase reporter gene, β-globin splice and poly(A) sequences, as well as polylinker sequences derived from pBLCAT3 (12; A.Hecht, unpublished results). It is driven by a minimal TK promoter. All fragments from the –3.9 enhancer region were cloned blunt in the sense orientation into the unique SalI site located in the polylinker in front of the promoter (9). Point mutants were introduced into transcription factor-binding sites using a Quickchange site-directed mutagenesis kit (Stratagene).

Transient transfection analysis

Aliquots of 2 × 10⁶ HD11, HD50 MEP and HD37 cells were transfected by electroporation with 80 µg reporter plasmid and 5 µg RSVlacZ plasmid, using a pulse of 260 V/1050 µF (HD11, HD37) or 300 V/1050 µF (MEP cells). After 48 h the cells were harvested, protein extracts were prepared and luciferase and β-galactosidase activity was measured, using kits from Promega and Clontech, respectively, and a Berthold LB9501 luminometer. Variations in transfection efficiency were corrected by normalising luciferase values to those of β-galactosidase.

Preparation of nuclear extracts and electrophoretic mobility shift assays (EMSAs)

Nuclear extracts from chicken cell lines (NUN extracts) were prepared exactly as described (13). For EMSAs, various amounts of NUN extracts were mixed with radioactively labelled fragments and 0.5 µg poly(dI·dC) in up to 20 µl of 25 mM HEPES, pH 7.6, 5 mM MgCl₂, 34 mM KCl and incubated at room temperature for 30 min. A one-fifth vol of Ficoll loading buffer was added (15% Ficoll, 2.5× TBE, 0.1% bromophenol blue) and the samples were quickly loaded onto a 6% polyacrylamide gel and run at 6 V/cm. After observing an electrophoretic mobility shift, we consulted the databases (MatInspector) to identify factor-binding sites. Oligonucleotides used are outlined in Figures 4–6. Consensus oligonucleotides were NF1 oligo (Santa Cruz), C/EBP oligo (14) and CP2α oligo (15).

Figure 4.

A Fox family member and a CP2-like protein bind to the AT-rich region downstream of the ‘minimal’ enhancer. EMSAs with various radioactively labelled DNA fragments and 10 µg protein extracts from HD11, MEP or HD37 cells were performed as described in Materials and Methods. Competitions (comp.) were performed using a 100-fold excess of unlabelled DNA fragment. Sequences of the different oligonucleotides used in these experiments are depicted below each figure as well as mutant oligonucleotides for the palindrome (4/Pal.m), Fox (5/Fox) and CP2 (6/CP2) and CP2α consensus sequence (CP2α).

Figure 6.

Mutation analysis of the –3.9 kb enhancer by transient transfection experiments. Various luciferase constructs based on StuP2 plus an internal control were electroporated into HD50 MEP precursor cells (A) or HD11 macrophage cells (B) as described in the legend to Figure 1. Subsequently, the cells were incubated with or without 100 nM TSA for 48 h. Mutant StuP2 constructs carry point mutations for the following binding sites, C/EBP (C/EBPm) and Ets (Etsm), NF1 (NF1m), palindrome (Pal.m), Fox (Foxm) and CP2 (CP2m). Mutants were as described for the EMSAs. Results were calculated as fold change relative to expression in the presence of the control expression vector alone (pTATA) and are presented as a percentage of the StuP2 basal activity without TSA treatment. Data are shown as means ± SEM and represent a minimum of three independent experiments.

In vivo DMS footprinting analysis

Cells were treated with 0.2% DMS in phosphate-buffered saline (PBS) for 5 min at room temperature. The DMS reaction was stopped by addition of ice-cold PBS, followed by two washes with PBS. Cells were lysed in proteinase K/SDS buffer overnight at room temperature and DNA was purified by phenol extraction and ethanol precipitation. In vitro DMS treatment of naked DNA was performed essentially as described (7,16). Cleavage sites were analysed by ligation-mediated PCR (LM–PCR) as described (7,8,17). PCR products were labelled by primer extension using ³²P-labelled nested primers and were analysed on 6% denaturing polyacrylamide gels. The following nested primer combination was used: D376 (–3716) (dGAGCTACACACCCTTCAGC) (–3735); D376A (–3732) (dCAGCTGTCTCTCCTTGATGG) (–3755); D376B (–3748) (dGATGGCAGCCTGCCCCACAAG) (–3768).

RESULTS

Transcription factors interacting with the –3.9 kb enhancer

We examined developmental changes in enhancer activity and chromatin structure of the –3.9 kb enhancer in retrovirally transformed haemopoietic cells lines representing multipotent progenitor cells (HD50 MEP), macrophages (HD11) and erythroblasts (HD37) (Fig. 1A). Only the HD11 cells express the lysozyme gene, as no transcripts can be detected in HD37 and HD50 MEP cells (2,8). Using MNase assays we previously showed that after the onset of lysozyme gene expression nucleosomes become specifically positioned over the active –3.9 enhancer (9,18). This finding was confirmed by a restriction enzyme accessibility assay using the frequently cutting enzyme HinfI (8). This experiment demonstrated that one region of increased MNase accessibility coincides with an increased accessibility for cleavage by the restriction enzyme HinfI. The accessible HinfI site separates the two parts of the enhancer (Fig. 1B), indicating that this region may comprise nucleosomal linker DNA.

Figure 1.

An AT-rich region confers tissue-specific expression of the –3.9 kb enhancer. (A) Transformed cell lines analysed in this study and their position in the haemopoietic hierarchy. (B) (Upper) Map of the chicken lysozyme locus with the cis-regulatory elements indicated above, the arrows marking the position of a DHS. The short vertical arrows indicate constitutive DHSs with their positions relative to the transcriptional start site indicated below, the black vertical arrow indicates the DHS at the silencer element (only absent in mature macrophages), the gray arrows display DHSs that appear at the myeloblast stage and the light gray arrow the DHS that appears late (in mature macrophages). E, enhancer element; S, silencer element; P, promoter. The position of the –3.9 enhancer is indicated by a gray box. (Lower) Map of the –3.9 kb enhancer. Position of essential restriction enzyme sites as well as accessible HinfI site are indicated. The ‘minimal’ enhancer is indicated by a gray box and the AT-rich region by a striped box. (C) The indicated luciferase reporter gene constructs plus RSV lacZ DNA as an internal control were co-transfected into HD37, MEP and HD11 cells by electroporation and luciferase and β-galactosidase activities were determined after 48 h. The StuP1 construct covers the –3.9 enhancer between the StuI and PvuII(1) sites, Stup2 between the StuI and PvuII(2) sites. Results were calculated as fold change relative to expression of the control expression vector alone (pTATA). Data are shown as means ± SEM and are calculated from at least three independent experiments.

Our previous experiments have assigned the tissue-specific enhancer activity of the –3.9 kb element to a 458 bp StuI–PvuII fragment between –4186 and –3729 bp upstream of the transcription start site. A minimal enhancer harbouring almost all stimulatory activity was assigned to a 138 bp NsiI–PvuII subfragment, however, this element did not act in a tissue-specific fashion (9). Figure 1C depicts a transient transfection assay which demonstrates that AT-rich 3′-flanking sequences are required to repress enhancer activity in lysozyme-non-expressing cells. Addition of the AT-rich region was sufficient to abolish relative enhancer activity in HD37 cells and caused a 7-fold reduction in enhancer activity in HD50 MEP cells. The StuP1 core enhancer has essentially the same activity in HD50 MEP and HD11 cells and addition of the AT-rich region gave only a modest 46% reduction in relative activity in HD11 cells.

In order to identify the transcription factors responsible for differential regulation of the –3.9 kb enhancer in haemopoiesis, we performed EMSAs with labelled overlapping wild-type and mutated oligonucleotides that span the entire enhancer core (Oligos 1–6, summarised in Fig. 2). Figure 3 displays EMSAs with Oligos 1–3 covering the 5′-part of the enhancer. Figure 3A demonstrates specific binding of different proteins that display the characteristic binding pattern of members of the C/EBP transcription factor family (14). Binding of this complex can be competed with a 100-fold excess of unlabelled oligonucleotide carrying a consensus C/EBP-binding site and by an oligonucleotide carrying the well-characterised C/EBP-binding sites of the –6.1 kb enhancer (14) (data not shown). In turn, mutation of the binding site (1/C/EBP–oligo) resulted in a loss of binding activity. Significant C/EBP binding activity was only found in HD11 cells. A similar strategy was used to show that Oligo 2 is recognised by an ets protein (Fig. 3B), which binds to the ets consensus sequence TTCCGTT. However, because the ets family contains several members, we do not yet know the nature of the ets family member which is present in the cell types tested here (Fig. 3C). Mutation of the ets consensus sequence to TGCAGTT in Oligo 2/ets abolished binding of the ets factor, but revealed that Oligo 2 also bound a second protein (Fig. 3B, right). This additional binding activity was present in all cell types analysed, but has not yet been identified. The second factor appeared to bind upstream of ets, because it was not competed by Oligo 3, which also carries the ets-binding site (Fig. 3C). We originally speculated that it was myc/USF recognising the sequence CACATG, but this could be ruled out by competition assays (data not shown). Mutation of a myb-binding site core (CAAT) next to the ets recognition sequence did not result in any change in retarded complexes or in the competition ability of Oligo 2 (data not shown).

Figure 2.

Sequence of the –3.9 kb enhancer and summary of DNA–protein interactions. Relevant restriction enzyme sites are indicated as well as the accessible HinfI site. Black arrows indicate positions and sequences of the different oligonucleotide designs for gel shift experiments. White boxes cover identified protein-binding consensus sequences as well as binding sites for unknown factors. The stippled bar indicates the AT-rich region of the enhancer. The position of the palindrome is indicated. Open circles indicate diminished DMS reactivity as compared to naked DNA.

Figure 3.

C/EBP, NF1 and an Ets family member bind to the ‘minimal’ enhancer. EMSAs with various radioactively labelled DNA fragments and 10 µg protein extracts from HD11, MEP or HD37 cells were performed as described in Materials and Methods. Competitions (comp.) were performed using a 100-fold excess of unlabelled DNA fragment. Sequences of the different oligonucleotides used in these experiments are depicted below each figure as well as oligonucleotides mutant for C/EBP (1/C/EBP), Ets (2/ets) and NF1 (3/NF1) and C/EBP and NF1 (NF1) consensus sequences. The black dot corresponds to the ets bandshift and the open circle to an unidentified one.

We have previously shown that Oligo 3 is recognised by members of the NF1 transcription factor family present in all cell types tested and binds to the characteristic palindromic sequence TGGCAAATCTGCCCA (9). Figure 3C reproduces this result and additionally demonstrates that point mutation of the NF1-binding site in Oligo 3/NF1 does not abolish binding of the ets protein.

EMSAs analysing the 3′-part of the enhancer with Oligos 4–6 are shown in Figure 4. Oligo 4 covers a large palindrome. This sequence is highly AT-rich and has a different electrophoretic mobility as compared with other oligos of the same size in the –3.9 kb enhancer region (C.Bonifer, unpublished observations). It binds a tissue-specific set of factors that are only present in HD11 macrophage cells. Destruction of the palindrome resulted in a major change in protein binding to this sequence (Fig. 4A). The nature of these factors is presently unknown. Although Oligo 4 harbours a consensus sequence for HMGI(Y) proteins, we were not able to find any interaction of this particular factor with this sequence (P.Lefevre, R.Reeves and C.Bonifer, unpublished data).

Competition and mutation experiments demonstrated that Oligo 5 was recognised by a protein binding to the Fox consensus binding site ATAAACA (Fig. 4B) in all cell types tested. The Fox family of tissue-specific and developmental gene regulators uses a 100 amino acid winged helix domain that consists of a modified helix–turn–helix, for monomeric recognition of specific DNA target sites (19,20). The binding sites for the different Fox proteins share a core sequence RTAAAYA, but differ in the positions flanking the core (21,22). Oligo 5 was also recognised by a factor forming a complex with a higher electrophoretic mobility. This complex did not appear to be a Fox protein because it still bound to the mutated Oligo 5 probe and it could be competed by unrelated sequences (data not shown).

Oligo 6 was recognised by a protein present in all cell types tested that could be competed with a consensus sequence for CP2α (Fig. 4C). The ubiquitous transcription factor CP2α (LBP-1c/LSF) is the best characterised member of the NTF (neurogenic element binding transcription factor) family. A role for CP2α has been proposed in α-globin gene expression (15), T cell responses to mitogenic stimulation (23) and several other cellular processes. However, the factor we have detected did not appear to be CP2α, because an oligonucleotide carrying a CP2α consensus sequence yielded a different binding pattern and Oligo 6 did not compete binding to a CP2α probe (Fig. 4C, middle).

In order to confirm the significance of our in vitro assays, we examined which of the transcription factor binding sites of the –3.9 kb enhancer are occupied in HD11, HD37 and HD50 MEP cells in vivo. We performed DMS treatment and subsequent LM–PCR analysis of the top strand. The results are indicated in Figure 5 and summarised in Figure 2. The results could be partially reproduced from the other strand. In HD11 cells, several in vivo DNA–protein interactions can be seen over the 5′-part of the enhancer element, the most prominent ones over the NF1-binding site and the juxtaposed ets consensus sequence. However, due to the structure and sequence composition of the DNA in this area, we were not able to find suitable LM–PCR primers near the enhancer core and had to use primers starting from far upstream that did not reach into the 3′-part of the enhancer (data not shown). To what extent the C/EBP site is occupied in vivo is presently unclear, since the footprint over the C/EBP site on this strand was very weak and could only be seen in lipopolysaccharide-treated HD11 cells (data not shown). No in vivo footprints were visible in the 3′-part of the enhancer in HD11 macrophage cells. This contrasts with lysozyme-non-expressing cells, where in the AT-rich region a clear protection of a G at –3803 bp can be seen, which is located over the consensus sequence for Fox proteins.

Figure 5.

In vivo footprinting analysis of the –3.9 kb enhancer. LM–PCR was carried out with primer set D376, which amplifies the upper strand. The chicken cell lines used (HD37 erythroblasts, HD50 MEP multipotent progenitor cells and HD11 macrophage cells) are shown at the top. (+) or (–) indicates cells grown in the presence or absence of 100 nM TSA. Lanes 1 and 2 show Maxam–Gilbert G+A and G reactions, respectively, examined by LM–PCR using a linker to give the same size as for TD–PCR. Open circles indicate diminished DMS reactivity as compared to naked DNA with the position relative to the transcriptional start site indicated at the right. In vitro transcription factor-binding sites are indicated at the right. The position of an accessible HinfI site over an AT-rich palindromic sequence is indicated by a black arrow.

A member of the Forkhead (Fox) family of transcription factors is required for enhancer repression in lysozyme-non-expressing cells

The role of the different factor-binding sites in regulation of the –3.9 kb enhancer was examined by transient transfections of luciferase reporter gene constructs harbouring wild-type and mutated versions of the 458 bp StuI–PvuII(2) fragment in front of a minimal promoter (Fig. 6). We transfected these constructs into HD50 MEP multipotent progenitor cells and HD11 macrophage cells. Mutation of the various factor-binding sites had a pronounced but differential effect on reporter gene expression in the two cell types. In multipotent progenitors (HD50 MEP, Fig. 6A) mutations of the CP2-, NF1- and ets-binding sites resulted in a reduction in enhancer activity. In contrast, mutation of the Fox-binding sites resulted in a 2-fold increase in reporter gene activity. Mutation of the palindrome and the C/EBP site did not impair enhancer function. A markedly different picture was seen in HD11 cells (Fig. 6B). Here, the most important factors for enhancer activity were ets and C/EBP, and mutations of the NF1 and CP2 sites only slightly (but reproducibly) reduced enhancer activity. Similar to HD50 MEP cells, mutation of the palindrome had no effect and an even more dramatic 8-fold derepression of enhancer activity was observed with the Fox site mutation. In summary, we conclude that in contrast to HD37 cells, HD50 MEP and HD11 cells share a number of factors that are able to drive reporter gene activity in transfected cells via the same consensus sequences (ets, CP2 and NF1). However, none of these factors appear to bind in vivo in HD50 MEP cells. This contrasts with the Fox factor, which represses the activity of the –3.9 kb enhancer. The increase in activity resulting from Fox mutation was essentially the same as the increase resulting from deletion of the AT-rich region in both HD11 and HD50 MEP cells.

TSA treatment can override enhancer repression, but only in multipotent progenitor cells

The acetylation of core histone N-termini is an essential biochemical feature of active and potentially active genes in eukaryotes. The modification is dynamic, a consequence of the competing activities of histone acetyltransferases (HATs) and deacetylases (HDACs). However, histones are not the only targets for HATs and HDACs. More recently, it was shown that HATs can not only acetylate histones, but also transcription factors, such as GATA1 (24) and HMGI(Y) (25). We have previously shown by UV-photofootprinting experiments looking at the endogenous lysozyme gene that treatment of HD50 MEP cells with the HDAC inhibitor TSA has a pronounced effect on the chromatin fine structure of the AT-rich 3′-half of the –3.9 kb enhancer (8). The molecular basis of this change in DNA structure is presently unclear. However, TSA treatment was not sufficient to induce expression of the lysozyme gene in these cells. Figure 5 demonstrates, in addition, that TSA treatment was not sufficient to induce in vivo transcription factor binding to the –3.9 kb enhancer in HD37 or HD50 MEP cells. The same experiments also showed no change in the transcription factor binding pattern in TSA-treated HD11 macrophage cells. In order to clarify the role of HATs in regulation of the –3.9 kb enhancer we examined the activity of the StuI–PvuII(2) construct in HD37, HD50 MEP and HD11 cells with and without TSA treatment (Fig. 7A). Substantial stimulation of reporter gene activity was only observed in HD50 MEP cells and not in any other cell type and it was dependent on the presence of the AT-rich region. In HD11 cells only the StuI–PvuII(1) construct carrying binding sites for C/EBP, ets and NF1 responded with a slight, but reproducible, increase in its activity to TSA treatment. This stimulation could be mimicked by co-transfection of an expression plasmid coding for a functional p300 acetyltransferase (26; Fig. 7B). This also held true for the StuI–PvuII(2) construct in HD50 MEP cells, albeit to a lesser extent (Fig. 7C). No stimulation of reporter gene activity was observed with a control plasmid coding for a truncated p300 protein lacking HAT activity in HD50 MEP cells. Moreover, in HD11 cells enhancer activity was reduced, indicating a dominant negative effect of the truncated HAT protein.

Figure 7.

The histone acetyltransferase protein p300 mimics TSA activation of the –3.9 kb enhancer in transient transfection. (A) Activation of reporter gene expression by TSA treatment of cells. Reporter constructs and control constructs were co-transfected into HD37, MEP and HD11 cells by electroporation. Subsequently, the cells were incubated with or without 100 nM TSA for 48 h. Results were calculated as the ratio of relative luciferase activity in treated and untreated cells and normalised to relative activation of the control expression vector alone (pTATA). (B) Stimulation of the StuP1 construct by p300 expression in HD11 macrophage cells. The StuP1 construct covers the minimal enhancer between the StuI and PvuII(1) sites. (C) Stimulation of the StuP2 construct by p300 overexpression in HD50 MEP precursor cells. The StuP2 construct covers the entire enhancer between the StuI and PvuII(2) sites. Data are shown as means ± SEM and represent a minimum of three independent experiments.

Next, we sought to unravel the role of the different transcription factors in the stimulation of enhancer activity by TSA. We transfected constructs carrying the various point mutants described above into HD50 MEP and HD11 cells and treated these cells with TSA for 48 h (Fig. 6). These experiments proved again that TSA treatment has no stimulatory effect on any of the StuI–PvuII derived constructs in HD11 macrophage cells (Fig. 6B). However, enhancer activation as a result of removal of the Fox-binding site was reduced after TSA treatment. This is in contrast to HD50 MEP cells (Fig. 6A). Here the Fox mutant can still be stimulated by TSA treatment. The same holds true for a mutation of the palindrome, the C/EBP site and the CP2 site (albeit from a low level). C/EBP is not expressed in HD50 MEP cells and could also not be induced by TSA treatment (data not shown). However, mutation of the ets site severely reduced TSA stimulation and mutation of the NF1 site abolished it altogether, indicating a crucial role of these factors in TSA-induced enhancer activity.

DISCUSSION

Tissue-specific and non-tissue-specific transcription factors binding to the –3.9 kb enhancer

Our experiments define a number of transcription factors crucial for the activity of the chicken lysozyme –3.9 kb enhancer. However, we could only identify two sequences, a C/EBP site and the palindrome, which were recognised by DNA–protein complexes specific for HD11 cells. All other binding activities, NF1 (9) ets, Fox and CP2, were present in HD50 MEP and HD37 cells as well. Judging from the mutation analysis, the most important factors driving enhancer activity in HD11 cells are ets and C/EBP family members. The role of the NF1 site is somewhat puzzling, because although a strong in vivo footprint was seen, mutation of this site had only a slight effect. These results suggest a role of NF1 in enhancer activity in a chromatin context (see also below). NF1 is unable to bind to nucleosomal DNA, and A-tracts similar to those that flank the NF1-binding sites in the –3.9 kb enhancer can modulate its binding affinity in chromatin (27–29). The low, but measurable, basal activity of the –3.9 kb enhancer in HD50 MEP cells was mediated by the same factors (except for C/EBP). However, here the CP2-like protein seems to play a more important role in enhancer function than ets. Interestingly, similar to the experiments described here, CP2α binding has been shown to respond to HDAC inhibition (30). Its binding to the γ-globin gene promoter is induced in vivo in response to butyrate therapy. However, CP2α and the factor binding to the –3.9 kb enhancer are clearly different.

We have identified a factor resembling the Fox transcription factor family as an essential factor for tissue-specific activity of the –3.9 kb enhancer. This factor played a major role in repression of the chicken lysozyme –3.9 kb enhancer activity in lysozyme-non-expressing multipotent progenitor cells. This finding uncovers an interesting similarity between the –3.9 kb enhancer and the albumin enhancer, which is also organised in a specifically positioned nucleosome when it is active (31). Binding of the Fox factor HNF3 to this enhancer in vitro is necessary and sufficient to specifically position a nucleosome (32,33). Interestingly, the central globular domain of linker histone H5 and the ubiquitous linker histone H1 exhibit a winged helix structure which is similar to that of HNF3 except for the lack of a second wing (34,35). This may be the reason why HNF3 binds more stably to nucleosomal than to free DNA (36). This may also be true for the protein we have identified here, because it can bind to inactive chromatin. However, whereas HNF3 is an essential factor for enhancer activation, the protein identified here antagonises enhancer activity. Fox factors with repressive activity have been previously described (37–39). Our data also suggest that, in contrast to the albumin enhancer, binding of the Fox factor in lysozyme-non-expressing cells does not lead to nucleosome phasing. We are presently testing two hypotheses to address this issue: it is possible that Fox proteins recruit a HDAC activity (see below) that prevents binding of activators, which are required for nucleosome positioning as described for the specific positioning of nucleosomes over the mouse mammary tumour virus (MMTV) long terminal repeat (40). It could also be that the specific nucleosome position required for NF1 binding is the default state mediated by a sequence-specific DNA structure in the AT-rich region and that binding of the Fox factor interferes with this process.

The developmental activity of the –3.9 kb enhancer is mediated by the interplay of repressing and activating factors

Enhancer activity can be enhanced by TSA treatment in transient transfection assays of HD50 MEP cells, but not in HD37 cells. The experiments described in this study give the first indications of possible targets for HAT and HDAC activities involved in regulation of the –3.9 kb enhancer, since the different factors contributed differently to TSA-stimulated enhancer activity in HD11 macrophage and HD50 MEP cells. In HD50 MEP cells the binding of NF1 and ets was crucial for TSA stimulation of the –3.9 kb enhancer. This could mean that NF1 and ets factors are either involved in recruiting HAT activities to their binding sites or that their activities are directly regulated by acetylation. Our experiments argue against the latter hypothesis. EMSAs performed with untreated or TSA-treated cells showed no specific differences in the binding activities of any transcription factor involved in –3.9 kb enhancer regulation (data not shown). We also do not see any induction of new binding activities. Instead, we observed a general TSA-mediated 2- to 3-fold increase in all transcription factor activities (measured per microgram protein) as compared to untreated cells. However, we attribute this to a general alteration in protein composition, since TSA-treated cells have altered growth characteristics. We therefore suggest that in HD50 MEP precursor cells, NF1 and ets are involved in the recruitment of HAT activities to their binding sites. Both protein families were shown to have this capacity (41–43).

In HD11 macrophage cells, the most important factors for enhancer function were ets and C/EBP family members, whereas mutation of the NF1 site had only a small effect. C/EBP proteins are one of two macrophage-specific binding activities interacting with the –3.9 kb enhancer. C/EBP proteins have been shown to recruit chromatin-modifying activities. Transfection of C/EBPβ together with v-myb leads to transcriptional activation of the lysozyme gene in cells where it is normally not expressed, via recruitment of SWI/SNF (44,45). It is therefore possible that C/EBP proteins play an essential role in the formation of a stable enhancer complex in lysozyme-expressing cells. In addition, non-DNA-binding, tissue-specific cofactors could be involved in the assembly of a multi-component, tissue-specific transcription complex similar to that described for the human interleukin-3 enhancer (46).

TSA treatment did not stimulate activity of the complete enhancer in HD11 cells. Furthermore, overexpression of a truncated and possibly dominant negative version of p300 repressed the stimulatory activities of the ets-, C/EBP- and NF1-binding sites present on construct StuP1. We therefore conclude that in HD11 cells all relevant chromatin components, stimulatory transcription factors and cofactors are in the maximally active and, most likely, hyperacetylated conformation. In addition, TSA treatment reduced the increase in reporter gene expression caused by mutation of the Fox-binding site from 8- to 4-fold. Our interpretation of this result is that Fox protein counteracts ets-, NF1- and C/EBP-mediated activation either directly or by recruiting HDAC activities (see above). If HDAC activities are inhibited in TSA-treated cells, the repressive activity of the Fox factor is reduced. In the binding site mutation this leads to an increase in basal activity. The fact that TSA stimulation can be partially mimicked by overexpression of p300 indicates that it is the balance between repressive/HDAC-recruiting and HAT-recruiting factors that regulates the developmental activity of the –3.9 kb enhancer element.

Implications for the regulation of enhancer activity and chromatin structure of the endogenous lysozyme gene

Our data confirm the notion that in multipotent macrophage progenitor cells, in contrast to erythroblasts, most of the activities regulating the –3.9 kb enhancer are present. As regards the transcription factor complement, and this is what is measured in transient transfection assays, everything is ready to initiate transcription in multipotent myeloid progenitor cells. What is missing is a constitutively recruited HAT activity that is required for transcriptional competence of our enhancer constructs in transient transfection assays. At the molecular level, this may include acetylation of histone tails, architectural factors or transcriptional cofactors required by RNA polymerase II. However, the presence of a functional set of transcription factors sufficient to drive expression from the –3.9 kb enhancer from transiently transfected templates in multipotent progenitor cells after TSA treatment is not sufficient to induce chromatin remodelling at the enhancer of the endogenous gene. The reason for this is most likely the establishment of a repressive chromatin structure in these cells. Although the chromatin structure of the endogenous gene did not permit gene expression, we found a chromatin fine structure pattern specific for expressing cells in multipotent myeloid progenitor cells, but not in cells of a different lineage. The experiments in this study point to a possible mechanism for this finding. We speculate that factors like ets, NF1 and CP2 and their interaction partners are able to transiently recruit chromatin remodelling activities and leave behind a modified chromatin structure, without being able to form a stable transcription factor complex. The reason for this inability could be the recruitment of HDAC activities by repressive factors like the Fox protein and/or the absence of tissue-specific activator/co-activator proteins, like C/EBP. Our future aims will therefore be to unravel the hierarchy of the interactions leading to the formation of a stable enhancer complex.