Human H 4 Histone Gene Transcription Requires the Proliferation-specific Nuclear Factor HiNF-D AUXILIARY ROLES FOR HiNF-C (Spl-LIKE) AND HiNF-A (HIGH MOBILITY GROUP-LIKE)*

The proximal promoter of the human H4 histone gene FOlO8 contains two in vivo protein binding domains, sites I and 11. In this report we show that these se- quences interact with three nuclear factors: HiNF-D, HiNF-C, and HiNF-A. HiNF-C is a metal ion-requiring protein that binds to an Spl consensus binding site. HiNF-C and HiNF-A bind independently to the distally located site I, possibly in conjunction with other proteins, and deletion of site I reduces transcription rates 4- to 6-fold in vitro. Factor HiNF-D binds to an H4 histone-specific element (5”dGGTPyPyTCAATCNG- GTCCG, where Py indicates pyrimidine) present in site I1 that has previously been shown to be essential for in vivo expression of this H4 histone gene. All three bind- ing activities are present in human HeLa S3 cells throughout the cell cycle and in exponentially growing mouse C127 and human HL60 cells. This result is consistent with the transcription of H4 histone genes throughout the cell cycle. However, unlike HiNF-A and HiNF-C, HiNF-D is not present in terminally differentiated HL60 cells, in which histone gene transcrip- tion is down-regulated. These findings suggest a crucial role for HiNF-D, with an auxiliary role for HiNF-

The proximal promoter of the human H4 histone gene FOlO8 contains two in vivo protein binding domains, sites I and 11. In this report we show that these sequences interact with three nuclear factors: HiNF-D, HiNF-C, and HiNF-A. HiNF-C is a metal ion-requiring protein that binds to an Spl consensus binding site. HiNF-C and HiNF-A bind independently to the distally located site I, possibly in conjunction with other proteins, and deletion of site I reduces transcription rates 4-to 6-fold in vitro. Factor HiNF-D binds to an H4 histone-specific element (5"dGGTPyPyTCAATCNG-GTCCG, where Py indicates pyrimidine) present in site I1 that has previously been shown to be essential for in vivo expression of this H4 histone gene. All three binding activities are present in human HeLa S3 cells throughout the cell cycle and in exponentially growing mouse C127 and human HL60 cells. This result is consistent with the transcription of H4 histone genes throughout the cell cycle. However, unlike HiNF-A and HiNF-C, HiNF-D is not present in terminally differentiated HL60 cells, in which histone gene transcription is down-regulated. These findings suggest a crucial role for HiNF-D, with an auxiliary role for HiNF-C and possibly HiNF-A, in the regulation of H4 histone gene transcription. Furthermore, the conservation of potential HiNF-D binding sites in mammalian H4 histone gene promoters suggests that HiNF-D has an essential role in the coordinate transcriptional downregulation of the H4 histone multigene subfamily during the shutdown of proliferation.
Histone genes comprise a heterogeneous multi-copy family with gene products that are an essential component of eukaryotic chromatin. The expression of the most abundant class of histone genes (designated cell cycle-regulated) is tightly coupled to DNA replication and, hence, highly specific for proliferating cells (reviewed in Refs. 1-3). Thus, understanding the control mechanisms that influence the expression of the proliferation-specific histone genes should provide insight into gene regulatory events occurring in both normally dividing and transformed cells as a function of the cell cycle (1-3), cell aging (4, 5 ) and the onset of differentiation (6, 7).
Cell cycle-dependent histone gene expression has a prominent post-transcriptional component (reviewed in Refs. 8,9) * This work was supported by National Institutes of Health Grant GM32010, National Science Foundation Grant DCB88-96116, and March of Dimes Birth Defects Foundation Grant 1-1091. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. as evidenced by a 15-20-fold increase in histone mRNA levels when DNA replication initiates and a rapid destabilization of histone mRNA at the natural (or chemically induced) termination of DNA synthesis (10-15). Histone mRNA levels parallel the extent of histone protein synthesis, consistent with histone mRNA abundance as a rate-limiting step in histone gene expression. Histone mRNA turnover requires continued protein synthesis (16,17) and may involve 3' end processing events (18,19), particular histone mRNA sequences (20-23), a histone mRNA degrading exonuclease (24, 25), and possibly specific subcellular structures (26,27).
The transcriptional component of histone gene expression during the cell cycle involves a 3-to 5-fold transient increase of histone mRNA synthesis rates at the onset of S-phase (10-15) and contributes to the accumulation of histone mRNA in the early half of S-phase. However, in dividing cells replication-dependent histone mRNAs are synthesized throughout the cell cycle (albeit rapidly degraded) and only in terminally differentiated cells that have lost the ability to proliferate is there a complete shutdown of histone mRNA transcription (6).
Vertebrate histone gene promoters have a modular organization, and the transcriptional regulation of each individual histone gene involves multiple, distinct promoter elements (28)(29)(30)(31)(32)(33)(34). Several DNA binding proteins interacting i n vitro with such elements in the promoters of H1 (33,(35)(36)(37), H2B (38, 39), H3 (34,40), and H4 (40-43) histone genes have been characterized. These studies suggest that transcriptional control of histone gene expression involves a plethora of promoter factors that either are histone-specific or recognize a broad spectrum of gene promoters. A common principle underlying the coordinate transcriptional regulation of the histone genes remains to be established. Recently, our laboratory has established the i n vivo sites of DNA/protein interactions of human H3 and H4 histone gene promoters (44-46). From these in vivo DNA/protein interaction data it has become apparent that the histone gene promoter factors studied thus far (33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43) comprise only a subset of the full complement of factors that potentially can associate with each individual histone gene.
The active expression of the H4 histone gene F0108 in proliferating cells involves a dynamic cell cycle-dependent chromatin structure (47-50) and both distal and proximal promoter elements (28)(29)(30). We have previously shown that the H4 histone proximal promoter interacts i n vitro with a factor designated HiNF-A (40) that also binds to H3 (40) and H1 histone gene promoters (35). The H4 histone proximal promoter contains two in vivo DNA/protein interaction sites (sites I and 11) (44). In this study we have examined the binding i n vitro of three distinct nuclear factors (HiNF-A, HiNF-C, and HiNF-D) to these DNA sequences. The binding Human H4 Histone Gene Promoter Factors 15035 sites of the factors were defined by deletion analysis, DNaseI footprinting, and/or DMS' fingerprinting. We have explored the functional significance of these DNA/protein interaction sites by in uitro transcription assays and by monitoring the presence of the factors involved in various cell lines and biological situations. We propose that the interaction of HiNF-D with the proximal promoter is crucial in the regulation of H4 histone mRNA synthesis and is implicated in the down-regulation of histone gene transcription at the onset of differentiation.

MATERIALS AND METHODS
Fractionation of Nuclear Extracts-Undialyzed nuclear extracts (UNE) from exponentially growing HeLa S3 cells (35) and nuclear extracts of mouse C127 monolayer cells (36) were prepared by the procedure of Challberg and Kelly (51). Dialyzed nuclear extracts (DNE) from HeLa S3 cells synchronized by double thymidine block (40) were prepared according to Dignam et al. (52) at multiple hourly intervals after release from the blockade, extending into the second S-phase (up to 25 h). The preparation of nuclear extracts from proliferating and differentiated human HL60 cells was done according to the Dignam procedure (52). Exponentially growing HL60 cells were induced to differentiate by treatment with 16 nM of the phorbol ester TPA (12-0-tetradecanoylphorbol-13-acetate) and extracts were prepared after 4.5 days.
The fractionation of nuclear proteins has been described previously (35). In brief, human HeLa S3 nuclei were extracted with 0.4 M KCl, and the extract was diluted and passed over a DEAE-Sephacel column equilibrated at 200 mM KC1. The DEAE-Sephacel flow-through (DO-200 fraction) was fractionated using a phosphocellulose column equilibrated at 200 mM KCl. The flow-through was collected (PO-200) and bound proteins were eluted with buffers containing 350 mM (P200-350) and 1000 mM KC1 (P350-lOOO), respectively. HiNF-A elutes in the P350-1000 fraction, whereas HiNF-D elutes under these circumstances in both the PO-200 and the P200-350 fraction (data not shown). None of the above chromatography fractions contains significant amounts of HiNF-C activity.
Reciprocal binding site deletion analysis (stairway assay) (35,36,40) was performed by digesting aliquots (10 ng) of a single end-labeled probe (>200 base pairs) with various restriction enzymes to yield progressively shortened probes with identical molar specific activity. After digestion and organic extractions the probes were ethanolprecipitated in the presence of 5 pg of glycogen (Boehringer Mannheim) and used for binding reactions in parallel.
Chelation experiments were performed by preincubating undialyzed nuclear extracts for 15 min on ice in dilution buffer (20% glycerol, 20 mM NaC1, 0.1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 25 mM Hepes/NaOH, pH 7.5) containing various concentrations of chelators; the preincubated extracts were diluted 2fold during the binding reaction. For reconstitution experiments, undialyzed nuclear extracts were preincubated for 15 min on ice in nuclear extraction buffer (10% sucrose, 0.4 M KC1, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 50 mM Hepes/NaOH, pH 7.5) containing 0.5 mM 1,lO-phenanthroline (stock solution: 0.5 M in isopropyl alcohol). The mixture was diluted 10-fold in dilution buffer containing the indicated amounts of divalent cations, incubated for an additional 15 min on ice, and finally diluted 2-fold in the binding reaction.
I n Vitro Transcription Analysis-Zn vitro transcription experiments were performed with supercoiled plasmids KUC8 and JUC50 essentially as described by Dignam et al. (52). RNA transcripts were analyzed by S1 nuclease protection using a single end-labeled probe overlapping the transcriptional initiation site. The reaction products were analyzed on a 6% sequencing gel and the intensity of a 280-nt fragment, corresponding to accurately initiated F0108 H4 histone mRNA, was quantitated by densitometry.

RESULTS
The proximal promoter of the human H4 histone gene F0108 contains multiple consensus regulatory sequences that can potentially function as recognition sites for distinct DNA binding proteins (28,44; refer to Fig. 5 for details). Several of these elements appear to interact in uiuo with factors present in HeLa S3 cells and are located in two domains designated site I (nt -151 to -114) and site I1 (nt -97 to -58) (44). Site I contains AT-rich sequences (distal), an AAATGACG-motif (distal), and an Spl consensus sequence (proximal); site I1 consists of an H4 histone-specific element (distal) and the TATA box (proximal). An additional Spl box located immediately upstream of site I1 does not appear to be bound by a factor in vivo in HeLa S3 cells.
Factor HiNF-A Binds to the Distal Part of Site I-The binding of HiNF-A to the promoter of the F0108 H4 histone gene ( Fig. 1) was studied using several deletion mutants and synthetic DNA fragments in order to define the binding site requirements for this factor. First, the binding of HiNF-A to a radiolabeled DdeIITqI fragment (nt -183/-133) was established in a gel-retardation assay (Fig. lA). Incubation of this probe with increasing amounts of nuclear extracts (lunes 1-5) or partially purified HiNF-A (P350-1000 fraction) (lunes 6 and 7 ) shows a fast migrating complex characteristic of HiNF-A. The binding of HiNF-A was also examined by the gel-retardation assay using BAL31 deletion mutants of the H4 histone promoter (Fig. 1B). HiNF-A binds to the -182/ -13 probe (lunes 1-3), but not the -147/-13 or -141/-13 fragments (lunes 4-6 and 7-9). The binding of HiNF-A to synthetic oligonucleotides spanning the distal part of site I (DSI: nt -152/128) and the distal part of site I1 (DSII: nt -91/-64) was studied ( probe, regardless of the type of competitor DNA (lanes 3 and 4). In summary, the results of the binding site deletion analysis establish the maximal boundaries of the HiNF-A binding site at nt -152 and -133, which coincides with the distal part of site I. However, several lines of evidence suggest that HiNF-A may not be the only factor binding to this region (see below).
Two Additional Factors, HiNF-C and HiNF-D, Interact with the H4 Histone Promoter-The DNA sequences in the proximal promoter of the F0108 H4 histone gene that correspond to the two in vivo sites of DNA/protein interaction have the potential to bind at least four distinct factors, based on the presence of consensus promoter elements and the distribution of protein/nucleotide contacts (Ref. 44; see Fig.  5). At least six in vitro DNA-protein complexes that occur in the H4 histone proximal promoter between nt -234 and nt -13 have been identified in previous work (40). One involves a nonspecific DNA binding activity and another is the HiNF-A complex. The binding site requirements of the factors present in the other complexes were initially examined by means of stairway assays (reciprocal binding site deletion analysis).
Fragments were labeled at the EcoRI site (nt -240) or BanII site (nt -13) and digested with various restriction enzymes (Fig. 2, A and B, refer also to Fig. 3). Binding reactions containing these probes were electrophoresed in parallel in polyacrylamide gels. Sequences downstream of the ThaI site (nt -94) ( Fig. 2A, lanes 1-5) and upstream of the Tap1 site (nt -130) (Fig. 2B, lanes 1-4) are dispensable for the formation of the HiNF-C complex. Two faint bands migrating above and below the HiNF-C complex can be distinguished on shorter exposures (not indicated). These complexes appear to require the same DNA sequences and may reflect microheterogeneity of a similar factor or complexes involving HiNF-C. Sequences between the ThaI site (nt -94) and MspI site (nt -99) can be removed without abolishing the binding of HiNF-C (data not shown). Thus, the stairway assay establishes the maximal boundaries of the HiNF-C binding site between nt -99 and -130.
The binding requirements for another complex with a low relative mobility (HiNF-D) were determined using partially purified fractions of this factor. The HiNF-D complex requires sequences between the MboII site (nt -38) ( Fig. 2A,  lanes 6-10) and the ThaI site (nt -93) (Fig. 2B, lanes 5-8). Also, probes having deletions upstream of the AvaII site (nt -70) (Fig. 2B, lane 9) and downstream of the same site (nt -74) (data not shown) do not efficiently bind HiNF-D. The AvaII site coincides with a histone-specific element (see below), and, hence, the stairway assay demonstrates that the HiNF-D binding site is located between nt -93 and -38 and that HiNF-D requires sequences at or near the histone-specific element.
Factor HiNF-C Binds to the Proximal Part of Site I-The binding requirements for HiNF-C (nt -99 to nt -130) partially overlap with the sequences of in vivo DNA/protein interaction site I. The analysis was extended by using BAL31 deletion mutants in the assay to define further the HiNF-C binding site (Fig. 2C). We found that the HiNF-C complex is formed with the -182/-13 (Kg), -147/-13 (K6) and -141/ -13 (K2) probes, but the binding of HiNF-C to the -126/ -13 (J50) probe is strongly reduced. This result suggests that base pairs between nt -130 (TaqI site) and nt -126 in the proximal part of site I are required for optimal site-specific binding of HiNF-C (see Fig. 4 for summary).
The binding site of HiNF-C was also determined by DNaseI footprinting of the isolated HiNF-C complex, and we observed protection of nt -113 to -134 on the top (sense/+) strand (Fig. 3A). Dimethyl sulfate protection analysis (DMS fingerprinting) was performed using the top (sense/+) strand and total nuclear protein. Fig. 3B shows that nuclear factors protect guanines at nt -123, -125, -126, -127 and -129, coinciding with the Spl box (54). Together, the results of the deletion analysis, DNaseI footprinting, and DMS fingerprinting strongly indicate that HiNF-C is the factor that binds to the Spl box. Because the same protein/guanine contacts coinciding with the Spl box are observed in vivo and in vitro (refer to Fig. 5), this suggests that HiNF-C is the bona fide factor that binds to the proximal part of site I in vivo.
Apart from protein/guanine contacts in the Spl box, we observed DMS protection of guanines nt -139 and -142 using total nuclear protein (

Sitel
' O I ODO

HiNF-D
?ATAI!*=!P!l ""--40 lished by deletion analysis and DNaseI footprinting and most likely are not due to HiNF-A binding as the DMS fingerprinting results were obtained in the presence of excess poly(d1).

C C C G C C G G C G C G C T~T C G~T~_~_~S T C T G G T C~C G T C T T G~A -T~~C~~~G G A A G A C G G T G C T C G C C T T G A C G G G C G G C C G C G C G A A A G~C A A A~G~~~G~~~A~~~l A T G A G A A C A l A l A G l C C C C l l C l G C C A C G A G C G G A
(dC) DNA, which prevents detection of this factor i n vitro (see Fig. IC). The distal part of site I contains, apart from AT-rich repeats implicated in the binding of HiNF-A, a DNA sequence element (5'dAAATGACG) that is also present in the human H3 histone gene promoter ST519. The element is located in the i n uiuo DNA/protein interaction domain designated H3-site I, and both guanines are protected from DMS attack in vivo (45). We have established by DNaseI footprinting and DMS fingerprinting that the element in the H3 histone promoter is recognized by a factor designated HiNF-E, but we could not unequivocally identify a single DNAprotein complex corresponding to this factor in gel-retardation assays).' Hence, it is possible that the distal part of H4-

Human H4
Histone Gene Promoter Factors site I may interact with HiNF-E or a related factor.
Factor HiNF-D Binds to the Distal Part of Site 11-The interaction of HiNF-D with sequences at or near the histonespecific consensus elements in site I1 was confirmed and analyzed in more detail by DNaseI footprinting experiments. HiNF-D was bound to a probe (nt -130 to -40) labeled on the bottom (anti-sense/-) strand and the binding mixture treated with DNaseI. After electrophoretically separating the HiNF-D complex from unbound DNA on native gels, the DNaseI cleavage products were analyzed on a denaturing gel. Fig. 3C demonstrates that HiNF-D clearly protects a number of phosphodiester bonds between nt -66 to -90 on the bottom strand from DNaseI cleavage. Some nucleotides were less well protected than others, and we also observed apparent partial protections beyond nt -66 near the TATA box (not indicated); however, only between nt -66 and -90 did we observe contiguous protections. This region is entirely contained within the distal part of the in vivo DNaseI footprint designated site I1 and encompassses the histone-specific element (see Fig. 5 for summary).
The binding of HiNF-D to site I1 was further established by competition analysis using a DNA fragment spanning the entire site I1 (nt -98 to -40). The inclusion of 20-to 50-fold molar excess of this DNA fragment resulted in a corresponding decrease in the intensity of the band representing the HiNF-D complex, but not in the band representing the HiNF-C complex (Fig. 30). Competition experiments with the distal site I1 oligonucleotide (nt -91 to -64) showed a similar decrease in HiNF-D complex formation, although competition was less efficient, possibly due to the smaller size of the DNA fragment used (data not shown).
A number of vertebrate H4 histone gene promoters (31,44, 55-58) contain sequences remarkably similar to the HiNF-D binding site. These sequences are located at analogous positions within these promoters and conform to the (mammalian) consensus sequence: 5'dGGTPyPyTCAATCNGGTC-CG (Py is pyrimidine).
Effect of Site I and Site 1 1 on H4 Histone Gene Transcription-The functional properties of H4 histone promoter elements were studied utilizing BAL31 deletion mutants in an in vitro transcription system. We used supercoiled templates (containing the intact H4 histone gene with promoter segments truncated at nt -182 (K8) and nt -100 (J56)), incubated these in the presence of nuclear extracts, and subjected the transcription products to S1 nuclease analysis (Fig. 6). The amount of accurately initiated transcription obtained with mutant K8, comprising both sites I and 11, was 4.2 arbitrary densitometry units (n = 4, S.D. = 2.8), whereas mutant 556, which has a deletion spanning site I, yielded 0.83 units (n = 4, S.D. = 0.33). No significant effect was observed in vitro when deletions were introduced into site I1 (data not shown). Thus, the site I DNA/protein interactions are capable of augmenting H4 histone gene transcription rates 3-5-fold.
The in vivo expression of the human H4 histone gene F0108 has been studied in a heterologous murine cell system (mouse C127 cells) (29). The cross-species compatability of mouse and human promoter-binding factors is evidenced by gel retardation assays performed with mouse (2127 nuclear extracts. Complexes are detected that are similar to those formed by HiNF-D, HiNF-C (Fig. 7, lanes 1-4), or HiNF-A (lanes 5-8), and the binding sites of the murine factors coincide with those of the human counterparts as established by stairway assays using human and mouse nuclear proteins in parallel (data not shown). Although no effects with site I1 deletions up to nt -74 were observed in the homologous in uitro system, Kroeger et al. (29) have reported a profound effect of the same deletions extending into site I1 in the murine in vivo system. Accurate initiation of H4 mRNA transcription in vivo was observed when sequences up to nt -100 are present, but when sequences upstream of nt -74 were deleted no human H4 histone mRNAs were detected. The latter deletion perturbs the HiNF-D binding site, strongly indicating that the interaction of HiNF-D with site 11, although dispensable under our in vitro conditions, is essential for the in vivo regulation of human H4 histone gene transcription.
HiNF-D Is Specific for Proliferating Cells Synthesizing Replication-dependent H4 Histone mRNAs-Human H4 histone mRNAs are transcribed throughout the HeLa S3 cell cycle with a transient 3-5-fold increase in transcription rates occurring at the Gl/S-phase boundary (10, 11). One of the mechanisms by which the cell modulates histone mRNA transcription rates during the cell cycle could involve, for instance, alterations in the DNA binding activity of HiNF-A, HiNF-C, or HiNF-D. However, the overall DNA binding activity of these factors as a function of the HeLa S3 cell cycle does not vary (Fig. 8). The factors can be detected both in early S-phase and G1-phase HeLa S3 cells (A, lanes 1-8;

C, lanes
Human histone gene expression is down-regulated when proliferating HL60 promyelocytic leukemia cells become terminally differentiated after treatment with phorbol esters such as TPA (6, 7). Down-regulation is mediated a t least partially at the transcriptional level as human H4 histone mRNA synthesis rates in isolated nuclei are reduced below the level of detection following treatment of HL60 cells with TPA (6). This drastic difference in transcription rates, in contrast to the subtle fluctuations occurring during the cell cycle, provides the basis for establishing the extent to which the interaction of the H4 histone promoter factors with the F0108 H4 histone gene promoter is related to the transcription of this gene.
Factors HiNF-C (Fig. 8B) and HiNF-A (Fig. 8 D ) are present in both proliferating and terminally differentiated HL60 cells. Most interestingly, although HiNF-D activity is clearly present in exponentially growing cells (Fig. B ) , this activity is selectively lost during differentiation. Hence, the site I1 binding protein HiNF-D is specific for proliferating cells. These results suggest a principal role for HiNF-D in the regulation of transcription of the cell cycle-dependent H4 histone genes. In support of this are in uiuo DMS protection experiments in the living cell (6) in which protection of sites I and I1 was observed in proliferating HL60 cells, but only protection of site I, and not site 11, was seen in terminally differentiated HL60 cells.

Temperature and Detergent Stability of Factor HiNF-D-
The binding site specificity of HiNF-D is very similar to that. of factor H4TF-2, isolated by Dailey et al. (41,42), which binds to a different H4 histone promoter (designated Hu4a). However, the migration rate of the HiNF-D complex in gel retardation assays is very different from the H4TF-2 complex. We have studied the temperature and detergent stability of factor HiNF-D to investigate its physical properties and to facilitate a comparison with H4TF-2. Nuclear extracts were subjected to various mild denaturing conditions and subjected to the gel retardation assay (Fig. 9). The addition of low concentrations of SDS (above 0.002%) prevented the formation of complex HiNF-D (Fig. 9B), but the addition of various concentrations (up to 1%) of non-ionic detergents such as Nonident P-40 and Triton X-100 had no effect on the binding.
When nuclear extracts were gently heated during the binding reaction (Fig. 9A), we observed a decrease in the formation of the HiNF-D complex with increasing incubation temperatures from 37 to 55 "C (lanes 1-9). In a second experiment nuclear extracts were preincubated at the same temperatures prior to the binding reaction (10-18). In this case we observed that HiNF-D remains relatively stable up to 50°C and becomes rapidly destabilized at 52 "C and higher temperatures.

Human H4 Histone
at sites I and I1 has been tested in isolated human HeLa S3 nuclei by a salt extraction procedure (44). The H4 proximal promoter factors gradually dissociate in vivo at salt concentrations between 120 and 200 mM KC1. We studied the salt dependence of HiNF-C and HiNF-D complex formation in gel-retardation assays to allow a comparison between in vivo and in vitro salt stability, as an initial measure of the relative binding affinity of these factors in vivo and in uitro. The results (Fig. 9D) show that HiNF-D binding is unstable at KC1 concentrations between 80 and 110 mM but that HiNF-C binding is relatively stable up to 150 mM KC1 (Fig. 9D). From these experiments it appears that HiNF-D binding is much more sensitive to increased ionic strength conditions than HiNF-C binding.
Factor HiNF-C Requires Divalent Cations for Site-specific Binding-We observed a significant loss of HiNF-C DNA binding activity upon dialysis of nuclear extracts against EDTA during chromatographic fractionation, and mixing of various fractions did not lead to restoration of HiNF-C activity (data not shown). A number of DNA binding proteins have been described that contain amino acid motifs capable of divalent cation binding (e.g. zinc binding proteins; 59). Therefore, we considered the possibility that HiNF-C would have a similar divalent cation requirement for its DNA binding activity and possibly its stability.
Titration experiments were performed in which the binding of HiNF-C was studied in the presence of various concentrations of metal ion chelating agents. HiNF-C activity was virtually undetectable when nuclear extracts were preincubated in the presence of 1 mM 1,lO-phenanthroline (Fig. 10A). The binding of HiNF-C partially decreased after preincubation with 10 mM EDTA but was not affected by EGTA (10 mM). These data strongly suggest that HiNF-C has a requirement for divalent cations. Pretreatment with 1,lO-phenanthroline had a gradual, inhibitory effect at a range of concentrations between 0.1 and 1.0 mM (Fig. 10B, lanes 1-8) and its effect was also observed if the preformed HiNF-C complex  0.05, 0.1, 0.2, 0.5, and 1.0 mM 1,lO-phenanthroline; lane 9, the same chelator (5.0 mM) added after the binding reaction, 1 min before loading the sample on the gel. C, Lane 1, 2 pg of UNE protein, no chelator added (control); lanes 2-9: same, but preincubated in the presence of 0.5 mM 1,lO-phenanthroline; after this, samples were diluted 10-fold and reconstituted with ZnC1, at the following concen- trations (lanes 2-9), respectively, 0, 0.01, 0.02, 0.05, 0.10, 0.20, 0  was challenged with the chelator after formation in the binding mixture (lanes 9 and 10).
The metal ion requirement of HiNF-C was further explored by reconstitution experiments in which HiNF-C was first partially inactivated by 1,lO-phenanthroline and subsequently incubated in the presence of various amounts of divalent cations (Fig. lOC). Binding activity could not be restored by adding M e or Ca2+ (data not shown), but the addition of an optimal amount of Zn2+ (approximately 100 p~) restored HiNF-C binding. This result implies that HiNF-C has a selective metal ion requirement and that Zn2+ ions are sufficient for the restoration of binding activity.
Factor HiNF-D Is 1,lO-Phenanthroline Sensitiue at Elevated Temperatures-Chelation experiments aimed at showing a divalent cation requirement for factor HiNF-D did not result in decreased binding activity under the assay conditions described above. Dailey et al. (60) have reported that H4TF-'2 under similar conditions is very sensitive to 1,lO-phenanthroline and has a divalent cation requirement (Zn2+ or Fe2+). We reasoned that the metal ion in HiNF-D might be very tightly bound and therefore protected from chelation. The possibility was investigated whether HiNF-D would display sensitivity toward chelators at elevated temperatures.
HiNF-D was preincubated with 1,lO-phenanthroline at temperatures close to the point where its DNA binding ability is thermally inactivated. The binding of HiNF-D gradually decreased when the factor was preincubated at temperatures between 48 and 55 "C (Fig. 11). No effect was observed when nuclear extracts were preincubated with EDTA or EGTA (10 mM) (data not shown). Interestingly, when 1,lO-phenanthroline was added at these elevated temperatures HiNF-D displayed sensitivity toward this chelator (range 1-10 mM) that became more pronounced as temperature increased. These results demonstrate that HiNF-D and H4TF-2 are differentially sensitive toward divalent cation chelators and suggest that the protein structures of these factors are different.

DISCUSSION
The proximal promoter of the human H4 histone gene F0108 has been shown to contain two in vivo sites of DNA/ protein interaction (sites I and 11) (44). In this report, we have demonstrated that the DNA sequences of this promoter (nt -30 to -240) form at least three different, specific DNA/ protein complexes in vitro, the complexes of HiNF-A, HiNF-C, and HiNF-D. Factor HiNF-A binds to AT-rich sequences in the distal part of site I. HiNF-A has a moderate sequence specificity as its binding in gel-retardation assays is competed

Human H4
Histone Gene Promoter Factors 15041 out by a very small amount of the nonspecific competitor poly(d1-dC) DNA (<lo0 ng) and a moderate amount of E. coli DNA (>1 pg). The detection of HiNF-A in gel-retardation assays, despite its limited sequence specificity, may be a reflection of its abundance in nuclear extracts, thus excluding a role for HiNF-A as a limiting transcription factor. HiNF-A may belong to a class of DNA binding proteins that bind to A/T-rich DNA sequences such as human high mobility group-I (61) or monkey a-protein (62) and may have a chromatin structural role as proposed for these and other high mobility group proteins. In this regard, it must be noted that HiNF-A binding sites are present in the proximal promoters of human H4, H3, and H1 histone genes (35,40) and that the chromatin structure of these promoters is subject to dynamic changes during the cell cycle (47-50).
Factor HiNF-C binds to a DNA fragment (nt -99 to -130) containing an Spl consensus binding site (54; match; 8 of 10 basepairs) present in the proximal part of site I. The DMS protection pattern of total nuclear protein in this region shows only protection of nucleotides coinciding with the Spl box. This in vitro DMS protection pattern is identical to the DMS fingerprint observed i n vivo in the F0108 H4 histone promoter (44) and very similar to DMS protections observed i n vitro for Spl binding to Spl boxes (63). The estimated DNaseI footprint of HiNF-C spans approximately 20 nt, which is in close agreement to the DNaseI footprint size of Spl. Factor HiNF-C has a selective, divalent cation requirement for DNA binding, and Zn2+ ions are sufficient in regenerating HiNF-C binding activity after treatment with the metal ion chelator 1,lO-phenanthroline (<1 mM). Factor Spl has been shown to contain cysteine-rich motifs assumed to be involved in binding of Zn2+ ions (zinc-fingers; reviewed in Ref. 59), although Spl binding was influenced by the addition of EDTA (50 mM), but not 1,lO-phenanthroline (25 mM) (64). We propose that HiNF-C is a general promoter factor similar or identical to Spl.
Factor HiNF-D binds to the distal part of i n vivo site I1 (nt -50 to -90) containing a mammalian histone-specific consensus element (5' dGGTPyPyTCAATC(N) GGTCCG). the factor can be detected in several proliferating cell lines in vitro (this work) and i n vivo (6,44), but (unlike HiNF-A and HiNF-C) cannot be found in vitro or in vivo in terminally differentiated human HL60 cells in which replication-dependent histone gene transcription is down-regulated. Previously, we have described the mosaic CCAAT box factor HiNF-B that binds to the human H1 histone promoter (35,36). Although the histone-specific element of the H4 histone gene contains a CAAT-like sequence, the mobilities of HiNF-D and HiNF-B complexes and the binding site specificities of these factors are different. The human H4 histone promoter factor H4TF-2 isolated by Dailey et al. (41,42) has a binding site specificity similar to that of HiNF-D. However, there are several differences between the two factors: 1) the electrophoretic properties of the HiNF-D and H4TF-2 complexes in gel-retardation assays under comparable conditions are significantly different; 2) under chromatographic conditions identical to those used for the purification of H4TF-2, HiNF-D binds to phosphocellulose in the presence of buffer containing 100 mM KC12 whereas H4TF-2 does not (41,42); 3) H4TF-2 is very sensitive toward the divalent cation chelator 1,lO-phenanthroline (<LO mM; 4 "C (60) and its DNA binding activity is optimal at low Mg2+ concentrations (0-1 mM) (41); in contrast, HiNF-D is 1 , l Ophenanthroline sensitive to high concentrations (1-10 mM) only at elevated temperatures (>48"C), and the binding of HiNF-D is unaffected by a range of Mg2+ concentrations (0-10 mM) (data not shown). Hence, the metal ion requirements of H4TF-2 and HiNF-D appear to be different. We conclude that although the binding site specificities of HiNF-D and H4TF-2 are very similar, the factors are distinct in many other respects.
The proximal part of site I1 contains a TATA element that is protected in vivo in conjunction with the HiNF-D site. The distance between potential HiNF-D sites in several mammalian H4 histone genes and the adjacent TATA box sequence is strictly conserved (9 basepairs). This strict spacing constraint suggests that HiNF-D binds i n vivo in conjunction with a putative TATA box factor (65) to site 11. Consistent with this hypothesis is that HiNF-D binding by itself in vitro is stable only at moderate ionic strength conditions. This hints at the possibility that HiNF-D binding i n vivo requires a stabilizing factor(s). In this regard, the fact that HiNF-D is required for accurate transcription initiation in vivo but not i n vitro could indicate that the factor is involved in recruiting a TATA box factor that may be limited i n vivo, but not i n vitro.
In conclusion, the H4-Site I binding proteins HiNF-A and HiNF-C are not proliferation specific. The functional significance of binding events at site I is evidenced by a 4-6-fold reduction in the efficiency of i n vitro histone mRNA synthesis upon deletion of this site. We propose that the interaction of HiNF-C with site I, possibly in conjunction with HiNF-A and/or other uncharacterized factors, has an auxiliary (house keeping) function in augmenting H4 histone mRNA synthesis rates and that the roles of these factors are not confined to H4 histone gene transcription. The H4-site I1 binding factor HiNF-D is proliferation specific and the binding of HiNF-D is essential for i n vivo expression of the F0108 human H4 histone gene. We postulate that the interaction of HiNF-D with site I1 performs an essential function (on/off switch) without which no transcription can occur. Because the HiNF-D binding site is conserved in a number of mammalian H4 histone genes and the DNA binding activity of the factor is down-regulated during differentiation, it is conceivable that the factor is involved in the coordinate regulation of the proliferation-specific H4 histone multigene subfamily during the cell cycle and becomes a rate-limiting factor during the coordinate shut down of histone gene transcription at the onset of terminal differentiation.