An organizational model of transcription factor binding sites for a histone promoter in D. melanogaster

---

Abstract

The Drosophila H2A-H2B histone spacer, a small region that functions as a bidirectional promoter for the gene pair, was used as a test sequence for generation of a computationally derived organizational model of transcription factor (TF) binding sites. Expression studies of the spacer revealed that it contains the necessary sequences to confer replication-dependent transcription in partially synchronized cells in culture. Informatics analysis of the spacer uncovered a number of binding sites for specific TFs, none of which had been previously associated with this particular promoter. Each of the TFs in the promoter organizational model are also known to participate in stages of fly development that are characterized by DNA replication and/or cell division, thus providing a biologically functional rationale for an association. Moreover, phylogenetic analysis of the binding sites provides evidence for evolutionary conservation of the essential features of the organizational model. The model, if correct, provides information about the molecules that couple developmental specific demands and histone gene transcription.

Key words: organizational model, transcription factor binding site, histone genes, MatInspector, TRANSFAC, promoter analysis, replication-dependent transcription, phylogenetic footprinting, RAP1, non-coding sequences

---

Introduction

There are well-defined rules for translation of sequence patterns into proteins for the coding portions of genomes. These rules make identification and functional analyses of coding sequences relatively straightforward. In contrast, the relationships between sequence patterns and functions in non-protein coding regions of the genome are far less simple, and have only recently begun to emerge, making analyses more challenging. These gaps in knowledge and the resultant difficulty of computational analysis are especially critical since non-coding sequences comprise as much as 90% of the eukaryote genome and are important to its regulation, evolution and structure. Because of the vast size and complexity of such regions, in silico studies must be a component in their analysis.

Promoter regions have emerged as the first non-coding DNA amenable to informatics approaches, since motifs that bind trans-acting regulatory factors and the resultant effects on gene regulation are becoming known. For several years it has been possible to identify putative transcription factor binding sites using informatics research tools such as MatInspector [Quandt et al., 1995] that is linked to the TRANSFAC database of transcription factors and their DNA-binding profiles [Quandt et al., 1995; Wingender et al., 2000]. MatInspector identifies shared sequence similarities between the consensus binding sites for transcription factors that are annotated in the TRANSFAC database and motifs within a query promoter sequence. Using the reported motifs, DNA strand orientation, location, and cognate transcription factor, an organizational model for the promoter can be created.

We used a small spacer region from the histone genes of Drosophila melanogaster as a model for the creation and validation of an organizational model for transcription factor binding sites. This spacer functions as the bi-directional promoter for the divergently transcribed H2A and H2B gene pair and is ideal for promoter modeling studies. First, its extremely short length (232 nucleotides) is clearly delineated by flanking coding regions [Baldo et al., 1999], providing a small and unambiguous target for computational analyses. This structural arrangement eliminates the difficulties inherent with most other genes where upstream regulatory regions cannot be clearly defined [Werner, 1999]. Second, none of the transcription factors (TFs) known to regulate histone promoters in other organisms have been identified in flies, thus modeling provides a roster of TFs that can be used to guide future empirical studies within this gene family.

Replication-dependent transcription of histone genes is well established. However, it is not known (although often assumed) whether individual genes in a repeat possess an independently acting set of regulators, or how the embryonic type genes are regulated throughout development. Through transient expression studies, we show that the small H2A-H2B spacer possesses the necessary information for replication-dependent transcription of its flanking gene. By an in silico analysis, we demonstrate the presence of putative binding sites for a number of transcription factors that were not previously implicated in the regulation of histone genes. This roster of TFs is functionally assessed and shown to be relevant because it links known, development-specific, histone gene transcription with the proper embryonic [Strausbaugh and Weinberg, 1982] and ovarian [Di Nocera and Dawid, 1983] stages of development. Evolutionary support for the model has been provided by motif-based phylogenetic footprint analysis, since H2A-H2B spacer sequence is available for a number of species.

---

Methods

Demonstration of regulation by the H2A-H2B spacer

The histone genes of D. melanogaster and its close relatives are arranged in approximately 100 tandem repeats with the gene order and transcriptional directions: H1> - <H2B - H2A> - <H4 - H3>. An H2A gene in a cloned, embryonic-type 4.8 kb repeat [Strausbaugh and Weinberg, 1982] was tagged at its XhoI site by the ligation of an in-frame 20 base pair oligonucleotide (5'-ACCATGATTCAGCTGCTGGC-3') with XhoI overhangs at each end. Construct pDm4.8: H2AO contained the tagged gene in the entire 4.8 kb repeat cloned into pBR327; construct pDm4.8:H2AO Del contains only the H2A and H2B gene pair and the associated intergenic spacer cloned into pUC13.

Drosophila SC-2 cells were transformed with calcium-phosphate DNA complexes of each construct for transient expression assays [Di Nocera and Dawid 1983]. Transfected cells were left undisturbed for 22-48 hours prior to aphidicolin arrest to achieve a partial synchronization [Dalton et al., 1986, LaBella et al., 1988]. Following release from the aphidicolin block, cells were labeled with thymidine to monitor the course of DNA synthesis. RNA was extracted for primer extension analysis, using the oligomer tag sequences as primers. This less disruptive, but more cumbersome, strategy for tagging and expression analysis was chosen (over use of a fusion reporter-protein construct) because histone gene regulation has been shown in other systems to be affected by a number of features, including 3'-sequences, stability, turnover, and sequences internal to the histone coding regions.

Identification of TF binding sites using TRANSFAC

Histone H2A-H2B intergenic spacer sequences were downloaded from NCBI Entrez-Nucleotide (http://www.ncbi.nlm.nih.gov/) database for the following Drosophila species: D. melanogaster (AJ224808), spacer length 232 nucleotides; D. mauritiana (AJ224806), spacer length 227 nucleotides; D. simulans (AJ224807), spacer length 230 nucleotides; D. yakuba (AJ224809), spacer length 231 nucleotides; and D. hydei (X52576), spacer length 252 nucleotides. Each spacer sequence was used as the query for transcription factor binding site (TFBS) searches using MatInspector v7.2 (http://www.genomatix.de/cgi-bin/matinspector/matinspector.pl) that contains a link to the library of transcription factors contained with the TRANSFAC v6.0 database (http://www.biobase.de). Searches were performed in the "Insect" matrix with core and matrix similarity scores, 0.75 and 0.85 respectively. World Wide Web links are provided in both MatInspector and TRANSFAC programs for obtaining more detailed descriptions of the transcription factors and their functions.

Creation of Organizational Models

The TRANSFAC/MatInspector-identified binding sites (identity of the cognate transcription factors, nucleotide position number, and DNA strand orientation) were used to create organizational models for the H2A-H2B promoters. Arbitrary shapes and colors represent putative binding sites, arranged by sequential order and scaled to reflect reported distances. Shapes positioned above the line represent binding sites that were reported on the plus strand, while shapes below represent those on the minus strand. In addition to the sites identified using MatInspector, a highly conserved [Baldo et al., 1999] centrally located binding site [position 112-122 (+) in all spacers; except D. hydei, position 116-125 (+)] is depicted in the organizational model. Although the cognate transcription factor for this binding site remains unknown, its consensus binding site is highly conserved with that of the yeast transcription factor, Repressor Activator Protein 1 (RAP1) [Baldo, 1998].

Phylogenetic footprint analysis

Phylogenetic footprinting traditionally involves comparative analyses between orthologous and paralogous gene pairs to identify conserved and divergent nucleotides within DNA sequences. We modified this approach to compare combinatorial sets of transcription factors, which we call motif-based phylogenetic footprint analysis. Although another program at the Genomatix website, GEMS Launcher-FrameWorker is designed to perform a similar type of comparison, manual comparisons allow the addition of putative binding sites that may not be in the MatInspector database. Additionally, manual analysis allows more plasticity in evaluating the overall observed conservation among the promoters, as compared to adhering to precise rules for computational algorithms. The D. melanogaster organizational model is the reference to which others were compared for the identity, sequential order, and relative distance between cognate transcription factors.

---

Results

The H2A-H2B intergenic spacer confers replication-dependent transcription

To evaluate the relevance of an informatics-generated model of trans-acting regulators, it is necessary to know the functions performed by the sequence under study. A major regulatory aspect of the transcription of "embryonic-type" histone genes (the repeats under study herein) is coordination with DNA replication; in the case of flies, this linkage plays a role in the high levels of transcription in the early embryo and during metamorphosis, as well as during the endomitotic stages of nurse cell maturation in the ovary. Later ovarian histone gene expression is not cell-cycle dependent, but may be replication dependent with ovary-specific amplification events.

While it was well established that the major repeat types (4.8 and 5.0 kb) contain the replication-dependent genes, it had not been demonstrated that the small spacer, alone, was sufficient to couple replication and transcription. To assess the role of the small H2A-H2B intergenic spacer, a tagged H2A gene was created and expressed transiently in partially synchronized cells in culture. Figure 1, depicts levels of thymidine incorporation in transfected cells released from aphidicolin treatment. While we do not argue that the block and subsequent release accurately reflects "normal" cell-cycle events, this strategy does provide an effective way to create a population of cells undergoing replication. Figure 2 illustrates the time course of transcription of the tagged H2A gene; its transcription mirrors incorporation of thymidine, a standard marker for DNA replication. Figure 3 depicts the sizes of the RNA fragments generated by primer extension. The major fragment at approximately 300 bp is consistent with models for the size of the transcript in vivo. These findings are reproducible and consistent for constructs that contain either the full repeat or only the divergently transcribed H2A-H2B gene pair, supporting the long standing assumption that individual histone genes in flies have independent sets of regulatory sequences for replication-dependent gene expression.

TF Binding Site Search of Histone H2A-H2B intergenic spacer

To identify putative transcription factors, the D. melanogaster H2A-H2B intergenic spacer was used as the query sequence for TRANSFAC/ MatInspector searches; results are displayed in Table 1. Ten different transcription factors were identified, the majority having multiple binding sites. We require two conditions to consider a binding site to be above chance occurrences: it must share sequence similarity with a previously characterized consensus binding site for a transcription factor, and this similarity must equal or exceed the suggested MatInspector default parameter (based upon experimental data and defined as the percentage of divergence that still allows binding of the cognate transcription factor) [Quandt et al., 1995]. Each of the putative binding sites (Table 1) received a matrix similarity rating that equals or exceeds our chosen cutoff score of 0.85.

Table 2, top panel, depicts the organizational model for transcription factor binding sites in the H2A-H2B intergenic spacer of D. melanogaster. In the diagram for the organizational model, the endpoint arrows indicate the start sites for the H2A and H2B genes. Arbitrary shapes were selected to represent the putative binding sites. The position numbers and strand orientations were used to arrange the putative binding sites in the correct geometry.

Functional Validation of Putative Transcription Factors

The roster of potential transcriptional regulators generated by TRANSFAC/ MatInspector consisted of proteins that were neither previously proposed, nor empirically identified, to associate with fly histone gene promoters. Our results could be revealing new functions for combinatorial regulatory elements, or could reflect chance associations with the histone promoter. Two functional considerations help assess these alternatives. First, do the postulated transcription factors act in temporal or spatial patterns consistent with known expression of the histone genes? Second, is there evidence from other genes for combinatorial interactions of any subsets of these same factors? The second assessment is based on the lesser likelihood for the co-occurrence of non-interacting transcription factors at the same promoter than for interacting ones. This principle is documented by examples of families of promoters that are regulated by specific "sets" of transcription factors, designated modules [Arnone and Davidson, 1997; Werner, 1999]. Members of a given module tend to have either direct associations with one another, or participate in a common genetic pathway. Table 2 provides selective profiles of the functional roles for the transcription factors postulated to associate with the H2A-H2B spacer. All ten of the transcription factors are known to interact through complex genetic pathways, and are functional during many of the same developmental stages that embryonic type histone gene transcription is induced.

Transcription bursts from the embryonic type histone genes are associated with phases of rapid cell proliferation in early development. Seven putative transcription factors (HB, CAD, DL, CROC, DFD, KNI and TLL) are expressed in a sequential fashion during embryogenesis and are classified as gap genes. Hunchback (HB) is among the first zygotic genes to be transcribed, forming a gradient that establishes anterior/posterior boundaries after cycle twelve in embryogenesis [Tautz, 1988; Hülskamp et al., 1994]. Cycle twelve, which is 7 - 9 hours after fertilization [Campos-Ortega and Hartenstein, 1985], is also the earliest time-point at which basal levels of zygotic histone mRNA can be detected [Anderson and Lengyel, 1980; Tautz, 1988]. Also during this time-point, maternally deposited caudal (CAD) RNA and proteins are localized in an anteroposterior gradient, while zygotic CAD RNA and proteins are localized in the abdominal segments [Mlodzik and Gehring, 1987]. At the midblastula transition, early in cycle fourteen (10 -11 hours after fertilization), dorsal (DL), under the influence of the Epidermal Growth Factor Receptor (EGF-R) pathway, is transported into the cell nuclei of the blastoderm [Campos-Ortega and Hartenstein, 1985; Roth et al., 1989; Rushlow et al., 1989; Stewart, 1989]. In the later part of cycle fourteen, zygotic histone gene expression increases to its highest level [Lanzotti et al., 2002]. Following cycle fourteen the four remaining transcription factors, crocodile - CROC [Häcker et al., 1995], deformed - DFD [Regulski et al., 1991], knirps - KNI [Lunde et al., 1998], and tailless - TLL [Rudolph et al., 1997] participate in the genetic pathway that later lead to the development of structures within the adult fly. Although there has been no prior direct linkage between gap gene functions and histone gene regulation, the fact that all are expressed in the correct time frame, and have prior combinatorial associations with each other, makes a role in this gene family plausible.

Histone genes are also actively expressed during morphogenesis, when the adult is constructed from imaginal disk cells by standard proliferation, as well as from cells with polyploid and polytene replications patterns. Broad Complex (BRCZ4) and Ecdysone Induced Protein (E74A) are both members of the Ecdysone Response pathway (EcR) where they have well-characterized and essential functions during the metamorphic period in pupal development [von Kalm et al., 1994].

As mentioned previously, embryonic type histone genes are activated in complex, and unknown, ways at several points during oogenesis. The MatInspector program's identification of genes involved in oogenesis is especially intriguing. Oocyte maturation takes three days and is divided into 14 stages. In stages 1-10, nurse cells undergo polyploidization by a series of endomitotic replications; histone gene expression is probably replication-dependent. However, the majority of histone mRNA accumulation occurs after stage 10 (induction in nurse cells occurs by stage 11), well past the main period of nurse cell replication. The stored histone transcripts are sequestered in the developing oocyte to support the high demands of early embryonic cleavages. Nothing is known about the regulators that contribute to this bimodal fashion of regulation: coupled to replication prior to stage 11, independence of it afterwards [Ambrosio and Schedl, 1985; Ruddell and Jacobs-Lorena, 1985]. Three of the ten transcription factors (BRCZ4, E74A, and CF2II) are known to play roles during oogenesis. During mid-oogenesis, BRCZ4 and E74A are regulated by a second pathway, the EGF-R signaling pathway [Buszczak et al., 1999]. Chorion Factor 2II (CF2II) is involved in the establishment of dorsoventral polarity in the oocyte, and participates in the network of pathways between EcR and EGF-R. CF2II, which specifically represses the rhomboid gene, has the ability to initiate a cascade of gene-protein interactions, resulting in the persistent activation of the EGF-R pathway. Thus, interactions among the three transcription factors is establish because CF2II functions within the EGF-R pathway that is responsible for the regulation of both BRCZ4 and E74A genes during mid-oogenesis [Hsu et al., 1996; Deng and Bownes, 1997].

Motif-based Phylogenetic Footprint Predictions

An organizational model is a footprint that reveals three major features: 1) identity of the cognate transcription factors, 2) sequential order of the binding sites, and 3) relative distances between them [Frech et al., 1998; Werner, 1999]. Features required for proper histone gene expression are more likely to be conserved than those due to chance occurrences, especially given the fact that sequence divergence exists in this spacer [Baldo et al., 1999]. We have used a phylogenetic framework to assess the likelihood that the TFs in the organizational model are important. TRANSFAC/ MatInspector analysis was performed using the H2A-H2B intergenic spacer for the following members of the Sophophora radiation: D. melanogaster, D. mauritiana, D. simulans, D. yakuba, and D. hydei. Their aligned organizational models are shown in Figure 4, in increasing evolutionary distance from D. melanogaster. D. mauritiana is most closely related, having membership within the same species subgroup (the Melanogaster species subgroup), sharing 94% sequence identity and ~ 0.8 million years (Myr) diverged. In contrast, D. hydei belongs to the Repleta species group, which is the most distant lineage from the Melanogaster species group within the Sophophora radiation [Menotti-Raymond et al., 1991; González et al., 2002]. These lineages diverged 40-62 Mya [González et al., 2002], and only share 46% sequence identity between their H2A-H2B spacers.

We chose the D. melanogaster model as the reference for comparing the four other models. Alignments of the organizational models reveal similar patterning among all species. Each postulated transcription factor observed in the D. melanogaster model is also identified in the other species, supporting a high level of conservation of regulatory components. In addition to the conservation of components, the conservation of geometry of the binding sites is striking. All species contain a centrally located binding site for RAP1 (magenta star), which appears to partition the other TFBS into two clusters, one proximal to the H2A start site and the other proximal to the H2B start site. The same TFs occur within respective clusters among the different models with nearly identical orders and distances. We acknowledge that in this phylogenetic range, similar models would be expected among closely related species. Still it should be noted that binding sites may be lost or gained as a result of single nucleotide changes, thus small percent differences for overall sequence identity among the spacers could still allow notable differences to exist among the models. Thus the similarity in identity and patterning of TFBS between the D. melanogaster and D. hydei models, despite substantial sequence divergence (46% identity) provides a plausible argument for an association between the cognate TFs and this histone promoter.

---

Discussion

Among the longer-term goals of genome projects is obtaining the ability to identify components of regulation and how they function to achieve proper expression. A large component of this goal involves the recognition of transcription factor binding sites. Although there is a general understanding of the mechanics involved in basal promoter function, there are relatively few genes for which the complete set of required transcription factors are known. Using experimental approaches to obtain this information often requires laborious and time-consuming procedures. Resultantly, computational methods for identifying TFBS have become an increasingly active area of genome research because results from in silico analyses can be used to guide empirical studies. Computational programs such as CONSENSUS [Hertz and Stormo, 1999], MEME [Bailey and Elkan, 1994], and PROJECTION [Buhler and Tompa, 2002], that use position weight matrices for the identification of individual TFBS within local multiple sequence alignments, have been available for several years now. More recently, combinatorial programs such as rVISTA [Loots and Ovcharenko, 2004], and the FrameWorker program by Genomatix (http://www.genomatix.de) have become available, that combines pattern recognition with comparative sequence analysis of TFBS (sequential order and spatial relations) within promoters. These programs are essentially automated methods for performing phylogenetic footprint analyses.

Instead, we manually performed phylogenetic footprint analyses using the organizational models for the H2A-H2B histone promoters of Sophophora species that were created using the MatInspector v7.2 program. We found manual analyses to be advantageous because some plasticity was needed in evaluating the level of conservation among the TFBS patterns when patterning was similar but not identical among all the models. This is not always possible when using computational tools whose algorithms adhere to definitive rules for pattern recognition. For our analyses we also used a short promoter, of well-defined length, as a test sequence, however footprinting is not limited to short promoters. Furthermore, a defined length for a promoter does not guarantee its complete set of TFBS can be obtained. Still, we acknowledge that an organizational model created for a large promoter, whose exact length is not known, has an increased probability for yielding an incomplete roster of TFBS. In some instances (e.g. Drosophila histone promoters) an incomplete roster of TFBS is still of value because it provides more information than what is currently known. We also acknowledge that phylogenetic footprinting is restricted to genes that have promoter sequence available in homologous genes from several distantly related species. Given that genome databases are available for most of the commonly studied organisms and these databases are undergoing extensive annotations, it is expected that in the near future phylogenetic footprinting will be applicable to most promoters.

We believe the findings we obtained through this approach are significant because no other study has reported putative binding sites within the histone H2A-H2B promoter in Drosophila. Prior to this study, none of the identified transcription factors had been reported as having an association with histone gene regulation. An analysis of the transcription factors revealed that they perform principle functions during oogenesis and/or other periods in fly development when histone transcription levels are high. Furthermore, subsets of the ten transcription factors are known to act in concert at other gene promoters, lending credibility to the same associations within the histone promoter.

The motif-based phylogenetic footprints for the promoters provided additional evidence in support of the validity of the organizational models. An alignment of the organizational models for the widely divergent species (30-62 Myr) revealed shared identity among the TFs. This shared identity and organization occurred among multiple binding sites between five different organizational models, which decreases the likelihood that the binding sites are "chance hits" made by the MatInspector program. The clustered appearance of the TFBS, proximal to each flanking gene, could suggest that the H2A gene may possess a separate but similar set of regulatory elements from the H2B gene. It could also suggest that from this single promoter, separate, yet coordinated regulation of the flanking genes occurs, which is consistent with the known bi-directional regulation of these genes. Thus the binding sites and their cognate transcription factors become plausible candidates for further study.

Given the importance of histone proteins and the proper expression of histone genes, continued in silico analysis of this promoters, coupled with a phylogenetic footprint from additional widely diverged species may prove to be the ideal prelude to forthcoming experimental methods that are required for establishing a definitive associations among TFBS, their cognate transcription factors, and histone promoters.