In the work reported here we have analyzed the role of the GAGA factor [encoded by the Trithorax-like (Trl) gene] in the enhancer-blocking activity of Frontabdominal-7 (Fab-7), a domain boundary element from the Drosophila melanogaster bithorax complex (BX-C). One of the three nuclease hypersensitive sites in the Fab-7 boundary, HS1, contains multiple consensus-binding sequences for the GAGA factor, a protein known to be involved in the formation and/or maintenance of nucleosome-free regions of chromatin. GAGA protein has been shown to localize to the Fab-7 boundary in vivo, and we show that it recognizes sequences from HS1 in vitro. Using two different transgene assays we demonstrate that GAGA-factor-binding sites are necessary but not sufficient for full Fab-7 enhancer-blocking activity. We show that distinct GAGA sites are required for different enhancer-blocking activities at different stages of development. We also show that the enhancer-blocking activity of the endogenous Fab-7 boundary is sensitive to mutations in the gene encoding the GAGA factor Trithorax-like.
EUKARYOTIC chromosomes are subdivided into functionally and structurally autonomous chromatin domains. This was first recognized more than a half century ago in cytological studies of lampbrush chromosomes in amphibian oocytes and polytene chromosomes in insects (Alfert 1956; Gall 1956; Gall and Callan 1962; Ritossa 1962). More direct evidence for the subdivision of eukaryotic chromosomes into discrete domains has recently come from a combination of genetic, molecular, and biochemical experiments. These studies demonstrate that domains correspond to units of independent genetic activity, and that each domain has its own distinct nucleoprotein composition and chromatin structure depending upon the developmental stage, tissue, or cell type and whether the domain is active or inactive (Kellum and Elgin 1998; Gerasimova and Corces 1999; Udvardy 1999; Grewal 2000; West et al. 2002). Critical to ensuring the autonomy of each chromatin domain are special cis-acting elements called boundaries or insulators. While domain boundaries were first discovered in Drosophila (Welshons and Keppy 1975; Udvardy et al. 1985; Gyurkovics et al. 1990; Holdridge and Dorsett 1991; Kellum and Schedl 1991, 1992; Geyer and Corces 1992), they have since been identified in organisms ranging from yeast to humans (Gerasimova and Corces 2001; West et al. 2002). These elements define the limits of chromosomal domains and function to establish independent units of gene activity, insulating genes or regulatory elements within a domain from regulatory elements located in adjacent domains.
One of the better-characterized boundaries in Drosophila is the Fab-7 element from the bithorax complex (BX-C). BX-C contains three homeotic genes, Ultrabithorax (Ubx), abdominal-A (abd-A), and Abdominal-B (Abd-B), which specify segment identity in the posterior parasegments (PS) 5–14 of the fly. The parasegment-specific expression of these three homeotic genes is controlled by an ∼300-kb cis-regulatory region that is subdivided into nine parasegment-specific cis-regulatory domains: abx/bx, bxd/pbx, and iab2–9 (Lewis 1978; Sanchez-Herrero et al. 1985; Karch et al. 1985; Casanova and White 1987; Duncan 1987; Celniker et al. 1990; Sanchez-Herrero 1991). Each regulatory domain directs the appropriate parasegment-specific pattern of expression of one of the BX-C homeotic genes and the domains are sequentially activated, going from anterior to posterior parasegments. For example, the iab6 cis-regulatory domain is activated in PS11 where it promotes an Abd-B expression pattern that confers PS11 identity. In contrast, the cis-regulatory domain immediately distal to iab-6 in BX-C, iab-7, is silenced in PS11. In the next posterior parasegment, PS12, iab-7 is activated and in this parasegment it, not iab-6, controls Abd-B expression. Mutations that disrupt the activity of any of the nine PS-specific regulatory regions result in a phenotypic transformation of the corresponding parasegment into the parasegment immediately anterior (Karch et al. 1985). For example, iab-7Sz, a deletion within the PS12 cis-regulatory domain iab-7 results in the transformation of PS12 into a duplicate copy of PS11. This loss-of-function phenotype arises because Abd-B expression in PS12 is driven by iab-6 in the absence of a functional iab-7.
The regulation of the BX-C homeotic genes can be divided into two phases: initiation and maintenance. During the initiation phase, gap and pair-rule genes select parasegmental identity along the anterior-posterior axis of the embryo by interacting with target sequences in each BX-C cis-regulatory domain and either activating or silencing the domain (Qian et al. 1991; Muller and Bienz 1992; Shimell et al. 1994). Since the products of the segmentation genes are present only transiently in the early embryo, the pattern of homeotic gene expression that they establish must be maintained by another mechanism during the remainder of development. The maintenance mechanism depends upon the formation and propagation of active or silenced chromatin states and is mediated by proteins in the trithorax and Polycomb (Pc-G) groups. As with the segmentation genes, there are elements in each cis-regulatory domain that respond to trithorax or Polycomb proteins called TREs and PREs, respectively.
The Fab-7 boundary is located between iab-6 and iab-7 and is essential for ensuring the autonomous action of these two cis-regulatory domains. It was initially identified on the basis of the unusual dominant gain-of-function phenotype of a small deletion called Fab-71 (Gyurkovics et al. 1990). In the Fab-71 deletion, PS11 is transformed into a duplicate copy of PS12. Genetic analysis of the Fab-71 mutation showed that the iab-6 and iab-7 cis-regulatory domains are fused into a single domain and that this leads to the ectopic activation of iab-7 in PS11 by positive regulatory elements in iab-6. Subsequent studies revealed that Fab-71 lacks not only the Fab-7 boundary, but also the nearby iab-7 PRE, which functions in the silencing of the iab-7 domain in parasegments anterior to PS12 (Galloni et al. 1993; Mihaly et al. 1997). Deletions that remove only the Fab-7 boundary differ from Fab-71 in that they have a more complex mixture of gain- and loss-of-function phenotypes in PS11. As in Fab-71, the gain-of-function phenotype arises from the ectopic activation of iab-7 by positive elements in iab-6. The loss-of-function phenotype of the boundary deletions is due to the ectopic silencing of iab-6 by negative elements in iab-7. Since the iab-7 PRE is still present in these mutants, it is able to maintain iab-6 in the silenced state. Besides Fab-7, two other BX-C boundaries, Mcp and Fab-8, have been identified in the Abd-B region of the complex (Mihaly et al. 1998; Barges et al. 2000). Like Fab-7, these elements appear to be important for ensuring the functional autonomy of adjacent cis-regulatory domains.
The chromatin structure of the Fab-7 (see Figure 1) region contains three prominent nuclease hypersensitive sites (HS1, HS2, and HS3) plus one minor hypersensitive site (Karch et al. 1994). These hypersensitive sites range in size from nearly 400 bp (HS1) to ∼140 bp (the minor hypersensitive site) in length and are hypersensitive to both micrococcal nuclease and DNase I. They are separated from each other by nuclease-resistant sequences that are approximately the size of a nucleosome. Deletion analysis within the context of BX-C as well as various transgene assays have shown that the distal hypersensitive region HS3 corresponds to the iab-7 PRE, while the Fab-7 boundary spans an ∼1.2-kb DNA segment extending from the minor hypersensitive site through HS1 and HS2 (Galloni et al. 1993; Karch et al. 1994; Hagstrom et al. 1996, 1997; Zhou et al. 1996; Mihaly et al. 1997; Mishra et al. 2001). To identify trans-acting factors, which might be important for Fab-7 boundary activity, we examined the sequence of this 1.2-kb region for consensus-binding sites of known transcription factors. This search revealed the presence of nine consensus-binding sites for the GAGA factor (Figure 1). Supporting the idea that this array of GAGA-binding sites may be important for the boundary function of Fab-7, chromatin immunoprecipitation experiments demonstrate that the GAGA factor is associated with Fab-7 sequences in vivo (Strutt et al. 1997).
The GAGA factor was originally identified in in vitro transcription assays as an activator (Biggin and Tjian 1988; Soeller et al. 1988) and GAGA-binding sites are found in many fly promoters and enhancers. Subsequent studies have shown that the GAGA factor functions to antagonize the repressive effects of chromatin by promoting the formation of nucleosome-free regions over promoters and other regulatory elements (Kerrigan et al. 1991; Lu et al. 1993; Tsukiyama et al. 1994; Tsukiyama and Wu 1995; Wall et al. 1995; Leibovitch et al. 2002) and that it participates in a variety of other sorts of chromatin-based regulatory mechanisms in addition to facilitating transcription. Thus, the GAGA factor is required for chromosome segregation/centromeric heterochromatin function in the early embryo and for Pc-G-mediated silencing by the iab-7 PRE as well as other BX-C PREs (Bhat et al. 1996; Hagstrom et al. 1997; Horard et al. 2000; Mishra et al. 2001). It is also thought to play a role in the insulating activity of the eve promoter (Ohtsuki and Levine 1998) and of a boundary element in the Antennapedia complex (Belozerov et al. 2003). The GAGA factor is encoded by the Trithorax-like gene (Farkas et al. 1994) and there are two major isoforms of 519 aa and 581 aa (Soeller et al. 1993). These two isoforms share the same N-terminal domain, which includes a BTB/POZ domain and a DNA-binding domain, but differ in their glutamine-rich C termini. The N-terminal BTB/POZ domain is thought to mediate heterotypic and homotypic protein:protein interactions, while the C-terminal domain is thought to mediate homotypic protein:protein interactions (Katsani et al. 1999; Wilkins and Lis 1999; Espinas et al. 2000; Vaquero et al. 2000; Greenberg and Schedl 2001; Pointud et al. 2001; Pagans et al. 2002; Faucheux et al. 2003; Mishra et al. 2003).
In the studies reported here we have used enhancer-blocking transgene assays to test the functional significance of the consensus GAGA-binding sites in the Fab-7 element. We show that these GAGA-binding sites are required for full Fab-7 boundary activity. We have also found that the GAGA-binding sites are not functionally equivalent and that some of the sites seem to be more important for boundary function at one stage of development than at another. We also show that the boundary function of the Fab-7 element in BX-C requires the GAGA factor gene Trithorax-like (Trl).
MATERIALS AND METHODS
Generation of GAGA site point mutations within the 1.2-kb Fab-7 boundary:
Mutations in consensus GAGA-binding sites were introduced using a PCR-based strategy. All fragments generated by a first round of PCR were amplified from a 3.35-kb HindIII-to-XbaI Fab-7 fragment inserted into BlueScript as the template. The primers that were used are listed as follows:
KH4: (contains SpeI site) CCCTTTCTGCACTAGTTGTGCTT
CCCCG KH5: (contains SpeI site) CGGGGAAGCACAACTAGTGCAG
AAAGGG KH6: CAATACTCTTTCCAATAAACTTTGCTTATATTTAC
KH8: (contains XbaI site) CTCTTATCACGTCTAGATTAATT
KH9: (contains XbaI site) CACATAGATAAATTAATCTAGAC
KH12: (contains NotI site) CGAAACTCGCGGCCGCGAAAAACTAGAG.
Primers KH2–KH11 contain point mutations in consensus GAGA-binding sites. These point mutations are indicated in boldface type. KH4, KH5, KH8, KH9, and KH12 contain restriction enzyme sites, which are indicated by italic type. All PCR fragments were confirmed by sequencing.
Fab-7 fragment of 1.2 kb with mutations in GAGA sites 1–5:
To make a 1.2-kb fragment with point mutations in GAGA sites 1–5, six PCR fragments were amplified using the following primer pairs: KH1/KH2, KH3/KH4, KH5/KH6, KH7/KH8, KH9/KH10, and KH11/KH12. Pairs of adjacent and overlapping PCR fragments were combined for a second round of PCR amplification as follows. The KH1/KH2 and KH3/KH4 products were combined with primers KH1 and KH4 to generate a larger KH1/KH4 fragment, which was subsequently subcloned into the PstI/SpeI sites of Bluescript. The KH5/KH6 and KH7/KH8 products were combined with primers KH5 and KH8 to generate a larger KH5/KH8 fragment, which was subsequently subcloned into the SpeI/XbaI sites of Bluescript. The KH9/KH10 and KH11/KH12 products were combined with primers KH9 and KH12 to generate a larger KH9/KH12 fragment, which was subsequently subcloned into the XbaI/NotI sites of Bluescript. The PstI-SpeI, SpeI-XbaI, and XbaI-NotI fragments were combined to produce a continuous 1.2-kb fragment extending from PstI to NotI within the Bluescript vector.
Fab-7 fragment of 1.2 kb with mutations in GAGA sites 1–2:
To make a 1.2-kb fragment with point mutations in GAGA sites 1–2, three PCR fragments were amplified using the following primer pairs: KH1/KH2, KH3/KH4 and KH5/KH12. The KH1/KH2 and KH3/KH4 products were combined with primers KH1 and KH4 for a second round of PCR amplification to generate a larger KH1/KH4 fragment, which was subsequently subcloned into the PstI/SpeI sites of Bluescript. The KH5/KH12 product was subcloned into the SpeI/NotI sites of Bluescript. The PstI-SpeI and SpeI-NotI fragments were combined to produce a continuous 1.2-kb fragment extending from PstI to NotI within the Bluescript vector.
Fab-7 fragment of 1.2 kb with mutations in GAGA sites 3–4:
To make a 1.2-kb fragment with point mutations in GAGA sites 3–4, three PCR fragments were amplified using the following primer pairs: KH1/KH6, KH7/KH8, and KH9/KH12. The KH1/KH6 and KH7/KH8 products were combined with primers KH1 and KH8 for a second round of PCR amplification to generate a larger KH1/KH8 fragment, which was subsequently subcloned into the PstI/XbaI sites of Bluescript. The KH9/KH12 product was subcloned into the XbaI/NotI sites of Bluescript. The PstI-XbaI and XbaI-NotI fragments were combined to produce a continuous 1.2-kb fragment extending from PstI to NotI within the Bluescript vector.
Fab-7 control fragment of 1.2 kb:
To make a 1.2-kb Fab-7 control fragment (without mutations), a single round of PCR was performed using primers KH1 and KH12. The PCR product was cloned into the PstI/NotI sites of Bluescript.
Cloning of 1.2-kb mutated and nonmutated Fab-7 fragments into the wEN:mini-white and ftz:hsp70-LacZ vectors:
XhoI-NotI fragments of 1.2 kb with and without mutations in GAGA-binding sites were excised from Bluescript and inserted into the white-enhancer:mini-white vector (XN vector; Hagstrom et al. 1997) between the white enhancer and the mini-white gene. These XhoI-NotI fragments were also inserted into the ftz enhancer:hsp70/LacZ vector (pCfhL vector; Hagstrom et al. 1996) between the UPS/NE enhancers of fushi-tarazu (ftz) and the hsp70 promoter.
Generation of GAGA site multimer in the ftz:hsp70/LacZ vector:
Two single-stranded oligos, GAGAEXHO (TCGAGAGAATTCGGCTCTCTTCG) and GAGAESAL (TCGACGAAGAGAGCCGAATTCTC), were annealed to generate a double-stranded oligo with XhoI and SalI cohesive ends. Each monomeric double-stranded oligo contains two consensus-binding sites for the GAGA factor (indicated by boldface type) and contains one EcoRI restriction site (indicated by italics). Double-stranded oligos were multimerized using T4 DNA ligase in the presence of XhoI and SalI restriction enzymes to generate tandem arrays. The products of the ligation reaction were cloned into the SalI site of the Bluescript vector. Bluescript isolates containing four tandem copies of the GAGA oligo were selected. The XhoI/NotI fragment was excised from Bluescript and inserted into the ftz enhancer:hsp70/LacZ vector (pCfhL vector; Hagstrom et al. 1996) between the UPS/NE enhancers of ftz and the hsp70 promoter.
Preparation of DNA fragments for binding studies:
DNA fragments were generated by PCR using PfU polymerase. Either BS + 1.2-kb Fab-7 or BS + 1.2-kb Fab-7 with mutations in GAGA sites 1–5 was used as template DNA (see above). The primers that were used are:
Primers SES23 and SES24, SES29 and SES34, and SES33 and SES30 were used to generate fragments F12, F34, and F56, respectively. To generate biotinylated fragments, SES23, SES29, and SES33 were biotinylated prior to PCR amplification. All PCR products were gel purified.
The nuclear extract used for in vitro binding experiments was prepared from 0- to 16-hr embryos following the published procedure (Han et al. 1993) except that a multiple cushion centrifugation was used to isolate nuclei (Solano et al. 2003) prior to extraction of proteins. For gel shift, radiolabeled DNA (∼1 fmol) was incubated with nuclear extract for 30 min in 25 mm HEPES, pH 7.6, 100 mm NaCl, 1 mm DTT, 0.1 mm PMSF, 10% glycerol, 10 μg tRNA, and 0.5 μg poly(dI.dC) at room temperature. In addition, specific cold competitor DNA was used in the amount indicated in the legend of Figure 2. The mixture was then separated on 4% PAGE (80:1 of acryl amide:bis acryl amide) containing 2.5% glycerol. The gel was autoradiographed after drying.
Affinity matrix binding assay:
To prepare DNA affinity matrix for binding experiments, we amplified different subfragments of the Fab-7 boundary with one common biotinylated primer and the other primer of varying sequence. Control DNA matrix, 278 bp long, is a part of the minor HS site and has no consensus GAGA-factor-binding sites. Matrix HS1/1–2, 529 bp long, has the sequences to include GAGA sites 1 and 2 of the HS1. Matrix HS1/1–6, 834 bp long, includes all GAGA sites of HS1. The Ubx matrix is a 270-bp DNA from the Ubx promoter, −251 to +19 region. The other two matrices are from the bxd PRE region: bxd2 is a 441-bp fragment between BamHI and EcoRI sites (Horard et al. 2000) and bxd1 is a 289-bp subfragment truncated toward the EcoRI end. All PCR products were gel purified and bound to streptavidine-coated magnetic beads as per vendor's prescriptions. The matrix was incubated with nuclear extract for 30 min at room temperature in 25 mm HEPES, pH 7.6, 100 mm NaCl, 1 mm DTT, 0.1 mm PMSF, 10% glycerol, 10 μg tRNA, and 0.5 μg poly(dI.dC). Beads were separated from the unbound proteins and washed with 25 mm HEPES, pH 7.6, 1 mm DTT, 0.1% NP40 containing 0.1 KCl. The bound proteins were eluted in the same buffer but contained 0.4 m KCl. The samples were fractionated on 10% SDS PAGE and processed for Western analysis using anti-GAGA antibody raised against different isoforms (a kind gift from V. Pirrotta): see legend to Figure 2.
For each P-element construct, 0.5 mg/ml of construct DNA was coinjected with P-turbo helper plasmid (pUChspD2-3wc) into w1 embryos. Transformants were identified by the presence of the mini-white selectable marker and were outcrossed to w1. Individual flies of each transgenic line were then crossed to marked balancer chromosomes to generate balanced stocks and to determine the chromosome of insertion.
Embryos were collected overnight (∼0- to 12-hr collection) and stained for 20–24 hr for β-galactosidase activity according to the protocol in Bellen et al. (1989) except that embryos were fixed for 20–22 min by mixing with saturated heptane according to the following procedure. Ten milliliters of heptane was saturated by vigorously mixing with 5 ml PBS and 5 ml 50% glutaraldehyde, phases were then allowed to separate, and the top phase of heptane was recovered.
To compare the relative levels of staining for each transgenic line, stainings were performed simultaneously in the same dish, and the same positive and negative controls were included each time the staining was performed. Line 25.182 containing a random DNA insert was used as a standard for dark staining. Line 92.31 containing five binding sites for Suppressor of Hairywing [Su(Hw)] was used as a standard for medium staining. Line 86.70.1 containing a 1.7-kb Fab-7 fragment and lines 117.17B and 117.31B containing a 1.2-kb Fab-7 fragment were used as standards for light staining. Line 25.182, 92.31, and 86.70.1 were generated by Hagstrom et al. (1996), whereas lines 117.17B and 117.31B were generated in this work. Each transgenic line was scored a minimum of three times from independent staining reactions. In the event that a particular line did not fall into the same category for all three staining reactions, the line was retested (in some cases 5–10 more times) until a consistent pattern was evident and it could be placed into a specific category with confidence. Lines carrying mutations in GAGA-factor-binding sites were scored single blind at least one time to prevent any unintentional bias.
To compare the relative levels of mini-white expression between lines, transgenic lines were outcrossed to w1. Heterozygous progeny of the same sex and age were examined. Heterozygous females under 24 hr of age allowed for the most sensitive detection of differences in eye color. For autosomal transgenes, both males and females were used for classification. For X-linked transgenes, only females were used for classification to eliminate complications due to dosage compensation. Each transgenic line was scored against a panel of control lines that were representative of each category of eye color. The categories of enhancer-blocking eye colors were yellow, yellow/orange, and orange. Flies with orange-red, red, or bright-orange eye colors were classified as nonblocking lines. Every transgenic line was scored by comparing numerous adults on each of three separate days.
GAGA factor binds to GAGAG sequences within HS1 of the Fab-7 boundary:
The minimal 1.2-kb Fab7 boundary element defined in transgene assays (Hagstrom et al. 1996) contains nine sequences that resemble the consensus-binding site GAGAG for the GAGA factor (Figure 1). Six of these directly match the consensus sequence GAGAG, which has been shown to be a high-affinity-binding site in vitro (Pedone et al. 1996; Omichinski et al. 1997). Five of the six GAGAG sites map within HS1, while the sixth is just at the distal edge of the hypersensitive region. These GAGAG sites are arranged as pairs of direct repeats, 1–2, 3–4, and 5–6, and the sites within each pair are separated from each other by 25–50 bp (Figure 1). In addition to the high-affinity sites, Fab-7 also contains three lower-affinity GAGAA sequences. One of these lower-affinity GAGAA sequences marks the proximal edge of HS1, while the two other sites are within HS1 and HS2, respectively.
The fact that most, if not all, of the GAGA-binding sites map within the nuclease hypersensitive regions of the Fab-7 boundary argues that these sequences will be accessible for protein:DNA interactions in vivo. Consistent with this idea, chromatin immunoprecipitation (ChIP) experiments have shown that GAGA is associated with the Fab-7 region in tissue culture cells (Strutt et al. 1997). To provide additional evidence that GAGA binds to target sequences in HS1, we incubated embryonic nuclear extracts with DNA affinity matrix prepared by binding different biotinylated subfragments from the HS1 and various positive or negative control sequences to streptavidin-coated magnetic beads. After incubating the beads in the extract, they were washed and bound and proteins were then eluted with 0.4 m KCl. The presence of the GAGA factor was assayed by Western blotting the eluted proteins with GAGA antibody. As shown in Figure 2, GAGA binds to beads containing three positive control sequences from elsewhere in the BX-C, one from Ubx promoter and two from the bxd region of the complex (Biggin and Tjian 1988; Horard et al. 2000). GAGA also binds to a fragment spanning the HS1 hypersensitive site, which contains all six of the consensus-binding sites, and to beads containing a smaller HS1 subfragment that has only the proximal high-affinity GAGA sites 1 and 2. However, GAGA does not bind to an affinity matrix containing another sequence from the Fab-7 boundary that lacks the consensus GAGA-binding sites.
In the experiment shown in Figure 2B, an HS1 subfragment probe spanning GAGAG sites 3 and 4 is shifted by embryonic nuclear extract. This shift is competed by the addition of the excess of the same unlabeled wild-type GAGAG3–4 fragment. However, the shift is not efficiently competed when the HS1 GAGAG3–4 subfragment used as cold competitor has mutations in GAGAG sites 3 and 4 (see F34 in Figure 1 for the GAGAG3–4 mutations). Likewise we found that a DNA fragment spanning GAGAG sites 1 and 2 could compete the shift observed with the GAGAG3–4 probe. In contrast, when the same fragment had mutations in GAGAG sites 1 and 2 (see F12 in Figure 1 for the GAGAG1–2 mutations), it could not efficiently compete the shift observed with the GAGAG3–4 probe (not shown). We also tested whether GAGA factor expressed in bacteria could bind to HS1 or to various HS1 subfragments using the same gel-shift assay (not shown). As expected, the HS1 fragments containing GAGA-binding sites could be shifted by the bacterial extracts. Moreover, the shifts could be competed by the addition of HS1 subfragments containing GAGAG sites 1–2 or 3–4. On the other hand, the shifts were not efficiently competed with HS1 subfragments in which the GAGAG sites 1–2 or 3–4 were mutant.
To provide additional evidence that GAGA factor is responsible for the gel shifts observed with nuclear extracts, we tested whether the shifted band could be supershifted by the addition of antibodies directed against the GAGA protein. As shown for a fragment spanning GAGAG sites 5 and 6, the gel-shifted band seen in nuclear extracts is supershifted by the addition of anti-GAGA factor antibody (Figure 2C).
Mutation of GAGA-factor-binding sites disrupts Fab-7 boundary activity:
We next asked whether the GAGAG sequences in HS1 are important for the boundary activity of the Fab-7 element in vivo. To test whether mutations in the HS1 GAGAG sequences affect Fab-7 boundary function, we used two enhancer-blocking assays. As shown in Figure 3, the first blocking assay is a white enhancer mini-white transgene. In the absence of a boundary element, the white enhancer drives a high level of white expression in the eye, giving a red eye color. However, when a boundary element is interposed between the enhancer and the white promoter, it blocks enhancer:promoter communication. As a consequence, white expression is reduced, and the transgenic flies have a yellow or orange eye color depending upon the strength of the boundary element. While the Fab-7 boundary reduces white expression in this assay to light orange or yellow (see Figure 4), for reasons that are not known the blocking activity of the Fab-7 boundary in the wEN:mini-white transgene is insertion site dependent. It is able to block the white enhancer from activating mini-white in about one-half of the transgenic lines (Hagstrom et al. 1996; see Figure 4). In this respect, Fab-7 differs from another BX-C boundary, Fab-8, which blocks the white enhancer in the wEN:mini-white transgene independent of the site of insertion (Barges et al. 2000).
To determine whether the GAGAG sequences in HS1 are important for boundary activity, we generated a 1.2-kb Fab-7 element in which sites 1–5 were mutated as indicated in Figure 1. This mutant Fab-7:GAGAG1–5 element was introduced into the wEN:mini-white transgene, and transgenic lines were isolated. Whereas the wild-type Fab-7 element blocks in 48% of the wEN:mini-white lines, the boundary activity of the Fab-7:GAGAG1–5 mutant is reduced and only 25% of the transgenic lines have an eye color that is lighter than that of a control transgene that lacks the insert (Figure 4). In addition to a decrease in the frequency of enhancer-blocking lines, we also find that the range of eye colors of enhancer-blocking lines is darker for the Fab-7:GAGAG1–5 element than for the wild-type Fab-7 (Table 1; see example in Figure 4). We speculate that the darker eye colors of the blocking lines reflect a general weakening of boundary strength. Although blocking activity of the Fab-7:GAGAG1–5 element is clearly diminished, it is not completely eliminated. Thus, a wEN:mini-white transgene that carries a fragment of roughly similar size (1340 bp) but with no apparent blocking activity shows a reduction in eye color compared to the control in only 5% of the transgenic lines.
To confirm the results with the wEN:mini-white transgene, we used a ftz:hsp70-LacZ transgene as a second enhancer-blocking assay (see Figure 3). In this transgene the boundary element is inserted between the UPS stripe and NE neurogenic enhancers of the ftz gene and an hsp70-LacZ reporter gene. As shown for the random DNA control in Figure 5, the UPS enhancer drives a high level of LacZ expression in even-numbered parasegments in germ-band extended embryos while the NE enhancer drives a high level of LacZ expression in the CNS of older germ-band retracted embryos. We have previously shown that the enhancer-blocking activity of a larger 1.7-kb Fab-7 fragment is intermediate between that of 5 binding sites and 12 binding sites for the insulator protein Su(Hw) (Hagstrom et al. 1996). In addition, the blocking activity of the 1.7-kb Fab-7 fragment in the ftz:hsp70:LacZ assay is much less sensitive to the insertion site than it is in the wEN:mini-white assay, and blocking is observed in all transgenic lines. As illustrated in Figure 5, we found that the blocking activity of the wild-type 1.2-kb Fab7 element is about the same as the larger 1.7-kb fragment; it is stronger than 5 Su(Hw)-binding sites (compare the stripe and CNS expression in the 1.2-kb Fab-7 and Su(Hw) embryos) but weaker than 12 (not shown).
Just as was observed in the wEN:mini-white assay, the GAGAG1–5 mutations weaken the enhancer-blocking activity of the 1.2-kb Fab-7 boundary in the ftz:hsp70-LacZ assay (Figure 5). Moreover, the GAGAG1–5 mutations appear to make the boundary activity of the Fab-7 element in the ftz:hsp70-LacZ transgene more susceptible to position effects, and a greater range of enhancer-blocking activity between different transgene inserts is observed (Table 2). In addition, inserts that show reduced UPS blocking activity do not necessarily show reduced NE blocking activity. Thus, 7 of the 14 transgenic lines have reduced UPS enhancer-blocking activity while only 4 of these lines show defects in NE enhancer-blocking activity.
The GAGAG site pairs 1–2 and 3–4 have different functions:
The GAGAG sites in HS1 are arranged in three pairs: 1–2, 3–4, and 5–6. To obtain some idea as to whether these clusters are functionally significant, we generated Fab-7 elements that had mutations in either sites 1–2 or sites 3–4. We then tested the blocking activity of these mutant elements in both the wEN:mini-white and the ftz:hsp70-LacZ assays.
GAGAG site pair 3–4 is required for Fab-7 boundary function in the eye but not in the embryo:
As shown in Figure 4, the enhancer-blocking activity of the Fab-7 GAGAG3–4 mutant in the mini-white assay is reduced almost as much as the Fab-7 GAGAG1–5 mutant. Blocking of the white enhancer is observed in less than one-third of the Fab-7 GAGAG3–4 transgenic lines. In addition, as was observed for the Fab-7 GAGAG1–5 mutant, the eye color of the subset of lines that show evidence of blocking activity is generally darker than that in the wild-type Fab-7 control (see example in Figure 4 and Table 1). We presume that the increased expression of mini-white in these blocking lines reflects a weakening of the boundary activity of the Fab-7 element by the GAGAG3–4 mutations.
Although the GAGAG site pair 3–4 is required for full enhancer-blocking activity in the adult eye, mutation of these sites has only little effect on blocking activity in the ftz:hsp70-LacZ assay. As illustrated by representative embryos in Figure 5, the ability of the Fab-7 GAGAG3–4 mutant to block the UPS and NE enhancers is close to or nearly equivalent to the wild-type 1.2-kb Fab-7 element (see also Table 2). These results suggest that the GAGAG site pair 3–4 is required for blocking activity in the adult but is largely dispensable for boundary function during embryogenesis.
GAGAG site pair 1–2 is not required for Fab-7 boundary function in the eye:
Unlike the GAGAG 3–4 pair, mutations in the GAGAG 1–2 pair have no effect on enhancer-blocking activity of the Fab-7 element in the wEN:mini-white assay. As is observed for the wild-type 1.2-kb Fab-7 element, about one-half of the transgenic lines have a reduced eye color compared to control lines that do not contain insert (see Table 1). Similarly, mutations in sites 1–2 appear to have no effect on the blocking activity of the Fab-7 element in the CNS of germ-band retracted embryos, and the Fab-7 GAGAG1–2 mutant blocks ftz NE enhancer as well as the wild-type Fab-7 element. On the other hand, mutations in the GAGAG 1–2 pair do have a small effect on the ability of the Fab-7 element to block the UPS enhancer in germ-band extended stage 10 embryos (see Figure 5 and Table 2). This finding suggests that GAGAG1–2 sequences may contribute to the boundary activity of the Fab-7 element during the early stages of embryogenesis.
Genetic interaction with Trl, the gene encoding GAGA factor:
The findings described in the previous section indicate that the GAGA-binding sites in HS1 are important for the boundary function of Fab-7. To provide further evidence that the GAGA factor contributes to Fab-7 boundary function, we tested whether mutations in Trl, the gene encoding GAGA factor, perturb enhancer-blocking activity. We tested two alleles, Trl13c and TrlR67, in both the wEN:mini-white and the ftz:hsp70-LacZ assays. Trl13c is a weak hypomorph caused by the insertion of a P element into the 5′ UTR of Trl (Farkas et al. 1994). Imprecise excision of this P-element generated a second amorphic allele, TrlR67, that is homozygous lethal (Farkas et al. 1994). As Trl is an essential gene, it is impossible to determine whether the Fab-7 boundary is active in the complete absence of GAGA protein. For this reason, we were restricted to testing boundary function under conditions in which Trl activity is only partially reduced. There were no effects on the blocking activity of the wildtype Fab-7 boundary in the wEN:mini-white assay in flies heterozygous for the TrlR67 null allele or in Trl13c homozygotes. Transgenic lines from the different GAGAG mutants were tested as well; however, no effects were observed. There were also no discernible effects on boundary activity in the ftz:hsp70-LacZ assay in progeny from TrlR67/+ females. In the case of embryos from homozygous Trl13c mothers, we found that LacZ expression from the starting ftz:hsp70-LacZ reporter in the few embryos from homozygous mutant mothers that develop beyond the blastoderm stage (Bhat et al. 1996) is substantially reduced compared to embryos from wild-type mothers. Presumably this reflects the fact that the GAGA factor is required for the functioning of the hsp70 promoter in this report construct. Under these conditions, we were unable to observe any effects of the Trl13c mutation on the blocking activity of the Fab-7 element.
While these findings could indicate that Trl is not required for Fab-7 boundary function, an alternative possibility is that our transgene assays are not sufficiently sensitive to detect the effects of relatively modest decreases in Trl activity. For this reason we sought an assay that would be more responsive to a reduction in Trl activity. As shown in Figure 6, the bluetail transposon is inserted into the iab-7 cis-regulatory domain between the boundary and the iab-7 PRE. bluetail carries a Ubx-LacZ reporter (Galloni et al. 1993). This reporter is subject to regulatory elements located in the iab-7 domain but is protected from the immediately adjacent iab-6 cis-regulatory domain by the Fab-7 boundary. The iab-7 domain drives LacZ expression in PS12 and more posterior parasegments from early embryogenesis through to the adult stage. Mihaly et al. (1997) generated a series of Fab-7 boundary deletions by mobilizing the transposon. Three of these deletions removed Fab-7 boundary sequences but retained an intact bluetail transposon. The largest of these, iab-6,7P14.1 (P14.1), deletes an ∼1.0-kb DNA segment that corresponds closely to the minimal 1.2-kb Fab-7 boundary and extends from the bluetail insertion site to near the middle of the minor hypersensitive site (Figure 6). As would be expected from the fact that this deletion removes nearly the entire Fab-7 boundary, the bluetail Ubx-LacZ reporter is driven by iab-6 and the anterior limit of LacZ expression is PS11 from early embryogenesis onward (Figure 6). Two smaller deletions, iab6,7P6.1 (P6.1) and iab-6,7P18.1 (P18.1), which remove 510 and 594 bp, respectively, were also isolated. The proximal breakpoints of P6.1 and P18.1 are in HS1 and both lack GAGA sites 3–6, but retain GAGA sites 1–2. Both deletions have boundary function during the early stages of embryogenesis and the anterior limit of LacZ expression is PS12 just like the bluetail transposon with an intact Fab-7 boundary (see Figure 6; Schweinsberg and Schedl 2004). However, boundary function is lost later in embryogenesis after germ-band retraction and LacZ expression spreads into PS11 in both the developing CNS and the ectoderm (Mihaly et al. 1996: Schweinsberg and Schedl 2004). We reasoned that the early boundary activity of these two deletion mutants might be sensitive to changes in the dose of the GAGA factor. Since the bulk of the GAGA protein in early embryos is of maternal origin, we crossed TrlR67/+ females to males carrying the bluetail transposon. As shown in Figure 6, LacZ expression in the 510-bp P6.1 deletion is not affected by a reduction in the dose of maternally deposited GAGA factor and, as is observed when the mothers have two wild-type copies of Trl, the anterior limit is PS12 in germ-band extended embryos. In contrast, in the 594-bp P18.1 deletion LacZ expression can be detected prematurely in PS11 in embryos from TrlR67/+ mothers. These findings indicate that P18.1 boundary activity in early embryos is weakened by a twofold reduction in the maternal contribution of the GAGA factor. Since P6.1, which has an additional 84 bp from the HS1, is not sensitive to a reduction in the amount of GAGA protein, it would appear that there are target sequences in the Fab-7 boundary for other proteins that can compensate for a reduction or loss of GAGA-factor binding.
GAGA sites are not sufficient for enhancer-blocking activity:
Multimerized binding sites for boundary proteins can confer enhancer-blocking activity. For example, the naturally occurring insulator element in the gypsy transposon consists of an array of ∼12 consensus-binding sites for the Su(Hw) protein (Geyer and Corces 1992). Although it appears that not all 12 sites are occupied by Su(Hw) at the same time, multiple Su(Hw) proteins must bind to this element to confer insulator activity. Similarly, an oligo containing several copies of the binding site for the scs boundary protein Zw5 also has enhancer-blocking activity (Gaszner et al. 1999). Since Fab-7 enhancer-blocking activity seems to depend upon multiple GAGAG-factor-binding sites in HS1, an obvious question is whether boundary activity can be reconstituted by a multimerized array of GAGA-binding sites. To answer this question, we tested a multimer consisting of 8 GAGAG consensus-binding sites for enhancer-blocking activity in the ftz:hsp70/LacZ assay. Unlike multimerized Su(Hw) and Zw5 binding sites, multimerized GAGA-binding sites do not have enhancer-blocking activity. As illustrated for a representative transgenic line in Figure 6, most of the 8XGAGAG transgenic lines have a high level of LacZ, comparable in levels to lines carrying a random DNA insert. A couple of lines appear even darker than the random DNA control, whereas a few show moderate levels of staining (Table 2). Thus, while GAGAG sites in HS1 are required for the boundary function of the Fab-7 element, multimerized GAGA-binding sites do not appear to be sufficient to confer enhancer-blocking activity.
Elements that function as boundaries or insulators have been identified in many different species. A characteristic feature of these elements is chromatin-specific nuclease hypersensitive regions (West et al. 2002). These hypersensitive regions contain target sequences for proteins that are important for boundary function. In boundaries that are active in a wide range of developmental stages, tissues, or cell types, these hypersensitive regions are generally “constitutive.” In the case of regulated boundaries, such as the imprinted H19 insulator in mice, only the active element is a hypersensitive site, while the imprinted inactive element is not (Hark and Tilghman 1998; Bell and Felsenfeld 2000; Hark et al. 2000). Like many other boundaries, the minimal 1.2-kb Fab-7 boundary contains several prominent chromatin-specific nuclease hypersensitive regions. One of these hypersensitive regions, HS1, is ∼400 bp in length and contains multiple binding sites for the Drosophila GAGA factor. Since these GAGA sites are in an accessible configuration in chromatin (Galloni et al. 1993; Karch et al. 1994), it would be reasonable to expect that the GAGA factor will recognize the consensus-binding sites and interact with the Fab-7 boundary. Consistent with this expectation, we found that the GAGA factor in embryonic nuclear extracts binds to Fab-7 fragments containing GAGA consensus sequences from HS1 but not to Fab-7 fragments that do not have sequences resembling the consensus-binding site. We have also found that bacterially expressed GAGA binds to HS1 subfragments containing the consensus-binding sequences and that mutation of these sites perturbs binding. While these results demonstrate that the GAGA factor is capable of binding to HS1, the most compelling evidence that GAGA is actually associated with the Fab-7 boundary in vivo comes from the ChIP experiments of Strutt et al. (1997). They found that the Fab-7 boundary is highly enriched in GAGA immunoprecipitates while the flanking sequences are not.
The fact that GAGA is localized to the Fab-7 boundary in vivo and binds to sequences in HS1 in vitro raises the question of whether this association is functionally significant. We have addressed this question by examining the boundary activity of Fab-7 elements that have mutations in the HS1 consensus GAGA-binding sequences. Altogether there are eight consensus GAGA-factor-binding sites in the 400-bp HS1. Six of these are high-affinity sites and have the sequence GAGAG while two have the sequence GAGAA. While the presence of multiple GAGA consensus-binding sites in HS1 would favor the idea that GAGA plays an important role in Fab-7 boundary function, it also seemed likely that at least some of these sites may be functionally redundant. Thus, to determine whether the consensus sites are important for boundary function, we first asked whether it was possible to compromise enhancer-blocking activity of the Fab-7 element by mutating most, but not all, of these sites. We found that a Fab-7 element, which has mutations in five of the eight consensus GAGA sites, has significantly reduced enhancer-blocking activity. Two effects are observed in the wEN:mini-white assay. First, the frequency of blocking lines is reduced approximately in half. Second, among the GAGAG1–5 transgenic lines in which blocking is still observed, the white enhancer typically drives a higher level of white expression than in transgenes carrying the wild-type Fab-7 element. Both of these effects argue that the boundary function of the Fab-7 element in the adult eye is significantly compromised by the GAGAG1–5 mutations. Since the enhancer-blocking activity of the Fab-7 GAGAG1–5 mutant is also diminished in the ftz:hsp70-LacZ assay, it would appear that these GAGA-factor-binding sites are important for Fab-7 boundary function in the embryo as well.
The six GAGAG sites in HS1 are arranged in “pairs.” As a second test, we asked whether it was possible to compromise Fab-7 boundary function by mutating the GAGAG sites in two of these pairs: 1–2 and 3–4. We found that the effects of the GAGAG3–4 mutation on boundary function in the wEN:mini-white assay were essentially equivalent to those observed for GAGAG1–5. Thus, there was a reduction both in the frequency of blocking lines and in the strength of the blocking activity in the few lines that actually showed blocking. However, Fab-7 GAGAG3–4 differed from Fab-7 GAGAG1–5 in that there was little, if any, effect on boundary function in the embryo, and the Fab-7 GAGAG3–4 mutant blocked the ftz stripe (UPS) and the CNS (NE) enhancers almost as well as did the wild-type Fab-7 element. Mutations in GAGAG sites 1–2 had, if anything, the opposite effect on boundary function. In the wEN:mini-white assay, the blocking activity of Fab-7 GAGAG1–2 is indistinguishable from that of the wild-type control. In the ftz:hsp70-LacZ assay, there appears to be a reduction in the ability of the mutant element to block the ftz stripe UPS enhancer. Supporting the idea that this small effect is meaningful, two partial Fab-7 boundary deletions that retain GAGAG sites 1–2, but lack the other GAGAG sites in HS1, have boundary function in early embryos, but lack boundary function later in embryogenesis and in adults (see Mihaly et al. 1996; Schweinsberg and Schedl 2004). In addition, we found that a multimerized ∼250-bp fragment from the proximal edge of HS1 that includes GAGAG sites 1–2 blocks the ftz UPS enhancer, but not the ftz NE enhancer or the w enhancer (Schweinsberg and Schedl 2004).
At this point there is good evidence that the GAGA factor interacts with the GAGAG sites in HS1 both in vitro and in vivo. First, recombinant GAGA binds to HS1 subfragments in vitro and this binding depends upon the GAGAG consensus sequences. Second, GAGA factor in nuclear extracts binds to HS1 subfragments and this binding also appears to depend on the GAGAG consensus sequences. Third, ChIP experiments by Strutt et al. (1997) demonstrate that the GAGA factor is associated with Fab-7 in vivo. While the experiments of Strutt et al. were done with tissue culture cells, the GAGAG consensus sequences in HS1 are in a large nuclease hypersensitive region not only in tissue culture cells but also in embryos of different ages (Karch et al. 1994; P. Schedl, unpublished data). Consequently, these binding sites should also be accessible to the GAGA factor at least during embryogenesis. A more difficult problem is establishing that GAGA contributes to Fab-7 boundary function throughout development. There are two major complications. First, because Trl is a cell-vital gene, it is impossible with our current assays to determine whether Fab-7 retains boundary function in the complete absence of any GAGA factor. In adults Trl activity can be reduced but not eliminated using the semivital hypomorphic allele Trl13c; however, even in homozygous Trl13c animals no effects on Fab-7 blocking activity in the wEN:mini-white assay were detected. One likely reason why no effects were observed in the mini-white assay is that the levels of GAGA factor required for viability through to the adult stage are more than sufficient to confer Fab-7 boundary function. A second problem is that the intact boundary is likely to contain target sequences for factors that can compensate for even a relatively significant reduction in GAGA activity. Two lines of evidence suggest this possibility. The first is the fact that instead of severely compromising Fab-7 boundary function, the GAGAG1–5 mutations only weaken or diminish enhancer-blocking activity in transgene assays. The second is the finding that the boundary activity of the Fab-7 deletion mutants P18.1 and P6.1 differ in their sensitivity to a reduction in the dose of the Trl gene in early embryos. The enhancer-blocking activity of P18.1 is weakened by a reduction in GAGA protein levels while P6.1 is not. Since the P6.1 deletion has 84 bp from HS1 that are missing in P18.1, a plausible explanation for the difference in sensitivity to the Trl gene dose is that there are target sequences in this 84-bp sequence for proteins that can compensate for a reduction in the amount of GAGA factor available for interaction with GAGAG sites 1 and 2. Of course, the fact that the P18.1 deletion is sensitive to the Trl gene dose provides a strong argument that the GAGA factor contributes to the activity of the intact Fab-7 boundary. Moreover, taken together with the effects of the GAGAG1–2 mutations on the blocking of the UPS enhancer in the ftz:hsp70-LacZ assay, these findings would be consistent with the proposal that GAGA factor interactions with GAGAG sites 1 and 2 play an important role in Fab-7 boundary activity in early embryos. Unfortunately, in the absence of a sensitized assay system it is not possible to draw a similar conclusion concerning the four other GAGAG sites in HS1. However, we think that it is reasonable to believe that GAGA factor binding to these other sites will also turn out to be important for Fab-7 boundary activity. Of course, even if this is correct, it would not exclude the possibility that other factors, such as Pipsqueak (Horowitz and Berg 1996), also interact with one or more of the GAGAG sites in HS1 and contribute to boundary function.
Assuming that GAGA factor interactions with the consensus-binding sites in HS1 are important to Fab-7 activity, one question of interest is whether GAGA plays a direct or indirect role in boundary function. Since GAGA-binding sites have been shown to be required for the enhancer-blocking activity of other fly elements in addition to Fab-7 (Ohtsuki and Levine 1998; Belozerov et al. 2003), one must consider the possibility that GAGA plays a direct role in boundary function analogous to that of, for example, the Su(Hw). However, this view is difficult to reconcile with the fact that GAGA is required for the functioning of other elements unrelated to boundaries such as promoters, enhancers, and PREs, not to mention its role in centromeric heterochromatin and chromosome segregation (Biggin and Tjian 1988; Soeller et al. 1988; Raff et al. 1994; Bhat et al. 1996; Hagstrom et al. 1997; Wilkins and Lis 1997; Horard et al. 2000; Barges et al. 2000). Also arguing against a direct role in boundary function, we have found that enhancer-blocking activity cannot be reconstituted by multimerizing GAGA-factor-binding sites. On the other hand, one known activity of the GAGA factor that would account for its ability to participate in the functioning of such a diverse array of regulatory elements is the formation and/or maintenance of nucleosome-free regions of chromatin (Lu et al. 1993; Tsukiyama et al. 1994; Tsukiyama and Wu 1995; Wall et al. 1995). In this model, GAGA factor binding to sites in HS1 would ensure that this 400-bp sequence is nucleosome free and that target sequences within HS1 are readily accessible for the binding of other factors that actually confer boundary function. In this case, mutations in the HS1 GAGA sites would disrupt boundary function indirectly because of difficulties in generating a nucleosome-free region of chromatin, which is fully accessible for these other boundary proteins. While we favor the idea that a key function of the GAGA protein is to ensure DNA accessibility, we believe that it is reasonable to think that it may also play a more central role in Fab-7 boundary function because of its ability to participate in protein:protein interactions. The N terminus of the GAGA protein has a BTB/POZ domain that is present in numerous other proteins (Horowitz and Berg 1996). This domain has been shown to mediate interactions between GAGA and a large collection of other proteins, including several that have BTB/POZ domains and/or interact with DNA (Espinas et al. 2000; Pointud et al. 2001; Pagans et al. 2002; Faucheux et al. 2003; Mishra et al. 2003). Moreover, depending upon the particular protein partners, the GAGA factor appears to have rather different activities. Thus, a plausible hypothesis is that GAGA can have boundary functions when it is combined with one set of proteins, transcriptional activation functions when it is combined with another, and Pc-G-silencing functions when combined with yet a third set of proteins. This would explain why no boundary function was observed with the multimermized GAGAG sites. Moreover, if this hypothesis is correct, then it would be reasonable to think that the partners for GAGA protein bound at sites 3–4 are likely to be different from the partners for GAGA protein bound at sites 1–2 or, presumably, 5–6. Further studies will be required to identify these putative partners and to understand how they function together with the GAGA factor to provide a scaffold for building a boundary element.
We thank members of the Karch, Mishra, and Schedl labs for helpful discussions as well as for providing reagents used in these studies. Thanks also go to H. Gyurkovics and J. Gausz for helpful discussions and comments on the work. F.K. acknowledges support from the Swiss National Fund and the State of Geneva. R.K.M. was supported by a grant from the Human Sciences Frontier Program. P.S. acknowledges support from the National Institutes of Health. S.S. especially thanks the New Jersey Cancer Commission for a predoctoral research fellowship during the course of this work.
Communicating editor: A. J. Lopez
- Received April 2, 2004.
- Accepted July 12, 2004.
- Genetics Society of America