The CREB binding protein (CBP) is one of most extensively characterized coactivator proteins. CBP was first identified through its ability to link the cyclic AMP protein kinase (PKA)-phosphorylated form of CREB to components of the basal transcriptional machinery, including TFIIB (14, 34), TATA-binding protein (65), and the RNA polymerase II holoenzyme complex (28, 44). CBP is highly related to the adenovirus E1A binding protein p300 (17), and CBP and p300 are considered to be functional homologues (4, 38), although a few differences in their activities have been reported (27). CBP and p300 associate with a wide variety of transcriptional activators in addition to CREB, suggesting that each may serve as a transcriptional integrator of different signaling cascades (reviewed in references 20 and 60). Thus, one model for the function of CBP and p300 is bridging DNA binding transcription factors to components of the basal transcriptional machinery.

Another function of coactivators appears to be the modification of chromatin structure. In this regard, CBP and p300 have also been proposed to mediate transcriptional activation via intrinsic (6, 46) and associated (9, 63, 81) histone acetyltransferase (HAT) activity. Targeted HAT activity is thought to facilitate the access of nuclear factors to their target sites by relaxing the interaction between histones and the DNA (for a review, see reference 77). Moreover, recent studies suggest that transcriptional activation mediated through CBP or p300 occurs only in the context of chromatin (31, 32). Therefore, CBP and p300 may regulate gene expression by interacting with components of the transcriptional machinery as well as by augmenting the access of factors to DNA through their HAT activities. Acetylation of basal and sequence-specific transcriptional regulators may also contribute to CBP function (22, 24).

Genetic studies indicate an essential role for CBP in cellular function and development (reviewed in reference 21). In humans, CBP loss of function is associated with Rubinstein-Taybi syndrome, a haploinsufficiency disorder characterized by mental retardation, developmental defects, and an increased predisposition to cancer (42, 48). Chromosomal translocations that fuse CBP with MOZ (monocytic zinc finger protein) or MLL (mixed-lineage leukemia protein, a trithorax group-like protein) are associated with various types of myeloid leukemia (7, 62). In addition, somatic mutations of the p300 gene have been detected in colorectal and gastric carcinomas (43). Gene knockouts in mice indicated that CBP and p300 are required for normal embryonic development and viability (69, 82). Finally, mutations in the Caenorhabditis elegans homologue of CBP (CBP-1) affect the differentiation of several embryonic tissues (59).

In Drosophila melanogaster, Drosophila CBP (dCBP) loss-of-function mutations cause embryonic lethality. Specifically, dCBP serves as a coactivator for transcription factor Cubitus interruptus (CI) and mediates its activity in the hedgehog pathway (2, 12). dCBP is also a coactivator of the dorsal protein (D1) and Mad, mediating dl-dependent twist expression and dpp-induced transcriptional stimulation, respectively (3, 79). However, dCBP does not always function as a coactivator. Recent studies have shown that dCBP binds to the Drosophila homologue of the T-cell factor (dTCF) and facilitates dTCF-mediated repression in the Wnt/Wingless signaling (78). Therefore, dCBP can function as both a coactivator and a corepressor during embryogenesis.

To further define the developmental processes and the signaling pathways that require dCBP, we have taken advantage of the yeast GAL4 enhancer trap (8) system to generate transgenic flies that overexpress dCBP in a variety of cell types. The dominant overexpression adult phenotypes generated with this system were used to screen for suppressors of dCBP overactivity in specific tissues. In this report, we describe a functional and specific interaction between transcriptional coactivator dCBP and ASH1, a member of the trithorax group (trxG) of chromatin modifiers.

The trxG proteins are required to maintain the continued and efficient expression of homeotic and other genes throughout development. Loss-of-function mutations in the trithorax group genes cause homeotic transformations because they fail to maintain the expression pattern of homeotic selector genes. While the trxG proteins function as transcriptional activators, members of Polycomb group (PcG) genes form stable complexes that maintain a repressed pattern of homeotic gene function (for reviews, see references 49, 50, and 61). Current models envision that trxG and PcG proteins lock in the active or inactive state, respectively, by creating a stable chromatin organization. trxG represents a heterogeneous family of proteins with diverse functions. Some of these proteins, such as Trithorax (TRX), ASH1, ASH2, GAGA, and ZESTE, are associated with particular sites on polytene chromosomes (1, 13, 33, 51, 71, 74), while others, such as Brahma (BRM) and SNR1, are found in chromatin-remodeling complexes that may not be associated with specific chromosomal regions (16, 68). There is some evidence that one of the functions of the trxG proteins may be to recruit chromatin-remodeling complexes to DNA. GAGA is required for the function of one chromatin-remodeling complex, the Drosophila NURF complex (74), and TRX was shown to physically interact with SNR1, a component of the Drosophila SWI/SNF complex (54). These studies strongly support the model that trxG proteins are important regulators of higher-order chromatin structure. However, the precise role of each of the diverse trxG members and the functional relationships that might exist among them and with other transcriptional regulatory factors are still poorly understood.

Our studies show that mutations in the ash1 gene suppress a wing phenotype caused by the overexpression of dCBP in specific central nervous system (CNS) cells. This suppression is specific for ash1 because other members of the trithorax family do not have the same effect. At the cellular level, ASH1 expression coincides with the overexpression pattern of dCBP and, in wild-type flies, ASH1 and dCBP colocalize in the nucleus of the ASH1-expressing neurons. Finally, at the molecular level, we show that dCBP interacts strongly with ASH1 and that the two proteins colocalize to specific sites on polytene chromosomes. Our results strongly suggest that coactivator dCBP and trithorax factor ASH1 are part of a functional complex in vivo. These findings implicate a new type of chromatin-associated proteins in mediating dCBP function and imply that, in addition to its HAT activity, dCBP may participate in the regulation of higher-order chromatin structure.

Drosophila strains. All the ash1 alleles and the ash2 allele were kindly provided by Allen Shearn (Johns Hopkins University). They are described by Tripoulas et al. (70, 71) and Adamson and Shearn (1). The PKA-C1/DCO null allele, G9, (36) and the UAS-R* line containing the upstream activation sequence (UAS) dominant-negative catalytic subunit of PKA (37) were kindly provided by Dan Kalderon (Columbia University). The GMR-GAL4 line was provided G. Rubin, (University of California, Berkeley). Other mutant alleles and most of the deletions used in this study came from the Bloomington Stock Center. UAS-dCBP transformants (Tr21 and Tr36) were established by standard methods. The NotI dCBP fragment that includes the entire dCBP cDNA was cloned into the NotI site of pUAST (8). yw embryos were injected with this DNA and the pπΔ2-3 helper as described previously, and the transformants were mapped and put into stock (64). The UAS-dCBP Tr21 insert is on the fourth chromosome, and the Tr36 insert is on the X. The UAS-386 and UAS-363 lines were kindly provided by C. O'Kane (Cambridge University). The balancer chromosomes MKRS, TM3,Stubble (Sb), and TM6B,Tubby (Tb), which are used in these studies, have been described by Lindsley and Zimm (37a). The balancer chromosome strain pk-sple³³ pr cn/T(2;3)SM6.TM6B,Tb was kindly provided by J. Roote and M. Ashburner (Cambridge University). The wild-type strain used in these studies is Canton-S. All crosses were reared at 25°C on standard cornmeal yeast extract source media.

Detection of β-galactosidase (β-GAL) activity in imaginal discs and CNS. Third-instar larvae were dissected into saline, fixed for 1 h in phosphate-buffered saline (PBS) and 4% formaldehyde, and rinsed several times with PBST (PBS with 0.3% Triton X-100). Fixed samples were then placed in a 0.02% X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) solution (53) at 30°C until the appearance of the blue color (about 1 h for the GAL4-386 CNS, no staining after 24 h for the GAL4-386 imaginal discs). After staining, samples were washed several times in PBST. Larval tissues were mounted in PBS–50% glycerol and photographed with Echtachrome 64 film using Nomarski optics.

Screen to isolate dominant suppressors of the dCBP overexpression wing phenotype. The screen was performed at 25°C where the uninflated-wing phenotype has a 100% penetrance. GAL4-386/MKRS; Tr21/+ females were collected and crossed to males from the Bloomington Stock Center deficiency kit [Df(2)/Balancer or Df(3)/Balancer] (see Fig. 2). Approximately 150 different deletions, which cover more than 80% of the second and third chromosomes, were tested. From these crosses, one-fourth of the progeny were expected to be GAL4-386/+; Tr21/+ and have uninflated wings. In fact, this ratio is less and approximates one-eighth of the total population because this genotype is weak. Although Tr21 is inserted on the fourth chromosome, it is linked to a miniwhite transgene and produces a characteristic pale-orange eye color that can be monitored in a w¹ background. All of the second and third chromosome balancers used in this study were tested and had no effect on the uninflated-wing phenotype (data not shown). When the balancer chromosome of a deficiency stock had a dominant wing marker, for example, Curly or Serrate, the balancer was exchanged with one carrying a more convenient marker, i.e., SM6.TM6B,Tb or TM3,Sb. In this screen, a deletion was considered to have no effect on the uninflated-wing phenotype when the population of GAL4-386 heterozygotes had approximately the same number of flies with uninflated wings regardless of whether they carried the deletion chromosome or the balancer. When the number of uninflated-wing GAL4-386/+ flies with the balancer chromosome was greater than the number of uninflated-wing GAL4-386/+ flies with the deletion chromosome by at least a factor of two, then putative suppressed males (partially inflated or wild-type wings) were crossed to w′ females to determine the presence of the Tr21 transgene in the next generation. Approximately 20 to 30 putative suppressed males were tested for the presence of Tr21. When Df/+; GAL4-386/+; Tr21/+ males were obtained, the deletion was considered a potential suppressor. These deletions were rescreened in the same way to confirm the suppression. Only the deletions for which the number of Df/+; GAL4-386/+; Tr21/+ flies with partially inflated or wild-type wings (suppressed phenotype) was at least 50% greater than the number of flies with the same genotype and uninflated wings were considered suppressors (deletions in Table 2). The same procedure was used to determine suppression by mutations in single genes (alleles in Table 3). Each experiment was repeated at least twice.

FIG. 2 dCBP overexpression with the GAL4-386 driver coincides with the expression of the neuropeptides FMRFamide and PHM. Arrows, colocalization of dCBP (rhodamine) and FMRFamide (fluorescein isothiocyanate [FITC]) in the larval (A) and pupal (more ...)

TABLE 2 Deletions that can dominantly suppress the dCBP overexpression wing phenotype and suppressor genes contained in these deletions

TABLE 3 Suppression of the uninflated-wing phenotype by the ash1 alleles and alleles of the trxG or PcG families

β-Gal expression tests. This assay allows us to measure the possible effect of a deletion on the expression of UAS-transgene and distinguish between suppressors that repress the expression of the transgene and those that represent dCBP-interacting genes (see Table 2). Deficiency stocks were put over a balancer chromosome with a dominant pupal marker: SM6.TM6B,Tb or TM6b,Tb. Df(2)/SM6.TM6B,Tb or Df(3)/TM6b,Tb males were crossed to UAS-LacZ; GAL4-386 homozygous females. Three-day-old pupae were collected and assayed for β-Gal activity using o-nitrophenyl-β-d-galactopyranoside (ONPG) as the substrate. Briefly, a single pupa was squashed and permeabilized in 800 μl of Z buffer (Na₂HPO₄ · 7H₂O [16.1 g/liter], NaH₂PO₄ · H₂O [5.5 g/liter], KCl [0.75 g/liter], MgSO₄ · 7H₂O [0.246 g/liter], pH 7.0) containing β-mercaptoethanol (2.7 μl/ml) plus 50 μl of 0.1% sodium dodecyl sulfate (SDS) and 50 μl of chloroform. ONPG was used as the substrate (160 μl of a solution of 4 mg/ml in 0.1 M phosphate buffer, pH 7.0), and assays were performed at 30°C. After 30 min, the reaction was stopped with 0.4 ml of 1 M Na₂CO₃ and the optical density at 420 nm was determined. For each test, at least five pupae were analyzed separately. β-Gal activity in Tb flies was compared to β-Gal activity in the non-Tb flies that contained the deletion. If the activities were similar, then the deletion has no effect and we considered the test positive. If the activity was largely reduced in the non-Tb flies, then the deletion represses the transgene either by affecting GAL4, the UAS, or the stability of the gene product and we considered the test negative. Using the same strategy (Tb versus non-Tb larvae), we were also able to compare qualitatively the expression of β-Gal in fixed larval brains stained with X-Gal (53). In all cases, the qualitative results in whole tissues were consistent with the quantitative values obtained for the pupal squashes (data not shown).

Confocal analysis of larval and pupal CNS. Third-instar larvae or pharate adults were dissected into saline, and the CNSs were processed. They were fixed for 1 h in PBS–4% paraformaldehyde, rinsed several times with PBST, and blocked at least for 1 h in PBST–10% HS. The CNSs were incubated overnight at 4°C with the primary antibody, washed several times in PBST–10% HS (at least three times for 20 min each over 1 h), and then incubated for 1 h at room temperature with the secondary antibody, washed several times in PBST, and mounted in a “slow-fade” buffer specific for immunofluorescence (Molecular Probes). ASH1 was detected by a rabbit polyclonal antibody (71) (affinity purified; kindly provided by Allen Shearn, Johns Hopkins University) at a dilution of 1:40, and dCBP was detected by a chicken polyclonal antibody raised against the CREB binding domain (CBD) of dCBP at a dilution of 1:800. FMRFamide (PT-2) and peptidylglycine-α-hydroxylating mono-oxygenase (PHM) rabbit polyclonal antibodies (30, 67) (kindly provided by Paul Taghert) were used at dilutions of 1:2,000 and 1:500, respectively. Fluorescein anti-rabbit (Vector) and rhodamine anti-chicken (Jackson) secondary antibodies were used at a dilution of 1/200. CNSs were examined with a confocal laser scan microscope (Bio-Rad 1024ES laser and Nikon Eclipse TE300 microscope).

The specificity of the dCBP antibody was determined in three ways. First, the staining of wild-type embryos was competed with increasing dosages of dCBD antigen. Second, the antibody failed to stain embryos that have mutant dCBP and that do not express dCBP RNAs. Third, the antibody detected a glutathione S-transferase (GST)-CBD fusion protein made from mammalian CBP in a dosage-sensitive manner.

Plasmid constructions. Various dCBP fragments (from pBSK-dCBP [2]) were inserted by PCR cloning into the pGEX-KG vector (Pharmacia) to generate pGST-dCBP-825-1043, pGST-dCBP-1699-1997, and pGST-dCBP 2274-2508, corresponding, respectively, to the CBD, BrZn, and C/H3 domains. The BglII restriction fragment of pBKS-ASH1 (kindly provided by Allen Shearn, Johns Hopkins University) was inserted into a pCITE vector (Novagen) for in vitro translation of the nearly full-length ASH1 (amino acids [aa] 49 to 2011). Other ASH1 fragments were inserted by PCR cloning into the pCITE vector for in vitro translation of ASH1-47-456, ASH1-458-853, ASH1-855-1255, ASH1-1639-2011, and the ASH1-SET domain (aa 1245 to 1525). Two N-terminal fragments of ASH1 were inserted by PCR cloning into the pGEX-KG vector to generate pGST-ASH1-47-456 and pGST-ASH1-458-853.

In vitro binding assays. GST fusion proteins were produced in Escherichia coli BL21 and purified by affinity on GST-agarose beads according to the Pharmacia protocol. In vitro-translated [³⁵S]methionine-labeled proteins were incubated with immobilized GST fusion proteins for 1.5 h at room temperature in Harlow buffer (50 mM HEPES [pH 7.5], 100 mM NaCl, 0.2 mM EDTA, 0.01 mM NaF, 1 mM dithiothreitol, 0.5% NP-40) with 5 mg of bovine serum albumin (BSA)/ml. After four washes in high-stringency Harlow buffer (containing 300 mM NaCl), bound proteins were eluted by boiling in SDS loading buffer, resolved by SDS-polyacrylamide gel electrophoresis, and visualized by autoradiography. Proteins from Kc cell nuclear extracts (65 μg of total protein) were incubated with immobilized GST fusion proteins using the procedure described above. In this case, the incubation was performed in the presence of 2.5 mg of BSA/ml instead of 5 mg of BSA/ml. For the E1A competition assays, proteins from Kc cell nuclear extracts were incubated with immobilized GST-ASH1-47-456 in the presence of 1, 4, or 12 μg of purified E1A or E1A-RG2 (kindly provided by J. Lundblad, Oregon Health Sciences University). This corresponds to approximately 50, 200, and 500 nmol of the E1A proteins, respectively. Incubations were done in the absence of E1A and E1A.RG2 as controls. The presence of bound dCBP from nuclear extracts was examined by Western blotting using the dCBP chicken polyclonal antibody at a dilution of 1/1,000. The Kc cell nuclear extracts were prepared as described previously (23).

Confocal analysis of polytene chromosomes. Polytene chromosomes were fixed and squashed for immunohistochemistry as described previously (83). ASH1 was detected by the ASH1 rabbit polyclonal antibody at a dilution of 1/40, and dCBP was detected by the chicken antiserum at a dilution of 1/800. Fluorescein anti-rabbit (Vector) and rhodamine anti-chicken (Jackson) secondary antibodies were used at a dilution of 1/200. Polytene chromosomes were examined by confocal laser scanning microscopy.

Dominant adult phenotypes obtained by overexpressing dCBP in specific cell types. Our goal was to obtain dominant adult phenotypes by overexpressing dCBP in specific cell types during development. To generate transgenic flies that overexpress dCBP in various tissues, we used the yeast GAL4 enhancer trap system (8). We generated two independent transgenic lines, Tr21 and Tr36, that carry the UAS-dCBP construct. These two lines are viable and show no visible phenotypes. We crossed the Tr21 and Tr36 transgenic lines to approximately 50 different enhancer trap GAL4-expressing lines and characterized the progeny in which the UAS-dCBP transgene is transcribed in a specific GAL4-dependent pattern. In most cases, overexpression of dCBP resulted in lethality at different stages of the development. A few GAL4 strains generated visible adult phenotypes with the Tr21-UAS-dCBP line (Table 1). Two of the dominant phenotypes had 100% penetrance: a smooth-eye phenotype using the GMR-GAL4 driver (this phenotype will be described in a separate paper) and an uninflated-wing phenotype using the GAL4-386 line (Fig. 1A).

TABLE 1 Pattern of GAL4 expression in larval and late pupal stages and the phenotypes obtained by crossing Tr21 (UAS-dCBP) with different GAL4 lines

FIG. 1 Uninflated-wing phenotype obtained by overexpressing dCBP in specific cells of the CNS using the GAL4-386 driver. (A) The uninflated-wing phenotype. (B to D) Expression of a LacZ reporter gene driven by GAL4-386 in larval tissues. β-Gal staining (more ...)

To confirm that the wing phenotype resulted from the overexpression of dCBP, we determined that a strong loss-of-function dCBP mutation, nej3, could suppress the phenotype. We also observed that the uninflated-wing phenotype is temperature sensitive. At 25°C, the uninflated-wing phenotype is completely penetrant, while only 40% of flies have the phenotype at 18°C. At 28°C, most of the flies die as pharate adults.

To identify the cells in which the GAL4-386 driver activates transcription, the GAL4-386 line was crossed to a UAS-LacZ line and β-Gal activity in the progeny was analyzed. No expression was detected during embryogenesis. During larval development, the UAS-LacZ reporter gene was expressed in specific cells in the CNS, both in the brain lobes and the ventral ganglion (Fig. 1B). Expression was also detected in the salivary glands, the fat body, and the gut; expression in these tissues was seen in many of the GAL4 lines tested in this study (data not shown). However, no expression was detected in other tissues, including the imaginal discs (Fig. 1C and D). Expression in the CNS begins in the third-instar larva and persists at least until the pharate adult stage. To test whether dCBP was effectively overexpressed in these cells throughout development, we performed confocal microscopy on larval and pharate adult tissues using a chicken polyclonal dCBP antibody. In wild-type flies, dCBP was expressed in every neuron in the CNS (Fig. 1E) and this expression was constant throughout development. In the GAL4-386/+; Tr21/+ flies, dCBP was clearly overexpressed in very specific neurons in the larval CNS (Fig. 1F and G). We counted approximately 50 neurons that overexpressed dCBP in the larval CNS, with about 30 located in the ventral ganglion. In pharate adults most of these cells still overexpressed dCBP, both in the brain and in different compartments of the thoracic CNS (Fig. 1H to J). Therefore, overexpression of dCBP in these neurons does not cause cell death as it does in other tissues with other GAL4 lines (our unpublished observation). By their positions and patterns, many of the neurons that overexpress dCBP appear identical to previously identified peptidergic neurons (45, 56). By using a neuropeptide antibody (PT-2) directed against a specific FMRFamide peptide (67) and an antibody for the neuropeptide biosynthetic enzyme PHM (30), we were able to show that some of these dCBP-overexpressing cells correspond to peptidergic neurons (Fig. 2).

Another GAL4 line, GAL4-363, also drives expression of the dCBP transgene in specific cells of the CNS (Table 1). When the Tr21-UAS-dCBP transgene is driven at 25°C with this GAL4 line, about 40% of the flies that eclose have an uninflated-wing phenotype (Table 1). This phenotype has a lower penetrance than the uninflated-wing phenotype described previously. However, overexpression of dCBP in wing tissues during development using the GAL4-30A and GAL4-71B drivers does not cause an uninflated-wing phenotype (Table 1).

The uninflated-wing phenotype obtained with GAL4-386 is a dominant visible adult phenotype with 100% penetrance at 25°C. Therefore, this phenotype was used to screen for suppressors of dCBP overactivity.

Screen for genes that can suppress the uninflated-wing phenotype caused by the overexpression of dCBP in specific CNS cells. The premise of the screen was that a twofold reduction in the dose of a gene that is involved in dCBP signaling may suppress the uninflated-wing phenotype that results from dCBP overexpression. Using the Bloomington Stock Center deficiency kit that covers at least 70% of the genome, we screened for deletions on the second and third chromosomes that are dominant suppressors of the uninflated-wing phenotype. The strategy of the screen is represented in Fig. 3 and is described in Materials and Methods. A suppression was considered positive when at least 50% of the flies had a suppressed phenotype (partially or completely inflated wings). Deletions identified as potential suppressors were rescreened at least twice to confirm the interactions. We obtained 10 different deletions in various portions of the genome (six on the second chromosome and four on the third chromosome) that suppress the uninflated-wing phenotype (Table 2). We expected two types of suppressors from our screen: first, deletions that reduce the level of dCBP production, either by reducing GAL4 expression or by affecting the expression of the transgene, and second, deletions of genes that affect dCBP function, either directly or indirectly and in a dosage-sensitive manner.

FIG. 3 Screen for deletions that can dominantly suppress the dCBP overexpression wing phenotype. Males carrying deficiencies for the second [Df(2)] and the third [Df(3)] chromosomes over a Balancer (B) were crossed to GAL4-386/MKRS (more ...)

To eliminate the first class of suppressors, we measured the effect of the deletions on the expression of a UAS-LacZ reporter transgene during pupal stages using a liquid β-Gal assay (see Materials and Methods). Of the 10 deletions identified in the screen as suppressors of the dCBP overexpression phenotype, 3 [Df(2R)cn9, Df(2R)AA21, and Df(3L)VW3] reduced the expression of the UAS transgene and were discarded (Table 2). The remaining seven deletions were considered to be suppressors of dCBP function. The results from the liquid β-Gal assays were confirmed for each deletion by comparing the levels of expression of β-Gal in fixed larval brains stained with X-Gal (see Materials and Methods). Interestingly, Df(2L)C144 and Df(3L)vin7 were also able to suppress the smooth-eye phenotype caused by the overexpression of dCBP in the eye disc (data not shown), suggesting that different tissues may share common features of dCBP signaling. The remaining deletions did not modify the smooth-eye phenotype, suggesting that dCBP might also utilize different pathways and interact with different protein complexes in the two tissues tested.

We then searched for genes within the deletions that, when hemizygous, suppress the wing phenotype. By analyzing the fly data base, we identified approximately 100 lines with null mutations in known genes or lethal P-element insertions contained within the remaining seven deletions. These lines were then screened for their ability to suppress the uninflated-wing phenotype. In five of the deletions, the genetic element that causes the suppression remains to be identified (Table 2). However, we characterized two genes in two deletions that, when mutant, act as dominant suppressors of the uninflated-wing phenotype: the PKA-C1/DCO gene encoding the catalytic subunit of PKA, and ash1, a member of the trithorax family. Thus, as observed in mammalian cells (14), PKA seems to play an important role in dCBP signaling. Furthermore, expression of a dominant-negative catalytic subunit of PKA under the control of a UAS promoter (UAS-R* [37]) also suppressed the wing phenotype observed in the GAL4-386; UAS-dCBP flies (data not shown). Further characterization will be necessary to determine if the action of PKA in dCBP signaling is direct or indirect.

Ash1 is a TrxG gene defined by its structural and functional relationship to the Drosophila homeotic regulator trithorax (71). Members of the trithorax family are thought to be transcriptional activators that maintain chromatin in an “open” configuration. When heterozygous, a null allele of ash1, ash1²², suppressed the phenotype almost as well as Df(3L)JK18 (Table 3). More than 75% of the flies containing ash1²² in the presence of dCBP overexpression had partially inflated or wild-type wings (Fig. 3 and Table 3). To confirm the characterization of ash1 as a suppressor of the dCBP overexpression phenotype, we examined other ash1 mutations. ash1¹ and ash1¹¹, which are described as strong hypomorphs, also suppressed the phenotype (Table 3). It is important to note that all three alleles came from different sources. Thus, the observed effect on the wing phenotype is very likely due to the loss of ash1 gene function rather than an unknown modifier on the chromosome. Two weaker alleles of ash1, ash1²⁹, a hypomorphic allele, and ash1¹⁴, a heat-sensitive hypomorph, did not suppress the phenotype (Table 3). In these cases, the products of both alleles probably retain some degree of function and might still contribute to the effect of dCBP overactivity.

It is interesting that ash1 was the only trithorax member, among the ones tested, capable of suppressing the dCBP wing phenotype. The ash2, brahma, trithorax, and trithorax-like alleles were not able to suppress the phenotype (Table 3). As expected, an allele of Polycomb was similarly unable to suppress the phenotype. These results indicate that the genetic interaction between ash1 and dCBP is very specific and was only apparent because we were examining dCBP signaling in very specific cells.

Specific expression of ASH1 protein in the same CNS cells that overexpress dCBP. To understand more precisely the relationship between dCBP and ASH1, we analyzed their localization at the cellular level. Using confocal laser microscopy, we first localized the expression of ASH1 protein in the CNS when dCBP is expressed under the control of GAL4-386. For this study, we used an ASH1 rabbit polyclonal antibody and the dCBP chicken polyclonal antibody produced in our laboratory. ASH1 was expressed in specific neurons in the larval ventral ganglion (Fig. 4A) and the pupal and adult thoracic ganglion (Fig. 4D, G, and J). No expression of ASH1 was detected in the brain, either in third-instar larvae, late pupae, or adults (data not shown). dCBP, which is normally expressed in every neuron of the CNS, was overexpressed in specific neurons when placed under the control of the GAL4-386 driver (Fig. 4B, E, H, and K). The overexpression of dCBP is seen both in the brain and the ventral ganglia at larval, pupal, and adult stages. We then determined whether dCBP and ASH1 were colocalized. In the ventral ganglia of third-instar larvae, approximately two-thirds of the overexpressing dCBP neurons were ASH1 positive (Fig. 4C). The same high level of coexpression was observed in the thoracic CNSs of pharate adults before eclosion and of adults 24 h posteclosion (Fig. 4F, I, and L). In the prothorax and mesothorax, 12 neurons that overexpressed dCBP expressed ASH1 as well (Fig. 4F). In the metathorax and the abdominal ganglion, about half of the dCBP-overexpressing cells were ASH1 positive (Fig. 4I), representing also about 12 neurons. Colocalization of ASH1 and dCBP overexpression was still observed in the 12 ASH1-positive cells that persisted 24 h after eclosion (Fig. 4L).

FIG. 4 ASH1 expression coincides with dCBP overexpression in the CNS of GAL4-386/+; Tr21/+ larvae and pupae. An analysis by laser confocal microscopy was performed. (A to C) Colocalization in the thoracic region of a third-instar larval ventral (more ...)

It is conceivable that the overexpression of dCBP in our system could result in ASH1 misexpression. Therefore, we compared the pattern and the level of CNS expression of ASH1 in dCBP-overexpressing flies with those in wild-type flies (Fig. 5A to F). Similar patterns and levels of expression were obtained in larvae of both genotypes (Fig. 5, compare panels A and D and panels C and F from merged images) in pupae, and in posteclosion adults (data not shown). Thus, the overexpression of dCBP does not modify the expression of ASH1. We also examined the colocalization of the two proteins in a wild-type larval ganglion under higher magnification (Fig. 5G to I). Although ASH1 and dCBP proteins overlapped in expression, the two proteins did not have precisely the same localization pattern. dCBP was detected only in the nucleus, while ASH1 was present both within the nucleus and also around the nuclear border. Importantly, overexpression of dCBP did not modify this characteristic ASH1 distribution pattern (Fig. 5, compare panels A and D and panels C and F).

FIG. 5 CNS expression of ASH1 in dCBP-overexpressing or wild-type third-instar larvae, as analyzed by laser confocal microscopy. (A to C) Expression of ASH1 and dCBP in the larval ventral ganglion by dCBP-overexpressing cells. (A) ASH1 immunostaining with fluorescein (more ...)

We previously used the β-Gal expression test to show that neither the deletion containing ash1 (Table 2) nor the amorphic ash1 mutation (data not shown) affected expression of a UAS-LacZ reporter gene. However, the regulation by proteins involved in chromatin organization, such as trxG proteins, can be sensitive to position effects. Because the UAS-LacZ and UAS-dCBP constructs are not inserted at the same position in the genome, it was necessary to verify that an ash1 loss-of-function mutant, when heterozygous, has no effect on the level of expression of the UAS-dCBP transgene. GAL4-386/+; Tr21/+ females with uninflated wings were crossed to ash1²²/TM6B,Tb males. Out of this cross, non-Tb larvae were selected (one-fourth of this population will be GAL4-386/ash1²²; Tr21/+). The level of dCBP overexpression in the ventral ganglion was analyzed for these larvae and compared to the level of dCBP overexpression in GAL4-386/+; Tr21/+ flies. No significant variation was observed when the flies were heterozygous for ash1²² (Fig. 6). Therefore, the suppression observed with the ash1 mutations does not result from a reduction in the overexpression level of dCBP.

FIG. 6 dCBP overexpression in a wild-type or ash122 heterozygous genetic background. Shown is an analysis by laser confocal microscopy of dCBP normal expression and overexpression in the ventral ganglia of third-instar larvae. (A) dCBP ubiquitous expression (more ...)

Direct association between dCBP and ASH1. Because both dCBP and ASH1 are involved in transcription and appear to interact at both the genetic and cellular levels, we asked whether dCBP and ASH1 could interact biochemically.

We tested direct binding between ASH1 and dCBP in vitro with pull-down assays. First, we used in vitro-translated [³⁵S]methionine-labeled ASH1 polypeptides and various fragments of dCBP expressed as GST fusion proteins and immobilized on GST-Sepharose beads. Our results show that a polypeptide corresponding to a nearly full-length ASH1 (aa 47 to 2011) interacts strongly with GST-dCBP-2278-2678 containing the C-terminal C/H3 domain (Fig. 7A). No binding of ASH1 to GST fusion proteins containing the CBP CBD (GST-dCBP-825-1043) or the bromo-zinc finger domain (aa 1699 to 1997) was observed. Different domains of ASH1 were used to show that the N-terminal regions of ASH1 (domains containing aa 47 to 456 and aa 458 to 853) and the SET domain (aa 1245 to 1525) are responsible for the interaction with GST-dCBP-2278-2678 (Fig. 7B and C). Unlike the nearly full-length ASH1, the ASH1-458-853 polypeptide and the SET domain bind with the GST fusion protein containing the dCBP bromo-zinc finger domain (Fig. 7A to C). Thus, these results indicate that ASH1 interacts predominantly with the C/H3 domain of dCBP through both its N-terminal region and its SET domain.

FIG. 7 dCBP interacts with ASH1. (A) Equimolar amounts of immobilized GST and GST-dCBP fusion proteins were incubated with in vitro-translated 35S-labeled ASH1 protein (nearly full-length ASH1 protein; aa 49 to 2011). (B) Equimolar amounts of immobilized GST (more ...)

The two N-terminal regions of ASH1 that mediate the interaction with dCBP were also expressed as GST fusion proteins and immobilized on GST-Sepharose beads. Using nuclear extracts from Drosophila Kc cells, we showed that dCBP is retained on GST-ASH1(47-456) and GST-ASH1(458-853) proteins but not on the GST protein alone (Fig. 7D). These interactions are specific, because an E1A polypeptide that binds to the ASH1 binding region in dCBP blocks the GST-ASH1(47-456)-dCBP interaction more efficiently than an E1A-RG2 polypeptide that carries a mutation that reduces E1A binding to dCBP (Fig. 7E).

An antibody against a fragment of the ASH1 protein identifies approximately 100 sites on polytene chromosomes (71). A chicken polyclonal antibody directed against dCBP also detects specific sites on polytene chromosomes (Fig. 8). Several sites were found to stain both ASH1 and dCBP, suggesting that dCBP and ASH1 may cooperate in the transcription of specific genes.

FIG. 8 Endogenous dCBP and ASH1 proteins colocalize on wild-type polytene chromosomes. Shown is an analysis by laser confocal microscopy. (A) Chromosome arm showing localization of ASH1. (B) Chromosome arm from panel A showing localization of dCBP. (C) Merged (more ...)

Taken together, the binding assays and the colocalization at specific sites on the polytene chromosome strongly support the idea that dCBP and ASH1 can be part of the same transcriptional regulatory complex in vivo.

Previous studies have shown that CBP affects transcription through interactions with components of the basal transcriptional machinery and through its intrinsic and associated acetyltransferase activities. In this report, we used a genetic approach in Drosophila to further examine the in vivo function of dCBP. Overactivity of dCBP in particular cell types causes several distinct adult phenotypes. By screening for deletions that could suppress one dCBP overexpression phenotype, we identified ASH1, a member of the trithorax group of chromatin modifiers, as a potential interacting partner of dCBP. ASH1 and dCBP colocalize to a subset of CNS neurons and to specific bands in polytene chromosomes. Furthermore, dCBP and ASH1 interact specifically at the molecular level. Our genetic and biochemical analyses link dCBP to a second class of proteins involved in epigenetic gene regulation.

Screen for suppressor genes of the uninflated-wing phenotype. Despite the fact that the dCBP-overexpressing flies with an uninflated-wing phenotype were weak and difficult to culture, they were fertile and could be used for our screen. The use of the deficiency kit allowed us to rapidly define regions of the genome that contain genes capable of influencing the effect of dCBP overexpression. In five deletions, the genetic elements that cause the suppressions remain unknown and must be characterized. We anticipate that the sequencing of the entire Drosophila genome and the generation of more P-element mutations in these regions will help us to identify novel dCBP interactors.

In this report, we identified two genes in two deletions that, when hemizygous, suppress the uninflated-wing phenotype. The first suppressor gene is the PKA-C1/DCO gene encoding the catalytic subunit of the PKA. Interestingly, PKA has been involved in different pathways that require dCBP activity, both in mammalian cells and in Drosophila. In Drosophila, PKA negatively regulates the hedgehog (hh) signal transduction cascade by phosphorylating CI, the transcription factor that transduces the hh signal into the nucleus. In the absence of an hh signal, the phosphorylated CI is proteolyzed to a repressor form of the protein and can no longer be activated by dCBP (5, 10, 11). In this case, PKA has a negative and indirect effect on dCBP-mediated CI activation. Here, PKA seems to have a positive effect on dCBP signaling and may play a role in the signaling pathway that regulates wing inflation. It is possible that PKA is directly involved in dCBP phosphorylation as proposed by Xu et al. (80). However, PKA may regulate the proteins that interact with dCBP or transduce a pathway that acts in parallel with the dCBP pathway. Further characterization will be necessary to determine if the action of PKA in dCBP signaling and wing inflation is direct or indirect. The second suppressor gene detected in our screen is ash1, a member of the trithorax group genes (trxG).

Suppressing effect of ash1 on the overactivity of dCBP. Lethal mutations of ash1 cause homeotic transformations of imaginal disc-derived structures (58). Gene ash1 (absent, small, or homeotic discs 1) is a member of the trxG genes that encode a variety of proteins with different biochemical properties that are thought to play an important role in modulating chromatin structure during development. Therefore, the identification of ash1 as a possible effector of dCBP function suggested a novel role for dCBP in regulating gene expression. Amorphic or strong hypomorphic alleles of ash1 that are thought to retain very little function (70, 71), suppress the dCBP overexpression phenotype. However, two other alleles, ash1²⁹ and ash1¹⁴, had no effect on the dCBP uninflated-wing phenotype. ash1²⁹ is a weak hypomorph that retains some degree of function, while ash1¹⁴ is a temperature-sensitive allele (70) that certainly retains most of its function at the temperature of our assays (25°C). These results indicate that the dCBP-ASH1 interaction, whether direct or indirect, is dosage sensitive. One of the deletions isolated as a suppressor contains gene ash2, another member of the trx family. The ash1 and ash2 genes are functionally related (57), but their gene products are structurally divergent (1, 71). Despite the fact that they belong to the same family, a strong mutation in ash2 does not suppress the dCBP overexpression phenotype. It is interesting that none of the other trxG genes tested in this study (brahma, trithorax, and trithorax-like) were capable of suppressing the wing phenotype. The specificity of the interaction could be explained by the restricted range and action of trxG genes in different cell types. Alternatively, it could be that the interaction of dCBP with other members of the trxG genes is not dosage sensitive in these neurons.

Double heterozygotes of recessive alleles of ash1 and brahma have a high penetrance of homeotic transformations in specific imaginal disc- and histoblast-derived tissues (70). However, double heterozygotes of various recessives alleles of ash1 and dCBP do not show any homeotic transformations. Likewise, mutations in dCBP do not enhance the homeotic transformations seen in the homozygous viable ash1¹⁴ adults (F. Bantignies, unpublished observations). In this case, it is possible that one dose of the dCBP gene is sufficient to maintain the ash1 function. It is also possible that dCBP and ash1 do not interact in the tissues which are sensitive to ash1-mediated homeotic transformations. Recently, Florence and McGinnis (18) generated antimorphic alleles of dCBP that enhance hypomorphic mutations in the homeotic gene Deformed (Dfd). The antimorphic dCBP mutations also enhance mutations in the homeotic gene Ultrabithorax (Ubx). The null alleles of dCBP did not affect Dfd hypomorphs. Nor do the null alleles of dCBP enhance or suppress mutations in Ubx, Sex combs reduced (Scr), Antennapedia (Antp), abdominal A (abdA), Abdominal B (AbdB), or the Polycomb group genes (S. M. Smolik, unpublished observations). None of the dCBP recessive phenotypes include homeotic transformations. These results suggest that any involvement of dCBP in homeotic gene function is not dosage sensitive and can only be detected with mutations that actively interfere with wild-type function.

ASH1 expression coincides with the overexpression pattern of dCBP. The set of specific cells in the larval ventral ganglion and pupal thoracic ganglion that express ASH1 is a subset of the cells that overexpress dCBP in the GAL4-386 line. In the thoracic region of the larval ventral ganglion as well as in the prothorax and mesothorax of the pupal thoracic CNS, all of the ASH1-positive cells colocalized with dCBP-overexpressing cells. Some of the dCBP-overexpressing cells express the neuropeptide markers FMRFamide and PHM (30, 45, 56) and from their positions are likely to express ASH1 as well. This result suggests that dCBP and ASH1 could have common functions in peptidergic neurons.

ASH1 is required for the proper differential activation of Ubx and probably other genes in the larval ventral ganglion (35). Therefore, dCBP and ASH1 might regulate the function of homeotic genes as well as other developmental genes in specific CNS cells.

We also show colocalization of dCBP and ASH1 in the nuclei of specific neurons of the wild-type CNS, which strongly reinforces the biological significance of our observations. It is intriguing that, while dCBP is only nuclear, ASH1 is present in both the nucleus and the cytoplasm surrounding the nuclear periphery. Preparation of nuclear and cytoplasmic Kc cell extracts revealed the presence of the ASH1 protein in both compartments (data not shown). At this time, we do not know the significance of this pattern of localization, but it is possible that ASH1 nuclear localization might be regulated through posttranscriptional modifications.

Screens for enhancers and suppressors of overexpression phenotypes have been useful in identifying components of regulatory pathways. Nevertheless, overexpression systems have drawbacks and can potentially identify secondary effectors of a nonspecific phenotype. However, we believe that this screen has identified genes that affect dCBP function for several reasons. First, the number of deficiencies that suppress the uninflated-wing phenotype is small. A large number of suppressors might suggest that the overexpression of dCBP was not eliciting a specific cell phenotype. Second, two of the deletions suppressed both the wing and the eye overexpression phenotypes, suggesting that the overexpression of dCBP in the two tissues has some common effects. One of the deletions demonstrated that the dosage of PKA could affect the dCBP overexpression phenotype. CBP and dCBP are known to play a role in PKA signaling, so the fact that PKA was identified in this screen is consistent with the idea that dCBP overexpression reflects an overactivation of the PKA pathway. We have ruled out trivial explanations for the suppression of dCBP overexpression by ASH1; dCBP overexpression does not cause the death of ASH1-expressing cells, nor do ash1 mutations affect the overexpression of dCBP. A characterization of dCBP loss of function in these cells both in wild-type and ash1 mutant backgrounds is necessary to complete this analysis. A clonal analysis of dCBP mutant cells is not feasible because dCBP is required for cell viability and only small clones can be generated. This analysis will have to await reagents that allow us to knock out dCBP function in the GAL4-386 cells in the ash1 mutant background. In addition, it will be important to identify the targets of dCBP and ASH1 in these cells as well as the pathways that activate them. Although the genetic analysis is not complete, it is likely that the genetic suppression of dCBP overexpression by ash1 mutations reflects a functional association between ASH1 and dCBP because these two proteins have specific interactions in vitro.

Overexpression of dCBP in specific CNS cells causes wing inflation defects. In many tissues, overexpression of dCBP causes lethality, suggesting that the dose of this effector is important for its function. The overproduction of dCBP in specific cells of the CNS with two different GAL4 lines produced defects in wing inflation with various degrees of penetrance. However, overexpression of dCBP in wing tissues throughout development does not interfere with wing inflation.

Previous studies have implicated specific CNS cells in the regulation of wing inflation. In Drosophila, the death of specific cells is triggered after eclosion and is strongly correlated with wing inflation behavior (29). In addition, two specific neurons in the fly brain are responsible for the production of the neuropeptide eclosion hormone (EH). The specific knockout of EH-producing cells (EH cells) during early development results in eclosion delays and a disruption of eclosion behaviors, such as wing inflation (41). In the moth Manduca sexta, EH triggers a neuroendocrine cascade that regulates both ecdysis and postecdysis processes such as wing inflation. It was suggested that the frequent failure of EH cell knockout flies to inflate their wings successfully is due to a lack of excitability of neuroendocrine-responsive EH cells that release important signals for proper eclosion behaviors (41). In Manduca, different neuropeptides, such as bursicon and the cardioacceleratory peptides, are usually released after eclosion to aid in wing expansion (72, 75, 76). It may be that the neurons which overexpress dCBP are the neurosecretory cells that are targeted by the EH cascade and that produce the peptides that signal the wing inflation process. In this case, the overexpression of dCBP interferes with normal cell function. Of course the wing inflation defect could be due to the death of the neurons caused by the overexpression of dCBP. However, the pattern of cells that overexpress LacZ and dCBP in the GAL4-386 background remains the same throughout development, and cells that overexpress dCBP and express ASH1 are viable at least 24 h posteclosion, so the overexpression of dCBP does not appear to affect the viability of these cells. Two additional GAL4 lines, GAL4-c929 and GAL4-c191, also drive specific expression in the CNS, specifically in most of the peptidergic neurons of the brain and ventral ganglion (R. Hewes and P. Taghert, personal communication). At 25°C, escapers were obtained only with the GAL4-c191 line. Approximately 30% of these flies have uninflated or partially inflated wings.

We propose that the overexpression of dCBP in specific CNS cells affects the regulation of signaling pathways that involve dCBP and that are important for proper eclosion behaviors. Our preliminary data suggest that at least some of the cells that overexpress dCBP are neuropeptidergic neurons and colocalize with the neuropeptides FMRFamide and PHM. However, antibody incompatibility does not allow us to determine whether these cells also express ASH1. Clearly, more characterization will be required to determine the exact pathways affected by dCBP. The dominant wing phenotype obtained by overexpressing dCBP with GAL4-386 is a good model to elucidate some of the cells and signaling pathways involved in wing inflation.

Specific interaction between ASH1 and dCBP. Our biochemical experiments show that coactivator dCBP binds strongly to trxG protein ASH1. This observation supports the idea that ASH1 and dCBP interact in vivo and implicates a novel class of chromatin binding proteins in mediating dCBP function.

The ASH1 protein contains three motifs that are characteristic of some proteins that regulate transcription and/or are bound to chromosomes: there are two AT hook motifs in the N-terminal region, a SET domain, and a PHD finger in the C-terminal domain. The AT hook motif is important for the binding of some proteins to DNA (52). PHD fingers are Cys-rich Zn finger-like motifs implicated in protein-protein interactions and are found in other trxG proteins (1, 40, 70). The SET domain is an approximately 130-aa region found in a number of other chromatin-associated proteins, including the TRX factor (40), PcG protein Enhancer of Zeste [E(Z)] (26), and the modifier of position effect variegation Su(Var)3-9 (73). The TRX SET domains have been proposed to mediate association with components of chromatin-remodeling complexes (54), and ASH1 and TRX interact directly through their SET domains (55). Our binding assays indicate that two N-terminal regions and the SET domain of ASH1 interact strongly with dCBP. However, no interaction with the PHD domain was observed. Thus, the SET and the PHD domains of ASH1 might function for the recruitment of other chromatin-associated proteins, such as TRX, and the N-terminal region could serve to interact with the DNA, possibly through the AT motifs, to direct the targeting of HATs to the promoter. Further biochemical characterization will be necessary to confirm this model, but the interaction between dCBP and ASH1 provides new insights on the possible function of ASH1 in gene regulation.

The binding of ASH1 to dCBP requires the C-terminal C/H3 domain. In mammalian CBP and p300, this region mediates interactions with numerous sequence-specific transcription factors, the adenovirus E1A protein, TFIIB, RNA helicase A, and P/CAF, a GCN5-like histone acetylase (19, 25, 60). In dCBP, the C/H3 domain mediates the interaction with transcription factor dTCF and Mad (78, 79), demonstrating an important role for this domain in dCBP function. Our findings reveal that this domain contributes to the interaction with chromatin-associated protein ASH1, suggesting that dCBP may function in epigenetic regulatory complexes. The C/H3 domain is adjacent to HAT and might contribute to the regulation of the histone acetylation activity of CBP and p300 (32) or might recruit targets of acetylation close to the enzymatic domain. Thus, it will be interesting to determine whether ASH1 has any effect on dCBP HAT functions or if it is a target of dCBP acetyltransferase activity.

The recent work of Dhalluin et al. (15) has shown that the bromodomain of P/CAF binds histone peptides in an acetylation-dependent manner. The bromodomain of GCN5, a member of the SAGA complex, is required for SWI/SNF remodeling of the nucleosome and stabilizing the SWI/SNF complex on the promoter (66). Thus, it appears that the bromodomain interacts with acetylated proteins and may form a link between different regulatory complexes. Although the full-length ASH1 does not interact with the bromodomain of dCBP, both the ASH1-458-853) polypeptide and the SET domain do interact with this domain. It may be that full-length ASH1 undergoes a modification, upon binding with the dCBP C/H3 domain, that allows other regions of ASH1 to interact with the dCBP bromodomain. In this case, it would appear that the interaction is not dependent on acetylation.

Our results also show that dCBP and ASH1 colocalized to a number of specific sites on polytene chromosomes, suggesting that they might serve as coregulators of a specific set of genes including the homeotic selector genes. The mapping of the specific sites where dCBP and ASH1 colocalize will help us to identify target genes that are regulated by ASH1 and dCBP. An analysis of these genes, their promoters, and their regulation by dCBP and ASH1 will further define the functional role of the dCBP-ASH1 interaction.

It has been recently shown that ASH1 is a component of a large-molecular-weight complex (47). The components of this ASH1 complex have not been identified, and it will be interesting to test whether it includes dCBP. BRM, another trxG member, is also contained in a large-molecular-mass complex of 2 MDa. Four of the subunits of the BRM complex are related to subunits of the yeast chromatin-remodeling complexes SWI/SNF and RSC (remodels the structure of chromatin) (47), suggesting that trxG proteins are important regulators of chromatin structure. However, no other trxG members are present in this complex, suggesting that each trxG member is involved with different complexes and probably has divergent functions. Thus, it might be anticipated that ASH1 could be the only trithorax member that affected the overexpression of dCBP function.

Our genetic approach allowed us to characterize a specific cellular and biochemical interaction between dCBP and ASH1, a member of the trithorax group of chromatin modifiers. This finding may provide important new insights into the functions of both proteins. trxG proteins are thought to be important components of chromatin-remodeling complexes, and our study provides evidence that they might also be involved in the recruitment of transcriptional activators. While the molecular mechanism of the interaction between dCBP and ASH1 is not known, it probably involves the modification of chromatin structure, and this suggests that CBP may not only affect the nucleosome but may also be involved in the regulation of higher-order chromatin structure as well.

We thank A. Shearn for the ash1 and ash2 mutants, ash1 cDNA, and ASH1 antibody; we are grateful to J. Lundblad for the E1A and RG2.E1A proteins. We also thank D. Kalderon for the Pka mutant and transgenic flies; C. O'Kane and A. Brand for GAL4 lines; G. Rubin for the GMR-GAL4 line; J. Roote for the pk-sple³³ pr cn/T(2;3)SM6.TM6B,Tb balancer chromosome strain; and R. Hewes and P. Taghert for GAL4 lines, the FMRFamide, and PHM antibodies and for sharing unpublished information. We also thank the Bloomington and the Umea Drosophila Stock Centers for providing numerous stocks. We are very grateful to A. Snyder (MMI department, OHSU, and the Oregon Hearing Research Center) for confocal analysis.

This work was partly supported by grants from the Association pour la Recherche contre le Cancer and the National Institutes of Health (DK4Y239).