Modulation of a Transcription Factor Counteracts Heterochromatic Gene Silencing in Drosophila

Cell, Vol. 104, 839–847, March 23, 2001, Copyright 2001 by Cell Press

Modulation of a Transcription Factor Counteracts Heterochromatic Gene Silencing in Drosophila Kami Ahmad and Steven Henikoff* Howard Hughes Medical Institute Fred Hutchinson Cancer Research Center Seattle, Washington 98109

Summary Variegation is a common feature of gene silencing phenomena, yet the basis for stochastic on/off expression is unknown. We used a conditional system that allows probing of heterochromatin at a reporter GFP gene by altering GAL4 transcription factor levels during Drosophila eye development. Surprisingly, the frequency of gene silencing is exquisitely sensitive to GAL4 levels, as though binding site occupancy affects the silenced state. The silent state is plastic, as spontaneous derepression occasionally occurs in both mitotically active and differentiating cells. By simultaneously assaying expression of a nearby gene, we further show that the size of an activated region within heterochromatin is small. We propose that variegation occurs because heterochromatin inhibits the transient exposure of factor binding sites. Introduction Mutagenesis has shown that nearby regulatory elements and the position of a gene along the chromosome modulate its expression. Many regulatory elements bind proteins involved in transcription, and act in concert to activate or repress a promoter. The effect of chromosomal position on gene expression is less well understood. Some cases of position effect result from the placement of a gene close to regulatory sites, but others are thought to reveal the effects of distinctive chromatin structures on genes. The silencing phenomenon referred to as position-effect variegation (PEV) in Drosophila melanogaster fits in this latter category. PEV can be observed when a euchromatic gene is juxtaposed to heterochromatin. Mosaic silencing is perhaps the most enigmatic feature of PEV and other epigenetic phenomena. To account for mosaicism, cell-to-cell variations in heterochromatic features have been proposed, with inactivation or expression indicating whether a heterochromatic structure has encompassed the gene (Tartof et al., 1989). Heterochromatin is condensed throughout the cell cycle and has distinct protein components, and in the heterochromatic state, a gene is thought to be inaccessible to transcription factors (Spofford, 1976; Henikoff, 1990; Elgin, 1996). Cell-to-cell variations as the basis for on/off expression is supported by correlations between gene silencing and variations in salivary gland chromosome morphology (Zhimulev et al., 1986), and subnuclear positioning in diploid cells (Henikoff, 1997; * To whom correspondence should be addressed (e-mail: steveh@ fred.fhcrc.org).

Bonifer, 2000), but examples showing no correlation have also been reported (Zhimulev et al., 1988; Hayashi et al., 1990; Sass and Henikoff, 1999). Thus, the basis of mosaic gene silencing remains unclear. We describe here a novel system for probing a silenced gene. Activity of a GAL4-dependent promoter near heterochromatin is altered by turning on GAL4 activator at different times in development. We find that promoter activation counteracts silencing at this and at neighboring genes, depending on the degree and timing of GAL4 synthesis. Live analysis of silencing in cultured imaginal discs reveals that derepression can occur spontaneously in both mitotically active and in differentiating cells. To explain these results, we propose that heterochromatin alters the rates of factor binding site exposure, and thereby reduces the occupancy of limiting transcription factor binding sites. At any one moment, binding sites in only some cells will be occupied, giving mosaic expression; more factor drives the occupancy of sites, giving more frequent expression. In this way, a bistable equilibrium underlies variegation. Results A Novel System for Probing Heterochromatin during Drosophila Development Our experimental rationale is outlined in Figure 1A. We started with a fly line carrying a P transposon containing a set of 5 tandem GAL4 binding sites upstream of a green fluorescent protein (GFP) gene (UASGFP) inserted at cytological position 79D-E. The insertion at this euchromatic site serves as a control. A mini-white gene on the transposon provided a reporter that can be used to select for position-effect variegation mutants. We used X rays to induce chromosomal rearrangements, and recovered progeny showing white variegation in the eyes. Two lines were established, x18 and x21, each of which placed the transposon insertion next to heterochromatic Y chromosome sequences, consistent with the notion that mini-white is subject to PEV in these lines. We subsequently mobilized the P element in the x18 line with transposase, which can generate more severe variegators (Sabl and Henikoff, 1996). Variegation of mini-white in x21, x18, and all transposase-induced derivatives of x18 was suppressed by elevated temperature or the presence of a Su(var)205 mutation, as is typical of PEV (data not shown). In a line where one gene in a P transposon variegates, it is likely that other genes in the construct will as well (Roseman et al., 1993, 1995; Wallrath and Elgin, 1995; Lu et al., 1996; Wines et al., 1996). To determine whether the UASGFP gene adjacent to mini-white in the lines generated here also variegates, we crossed each line to a GAL4-producing line to activate the UAS promoter of the reporter gene (Brand and Perrimon, 1993). We used the GMRGAL “driver” (Figure 1B), which produces GAL4 in the posterior portion of the eye imaginal disc (Freeman, 1996), and examined the eye discs of third instar larvae. The parental (euchromatic control) line

Cell 840

Figure 1. Rationale for Probing Heterochromatin with GAL4 Activator The target transposon carries two reporter genes (green and red lines), and shows their respective transcripts (thin lines). GAL4 is produced from a separate transposon source. The control target insertion is located in euchromatin, and can be activated by GAL4. Variegator derivatives of this insertion are juxtaposed to heterochromatin (red wavy line). (B) Relative timing of activation by GAL4 driver lines. The top line indicates the developmental stages of Drosophila development (e: embryos, L1-L3: larval stages, and P: pupal stages), and the lower lines the expression profile of each driver.

gives uniform expression of GFP in the posterior portion of the eye disc, as expected for this driver (Figure 2). GMRGAL in combination with each of the mini-white variegator lines caused mosaic expression of UASGFP, demonstrating that this gene is subject to PEV (Figure 2—GMRGAL, and data not shown). Derepression by Early Transcriptional Activation We used GAL4 drivers that begin expression at different times to assay the importance of activation for variegation of a single reporter gene (Figure 1B). We focused on variegation in the eye, since numerous drivers that express in this organ are available. Cells that will ultimately give rise to the adult eye are set aside during embryogenesis and multiply during the larval stages in the eye-antennal imaginal disc (Ready et al., 1976). During the third instar larval stage, cells cease dividing and begin differentiating. Differentiation begins at the posterior end of the disc and proceeds anteriorly, and the boundary between the differentiating and the mitotically active regions is visible as the morphogenetic furrow (MF). One final mitotic wave follows closely behind the MF as it moves anteriorly. The GMRGAL driver that we used above begins producing GAL4 protein immediately behind the MF, and thus the variegation that we observed was principally in terminally differentiating cells. If the time at which a promoter is first activated affects its susceptibility to PEV, then a UASGFP variegator line should show more complete expression when activated by an early GAL4 driver than by the late GMRGAL. We

Figure 2. Derepression of UASGFP Variegation in the x21 Variegator Lines Depends on the Timing of the GAL4 Driver Eye-antennal discs from L3 larvae are oriented with the anterior and mitotically active region on the right, and the posterior differentiating region on the left. Arrowheads mark the position of the morphogenetic furrow. GFP fluorescence is in green, and the counterstained tissue is in gray.

crossed flies carrying the A5CGAL driver, which begins producing GAL4 in embryos, to each variegator line. In the eye disc, GFP is uniformly expressed in the control line (Figure 2—A5CGAL). Expression in the x21 line was also uniform, with only occasional and minor mosaicism. To quantitate the incidence of gene silencing for each combination of target and driver, we measured the area of eye discs behind the MF in which UASGFP was not expressed. Eye discs from larvae carrying the x21 target and GMRGAL showed 74%–96% silencing, while discs from larvae carrying the same target and A5CGAL showed 0%–6% silencing. Thus, many more cells express UASGFP in variegator lines when driven by A5CGAL. Derepression of variegation by the A5CGAL driver suggests that activation of UASGFP early in development prevents later silencing events. We examined UASGFP expression in combination with other drivers to confirm this dependence on activation timing. UASGFP variegation in the x21 line was derepressed by another early-activating driver, GawBT80 (Figure 2). Silencing with GawBT80 was reduced to 25%–38% of the posterior part of the eye disc, consistently more than that with

GAL4 Activator Counteracts Variegation 841

Figure 3. A Weak GAL4 Driver Fails to Derepress UASGFP Variegation Each target line (euchromatic control or variegator) was crossed to the strong A5CGAL or to the weak armGAL driver, and eye discs were examined from L3 larvae.

A5CGAL (Figure 2). GawBT80 initiates GAL4 production about one day later than A5CGAL, suggesting that this later activation allows more silencing of UASGFP. Furthermore, this demonstrates that derepression can occur both in embryos and in larvae. We conclude that transcriptional silencing by heterochromatin is strongly derepressed by early promoter activation.

ure 3), but in each case, more derepression was seen when a line was combined with the strong embryonic A5CGAL driver. The GawBT80 driver also suppressed variegation more efficiently than armGAL. Thus, a minimal level of GAL4 is required for derepression of silencing, implying that silencing is an equilibrium system instead of an on/off switch.

Derepression Depends on Transcription Factor Abundance Studies in yeast and in mammalian cells have demonstrated that gene activation can suppress variegation (Renauld et al., 1993; Aparicio and Gottschling, 1994; Walters et al., 1996). Lu et al. (1998) have tested this model in Drosophila by comparing two adjacent promoters near heterochromatin. They found no difference in variegation of a strong Hsp70 promoter and the weaker one of a mini-white gene. We wondered if modulating the activator levels for a UAS promoter might reveal an effect on variegation. To assess whether activator levels affect variegation, we crossed different drivers to our target variegator lines. The severity of variegation in all lines depended on the driver used, confirming our results with x21 that derepression results from early promoter activation. Importantly, we found differences in the level of derepression using drivers that are similar in timing but differ in strength. All target lines showed severe silencing when combined with the weak embryonic driver armGAL (Fig-

Developmental Restriction Occurs in Certain Target Lines Developmentally restricted expression of a variegating HS-lacZ gene has been attributed to complete repression by tightly compacted heterochromatin in mitotic cells, followed by the “relaxation” of heterochromatin in differentiating postmitotic cells (Lu et al., 1996, 1998). In two of our lines, x18.3 and x18.4, we noted a similar restriction with the weak armGAL driver (Figure 3). However, using stronger drivers (GawBT80 and A5CGAL), no such developmental restriction was observed (Figure 3 and data not shown). Developmental restriction does not reflect an absence of GAL4 in the anterior portion of the disc, since armGAL expresses GAL4 throughout the eye disc. Lack of developmental restriction was also observed for 8 other variegator lines using armGAL (Figure 3 and data not shown), suggesting that this developmental restriction is specific for only two target lines (x18.3 and x18.4), and then, only at low levels of activation. Our results indicate that complete repression in mitotically active cells is not a general feature of hetero-

Cell 842

Figure 4. New Derepressed Cells Appear in Cultured Discs Eye imaginal discs from larvae carrying the GawBT80 driver and the x21 variegator were photographed immediately after dissection (A) and 24 hr later (B). A merge of phase contrast images (gray) and GFP fluorescence (green) of the disc at each time point is shown. (C and D) Magnifications of the bracketed regions in (A) and (B). Only the GFP channel is shown. New GFP⫹spots have appeared in the later time point (D).

chromatic silencing. It is possible that in certain lines, complete repression in mitotically active cells results from regulatory elements nearby whose effects are seen only when a reporter gene is near heterochromatin. Switching of a Silenced Gene during Development The observation that strong, early-acting drivers were most effective at derepressing silencing might indicate that UASGFP activation at an early critical time prevents the establishment of silencing. Alternatively, derepression and silencing may occur throughout development, and the presence of GAL4 might prevent derepressed cells from switching back to a silenced state. Earlyacting drivers would then result in the accumulation of derepressed cells. We observed that GFP⫹ patches in imaginal discs could be quite variable in size (Figures 2 and 3), suggesting that derepression was occurring throughout development. To directly detect derepression events, we cultured eye imaginal discs from third instar larvae carrying an early-activating GAL4 driver (GawBT80) and the x21 variegator rearrangement. The patterns of GFP expression were then photographed at successive time intervals. If a cell became derepressed during the culture period, we expected to detect it as a new spot of GFP fluorescence. We found that 11/20 viable discs (55%) from variegator larvae did indeed show one or more new GFP⫹ spots within 24 hr of in vitro culturing (Figure 4). We could unambiguously identify spots that appeared in large silenced regions of discs as newly derepressed cells, but derepression events in or near patches that are GFP⫹ in the first time point may have been missed. Since GAL4 was constitutively expressed in the disc throughout the culture period, we conclude that a spontaneous event was occurring in some cells that allowed the silenced UASGFP gene to be derepressed. Interestingly, new

GFP⫹ spots appeared both posterior and anterior of the MF in roughly equal numbers. The majority of cells posterior to the MF are postmitotic and have begun differentiation, so that the appearance of new spots in this region of a disc shows that derepression can occur spontaneously in such cells. We also observed four discs (20%) in which GFP⫹ spots had disappeared after 24 hr in culture. These may be due to silencing of the UASGFP gene, but since apoptosis is common during normal differentiation of imaginal discs (Ready et al., 1976), the disappearance of some GFP⫹ cells might instead be due to their death. Even if all of the disappearances were silencing events, the number of derepression events exceeds that of silencing in these discs. We conclude that switching of silencing is biased toward derepression in these variegator lines when combined with early-acting drivers, and thereby results in the accumulation of derepressed cells. Derepression of a Nearby Gene Derepression of UASGFP silencing by GAL4 raises the question of whether an adjacent gene is affected. We therefore examined variegation of the mini-white gene present in the target P transposon. To quantitate the effects of GAL4 on mini-white mosaic expression, we estimated the severity of silencing in an eye as the fraction of the eye that was unpigmented, and compared siblings from crosses between the variegator and driver lines. Flies carrying each of the UASGFP variegators and the A5CGAL driver showed more derepression than siblings that lacked GAL4 (Figure 5, P ⬍ 0.0001 for each variegator). GAL4 driver chromosomes had no effect on variegation of w⫹ in males carrying In(1)wm4 (P ⫽ 0.69), ruling out that these stocks carry trans-acting modifiers. We conclude that bound GAL4 derepresses silencing at the nearby mini-white gene. Derepression at mini-white may be a consequence of transcription at UASGFP, or of the bound activator itself. Using the A5CGAL driver, we observed quenching of mini-white in the control line, seen as paler overall pigmentation (Figure 5, control), presumably due to occlusion or read-through of this downstream gene. Quenching is unrelated to heterochromatic silencing, because similar reductions in pigmentation are seen in both euchromatic control and variegator lines. This weak quenching did not interfere with our ability to assay derepression. However, quenching was complete in many lines when we used the stronger, late-expressing GMRGAL driver, obviating the assay for derepression (e.g., Figure 5, control, x18, and x18.4). We reasoned that quenching of the downstream gene by expression from the upstream UAS could be eliminated by a local P-induced rearrangement selected for GMRGAL-dependent mini-white expression. The GAL4 dependence of mini-white expression implies that the UAS is still nearby in a rearrangement derivative. This proved to be the case, as we could readily select GMRGAL-dependent lines with strong pigmentation derived from the control line (see Experimental Procedures). Similar P-transposase mutagenesis of variegator lines yielded six derivatives. Each showed mini-white silencing in the eye, but when GMRGAL was present, the silencing was dramati-

GAL4 Activator Counteracts Variegation 843

Figure 6. Uncoupled Expression of UASGFP and mini-white in Variegator Lines Images of an eye from an adult fly carrying (A) x21 and A5CGAL (B) x18.4.1 and GMRGAL. Light microscope images of the eyes (left), and epifluorescent images (middle and right) are shown. In the epifluorescent images, pigment of the eye is in red and in the merged images, GFP is in green. The red pigments of the eye obscure GFP fluorescence where they coincide. The lines mark patches where mini-white is silenced but UASGFP is expressed.

Figure 5. GAL4 Binding at a UAS Counteracts Silencing at a Nearby mini-white Gene Each set of heads carries the indicated target chromosome (control or a variegator). Flies in each row are male siblings from a cross that either carry the indicated GAL4 source (“⫹” ⫽ A5CGAL or GMRGAL) or a dominantly marked homolog (“⫺”). The last pair of heads from the x18.4.1 variegator line also carry a dominant suppressor of silencing, Su(var)20502.

cally derepressed (x18.4.1 and x18.3.1 in Figure 5). The additional inclusion of a suppressor of variegation mutation almost completely derepressed silencing (x18.4.1; Su(var)20502/GMRGAL in Figure 5), further confirming that the white patches in this variegator line are due to heterochromatic silencing, and not to some form of quenching. These results demonstrate that substantial derepression of variegation at a nearby mini-white gene occurs when GAL4 is expressed. We asked whether binding of GAL4 near the miniwhite gene is sufficient for its expression, or whether promoter elements necessary for tissue-specific expression of white participate. Therefore, we examined tissues in which mini-white was not normally expressed but which have GAL4. If GAL4 is sufficient, then these cells should express mini-white. Pigmentation in the adult testis sheath requires white protein, but the miniwhite gene lacks the enhancer that directs expression in this tissue and thus w⫺ males carrying this gene have clear testes (Levis et al., 1985; Pirrotta et al., 1985). A5CGAL expresses UASGFP in this tissue (data not shown), and if GAL4 directs expression of mini-white in

these lines, the testes of w/Y;A5CGAL/P[EP] flies should be pigmented (these flies have fully pigmented eyes). However, their testes were clear (data not shown), implying that tissue-specific white regulatory elements participate in mini-white expression, even when GAL4 is present. Silencing of UASGFP and mini-white Can Be Uncoupled GFP fluorescence can be observed in adult eyes (Edwards et al., 1997), allowing us to assay derepression of UASGFP and mini-white simultaneously. We examined their expression in flies carrying a GAL4 driver and the x21 and x18.4.1 variegator chromosomes. Because red pigments of the adult eye obscure GFP fluorescence, we could distinguish three of the four kinds of facets in the eyes of adult flies: w⫺ GFP⫹, w⫺ GFP⫺, and w⫹ (w⫹ GFP⫹ and w⫹ GFP⫺). The eyes of flies carrying x21 showed weak w⫹ pigmentation and spots of GFP⫹. The promoters of the UASGFP and mini-white genes in x21 are only ⵑ2 kb apart, yet all eyes showed many w⫺ regions that contained GFP⫹ spots, as well as regions in which w⫹ and GFP⫹ overlapped (Figure 6). All the eyes examined from flies carrying x18.4.1 also showed a large number of w⫺ GFP⫹ ommatidia; this is clearest in large w⫺ patches, which seem to be almost entirely GFP⫹ (Figure 6). In w⫺ GFP⫹ regions, GAL4 has bound at and activated UASGFP but the nearby mini-white remains silenced, demonstrating that silencing by heterochromatin at the UASGFP and mini-white genes can be uncoupled. The predominance of w⫺ GFP⫹ cells over w⫺ GFP⫺ cells in the adult eye is consistent with the expression of UASGFP seen in eye imaginal discs from larvae carrying GMRGAL and x18.4.1: these discs show near-uniform expression of UASGFP, implying that the majority of

Cell 844

the adult eye will be GFP⫹. Indeed, nearly uniform GFP fluorescence was observed in flies that carried a background mutation that reduces the level of bright red pigmentation (data not shown). Therefore, UASGFP derepression is more extensive than mini-white derepression in the same tissue, and we expect that most w⫹ cells, if not all, are also GFP⫹. In eyes with pale pigmentation, we could detect GFP⫹ in the photoreceptors of w⫹ ommatidia (Figure 6A), supporting the idea that individual w⫹ cells in these ommatidia are also GFP⫹. Such an association between activity of mini-white and UASGFP suggests that binding of GAL4 to UASGFP causes derepression at the neighboring mini-white promoter. Discussion We have shown that the binding of GAL4 activator at a promoter counteracts heterochromatin-mediated silencing. UASGFP genes near heterochromatin showed mosaic expression when GAL4 was produced in the differentiating eye, but more cells expressed UASGFP when GAL4 was produced throughout development. High levels of GAL4 were required to efficiently counteract silencing in mitotic cells. GAL4 also counteracted silencing of a mini-white gene near UASGFP. These results demonstrate that activation at the promoter is a crucial determinant of whether a gene is silenced. Our results bear directly on the basis of mosaicism that is characteristic of PEV and other silencing phenomena. The prevailing view has been that mosaic silencing results from cell-to-cell variations in whether heterochromatin extends to the reporter gene, rendering it inaccessible to transcription factors. However, some evidence has raised the possibility that sensitivity to heterochromatin differs between genes (Clark and Chovnick, 1986; Roseman et al., 1995; Wines et al., 1996; Talbert and Henikoff, 2000), and it is not clear how reporter gene sensitivity could determine cell-to-cell variation of heterochromatin. Furthermore, this scenario does not predict that varying the levels of a transcriptional activator would efficiently interfere with the heterochromatic state, as we have shown here. This paradox also arises from studies of the mammalian ␤-globin locus, where compromised enhancer activity led to mosaicism rather than reduced uniform expression (Walters et al., 1995, 1996). To accommodate these results, we would need to imagine that the inaccessible state is actually an ensemble of states that could each be “opened” at different activator levels. A simpler alternative is that the reporter gene juxtaposed to heterochromatin is in only one chromatin state, and in this state, a transcription factor binding site is compromised. Reduced protein footprints (Cryderman et al., 1999) and reduced enzyme accessibility (Wallrath and Elgin, 1995; Boivin and Dura, 1998) at sequences near heterochromatin provide support for this explanation. Extensive work indicates that a binding site on DNA wrapped in a nucleosome is inaccessible to its cognate factor, but becomes transiently exposed as contacts between DNA and histones break and reform (Polach and Widom, 1995; Widom, 1999). Thus, the formation of an intermediate in which the site is exposed is required to allow binding of a cognate factor. We suggest

Figure 7. The Site Exposure Model for Variegated Silencing (adapted from Widom, 1999) When DNA containing a factor binding site (black) is wrapped around a nucleosome (gray), a single factor site is inaccessible (the N conformation). Transient unwrapping of the DNA exposes the site (conformation N⬘), which can then be bound by its cognate factor (F) (giving conformation N⬘F). N and N⬘F correspond to the silent and active states, respectively. The production of the N and N⬘F conformations is determined by the relative reaction rates (arrows) involving the common substrate N⬘. In euchromatin, k1 weakly favors the N conformation, but k2 is large, and the production of N⬘F is strongly favored. We postulate that in heterochromatin, k2 remains large, but k1 has increased. Thus, both N and N⬘F conformations will be frequently produced, resulting in mosaic expression.

that a factor binding site in heterochromatin becomes compromised because the stability of DNA–histone contacts is increased, perhaps due to histone modifications, associated heterochromatin proteins, or effects of higher-order chromatin structure. In the wild-type euchromatic state, DNA–histone contacts are easily broken, and a binding site will be readily exposed as a segment of DNA unwraps from the nucleosome. Since this intermediate is efficiently bound by factor, the two steps produce the active conformation in vast excess (Figure 7, N⬘F). In the heterochromatic state, exposure of the site becomes rare. While the intermediate will still be efficiently bound by factor, the opposing reactions involving the intermediate now produce both the active and the silenced conformations (Figure 7, forms N and N⬘F). Thus, variegation results from a bistable equilibrium. In this model, the equilibrium of a factor with its binding sites is inherently a determinant of mosaicism: increased availability of the factor simply drives occupancy of the site and concomitantly reduces the concentration of the silent state. In a similar vein, modifying the stability of DNA–histone contacts and chromatin “fluidity” may be the regulatory function that chromatin remodeling factors perform during transcriptional activation (Kingston and Narlikar, 1999). While gene silencing always shows stochastic on/off mosaicism, certain lines also show clone-like patterns of expression. Perhaps clones result when the stability of DNA–histone contacts is altered. For example, acetylation of histone tails reduces the affinity of histones for DNA (Grunstein, 1997), and moderately increases the rate of site exposure (Polach et al., 2000). The inheritance of modified histones has been previously suggested to account for clonal inheritance (Steger and Workman, 1996; Wade et al., 1997). This may be because modified histones bias the outcome of the opposing reactions diagrammed in Figure 7. In support of these possibilities, the association of hyperacetylated histone H4 with an inherited active state has recently been reported in Drosophila (Cavalli and Paro, 1999). We have observed that derepression can occur

GAL4 Activator Counteracts Variegation 845

throughout development, in both mitotically active and differentiating regions of cultured imaginal discs. Thus, progression through the cell cycle is not necessary for derepression of heterochromatic silencing in Drosophila, although progression appears to be required for derepression of telomeric silencing in yeast (Aparicio and Gottschling, 1994). Early availability of GAL4 in Drosophila increases the number of cells in which the promoter becomes active. We suggest that the basis for increased derepression is that the constitutively available activator binds to and traps its binding sites whenever they are exposed. Because earlier availability of activator allows a longer period for site exposure to occur at a compromised UAS, derepressed cells will accumulate. Derepression of a silenced mini-white gene occurs when GAL4 is bound at a nearby UAS; thus, it is clear that factor binding sites that do not normally direct expression of a nearby gene can become significant when that gene is repressed by heterochromatin. However, we also observed cells where GAL4 binds and activates UASGFP but the nearby mini-white reporter remains silenced. The uncoupled variegation of genes ⵑ2 kilobases apart implies that promoters individually respond to the heterochromatic state. This result indicates that the transition zone between silenced chromatin and active chromatin can be on the order of kilobases, in contrast to the notion that silenced domains are continuous and extensive. The fine scale structure of chromatin that our results imply extends the conclusion based on phenotypic observations that silencing is discontinuous on a large scale (Belyaeva and Zhimulev, 1991; Talbert and Henikoff, 2000). Many workers have used the term “heterochromatin-like” to describe a continuous compacted structure (Renauld et al., 1993; Orlando and Paro, 1995; Donze et al., 1999). For example, the original chromatin immunoprecipitation mapping of the silenced Bithorax Complex suggested that it was heterochromatinlike (Orlando and Paro, 1993), but later work argued against this, because discontinuities in the silenced region on the scale of a few kilobases were revealed (Strutt et al., 1997). Ironically, a similar picture now applies to heterochromatin itself. Experimental Procedures Drosophila Culture Mutations and chromosomes not discussed here are described at the Flybase (http://flybase.bio.indiana.edu) and at the Berkeley Drosophila Genome Project (http://www.fruitfly.org) websites. The parental control insertion line carrying the inducible green fluorescent protein gene UASGFP (P[UASGFPS65T, mini-white]T10) was obtained from the Bloomington Drosophila Stock Center (Bloomington, IN). All flies were raised at 25⬚C on standard cornmeal medium. GAL4 Driver Lines Lines expressing GAL4 used in this study were obtained from the Bloomington Drosophila Stock Center. The enhancer trap line GawBT80 has been reported by Wilder and Perrimon (1995), and the line we refer to as “armGAL” carries two insertions of a GAL4 gene under control of the armadillo promoter (4a and 4b; Sanson et al., 1996). The GMRGAL and A5CGAL drivers are modified versions of P[GMR-GAL4, mini-white]12 (Freeman, 1996) and P[A5CGAL4, mini-white]25 (Ito et al., 1997), respectively, in which the miniwhite genes were deleted by P-transposase mutagenesis. Activity of the GAL4 genes in the w⫺ derivatives was confirmed by crossing flies to P[UASGFP, mini-white]T10 and then examining GFP in the

eye imaginal disc. The w⫺ derivative of P[GMR-GAL4, mini-white]12 is referred to as GMRGAL. The w⫺ derivative of P[A5C-GAL4, miniwhite]25 is referred to as A5CGAL. The time at which drivers first became active was determined by scoring for GFP fluorescence in the diploid cells of embryos and staged larvae. A5CGAL becomes active in embryos, and GawBT80 first induces UASGFP in first instar larvae. Previous reports have described the expression timing of GMRGAL (Freeman, 1996), and of armGAL (Sanson et al., 1996). Higher expression from the Actin5C promoter relative to the armadillo promoter in L3 larvae was described by Riggleman et al. (1989). To compare the strength of drivers, we measured the expression levels from each driver line as the mean GFP fluorescence intensity, in arbitrary units (U), in the posterior regions of eye imaginal discs from male third instar larvae that also carried the UASGFP control insertion. GMRGAL, GawBT80, and A5CGAL all showed similarly high expression levels (mean intensities of 104, 90, and 81 U respectively), while armGAL was much weaker (48 U). Variegator Target Lines We generated lines that juxtaposed the P[UASGFP, mini-white]T10 insertion to heterochromatin. The mini-white gene is a derivative of the w⫹ gene and gives an orange color to the eye (Levis et al., 1985). Males carrying P[UASGFP, mini-white]T10 (inserted at 79D-E) were X irradiated with 3000 Rads and mated to w1118 females for four days, after which the males were discarded. Two males showing variegation of mini-white were recovered from ⵑ49,000 male progeny. Salivary chromosome analysis and neuroblast karyotyping identified the first as a translocation with breakpoints at 79E on chromosome 3 and at h19–24 on the long arm of the Y chromosome; this was designated T(Y;3)x18. The second variegator is also a translocation between the Y and chromosome 3, with breakpoints at 79E and h13–15 on the short arm of the Y chromosome, and is designated T(Y;3)x21. These rearrangements are abbreviated as “x18” and “x21”. The position of the P element in these lines was mapped to the 3PYD element in both x18 and x21 by replacing elements with a corresponding element from T(Y;3)R153, a translocation which is exchanged between 78A and the YL arm; thus the two translocations both break on the distal side of the UASGFP insertion. Lines showing enhanced variegation of mini-white were recovered after mobilization of UASGFP by P transposase (Robertson et al., 1988). Twenty-seven independent lines with enhanced variegation were isolated from 118 males of the genotype w;Sb P[⌬2–3, ry⫹](99B)/T(Y;3)x18 crossed to w1118 females. The lines used here are designated x18.1, x18.2, x18.3, and x18.4. Lines Expressing mini-white Near a UAS Promoter In the course of these experiments, we determined that the binding of GAL4 to a UAS site quenches the neighboring mini-white gene at euchromatic insertion sites (Figure 5). Flies carrying the control UASGFP insertion and GMRGAL have almost no detectable eye pigmentation. Reduced pigmentation is also observed when these drivers were combined with a UAS-lacZ insertion, but not with insertions of constructs that lack a UAS promoter. Quenching of miniwhite only occurs when this gene is downstream of a UAS promoter, since P[EP] insertion lines (which carry a UAS promoter oriented away from mini-white [Rørth, 1996]) show increased expression of mini-white. To create lines in which mini-white showed GAL4dependent enhanced expression, we crossed 53 males carrying the control insertion and Sb P[⌬2–3, ry⫹ ](99B) to w;GMRGAL females. We established 10 lines that showed increased pigmentation in the presence of GMRGAL, which demonstrate that the quenching effect can be eliminated at the euchromatic position. To create lines in which the insertion was near heterochromatin but quenching of miniwhite was eliminated, we crossed 52 males carrying the enhanced variegator derivative T(Y;3)x18.4 and Sb P[⌬2–3, ry⫹ ](99B) to w;GMRGAL females. We established five variegator lines (x18.4.1–5) that showed pigmentation in the presence of GMRGAL. Mutagenesis of the variegator line x18.3 generated one similar derivative line (x18.3.1). Examination of Eye Imaginal Discs For fixed samples, crawling third instar larvae were dissected in phosphate-buffered saline (PBS), fixed for 10 min on ice in 4%

Cell 846

paraformaldehyde/PBS, washed with PBS supplemented with 0.3% Triton X-100 (PBT), and counterstained with 50 nM Syto17 (Molecular Probes, Eugene, OR) in PBT. Tissues were mounted in PBT and examined using a TCS confocal microscope (Leica, Exton, PA). GFP was detected after illumination with the 488 nm line in the 500–550 nm range. Tissue from at least ten larvae for each of the genotypes discussed in the text were examined, and the discs shown are representative examples. Collected z series through the tissues were projected and exported to Adobe Photoshop (Mountainview, CA). For sequential live imaging of eye imaginal discs, we followed the disc culturing procedure of Li and Meinertzhagen (1995). Crawling third instar larvae were washed in 70% ethanol and dissected under sterile conditions in Insect-Xpress media (BioWhittaker, Walkersville MD). Discs were then cultured on a glass slide under a supported coverslip in Insect-Xpress media supplemented with 10% fetal bovine serum, and photographed with low light intensities on a Deltavision microscope (Applied Precision). All images were deconvolved using Deltavision software, and maximum intensity projections of a z series through the entire tissue were compared. Examination of Adult Eyes Brightfield images of adult fly heads were mounted under oil and photographed as described in Ahmad and Golic (1996), or with a Coolpix 990 digital camera (Nikon) mounted on a dissecting microscope. Epifluorescence images of heads were taken on a Deltavision microscope with a 20⫻ objective. GFP was detected using an FITC filterset (490/20 and 528/38) and the red pigments of the eye with Texas-Red filters (555/28 and 617/73). All images were adjusted in Photoshop. Statistical Analysis of mini-white Silencing Males carrying a variegator rearrangement were crossed to females heterozygous for a GAL4 driver (GMRGAL or A5CGAL) or a control chromosome (⫹), and CyO. Variegation at the w⫹ gene in In(1)wmw was used as a control for UAS-independent effects of GAL4 on PEV. Silencing was measured as the area of each eye that was unpigmented, and male progeny from each pair of genotypes were divided into 6 phenotypic ranks, depending on the severity of silencing. Each genotypic class contained at least 20 flies. Probability values (P) were determined by a Mann-Whitney test between each pair of genotypes. Acknowledgments We thank Devi Freudiger and Georgette Sass. K. A. is an ACS postdoctoral fellow. This work was supported by HHMI. Received August 7, 2000; revised February 5, 2001. References

Clark, S.H., and Chovnick, A. (1986). Studies of normal and positionaffected expression of rosy region genes in Drosophila melanogaster. Genetics 114, 819–840. Cryderman, D.E., Tang, H., Bell, C., Gilmour, D.S., and Wallrath, L.L. (1999). Heterochromatic silencing of Drosophila heat shock genes acts at the level of promoter potentiation. Nucleic Acids Res. 27, 3364–3370. Donze, D., Adams, C.R., Rine, J., and Kamakaka, R.T. (1999). The boundaries of the silenced HMR domain in Saccharomyces cerevisiae. Genes Dev. 13, 698–708. Edwards, K.A., Demsky, M., Montague, R.A., Weymouth, N., and Kiehart, D.P. (1997). GFP-Moesin illuminates actin cytoskeleton dynamics in living tissue and demonstrates cell shape changes during morphogenesis. Dev. Biol. 191, 103–117. Elgin, S.C. (1996). Heterochromatin and gene regulation in Drosophila. Curr. Opin. Genet. Dev. 6, 193–202. Freeman, M. (1996). Reiterative use of the EGF receptor triggers differentiation of all cell types in the Drosophila eye. Cell 87, 651–660. Grunstein, M. (1997). Histone acetylation in chromatin structure and transcription. Nature 389, 349–352. Hayashi, S., Ruddell, A., Sinclair, D., and Grigliatti, T. (1990). Chromosomal structure is altered by mutations that suppress or enhance position effect variegation. Chromosoma 99, 391–400. Henikoff, S. (1990). Position-effect variegation after 60 years. Trends Genet. 6, 422–426. Henikoff, S. (1997). Nuclear organization and gene expression: homologous pairing and long-range interactions. Curr. Opin. Cell Biol. 9, 388–395. Ito, K., Awano, W., Suzuki, K., Hiromi, Y., and Yamamoto, D. (1997). The Drosophila mushroom body is a quadruple structure of clonal units each of which contains a virtually identical set of neurones and glial cells. Development 124, 761–771. Kingston, R.E., and Narlikar, G.J. (1999). ATP-dependent remodeling and acetylation as regulators of chromatin fluidity. Genes Dev. 13, 2339–2352. Levis, R., Hazelrigg, T., and Rubin, G.M. (1985). Separable cis-acting control elements for expression of the white gene of Drosophila. EMBO J. 4, 3489–3492. Li, C., and Meinertzhagen, I.A. (1995). Conditions for the primary culture of eye imaginal discs from Drosophila melanogaster. J. Neurobiol. 28, 363–380. Lu, B.Y., Bishop, C.P., and Eissenberg, J.C. (1996). Developmental timing and tissue specificity of heterochromatin-mediated silencing. EMBO J. 15, 1323–1332. Lu, B.Y., Ma, J., and Eissenberg, J.C. (1998). Developmental regulation of heterochromatin-mediated gene silencing in Drosophila. Development 125, 2223–2234.

Ahmad, K., and Golic, K.G. (1996). Somatic reversion of chromosomal position effects in Drosophila melanogaster. Genetics 144, 657–670.

Orlando, V., and Paro, R. (1993). Mapping Polycomb-repressed domains in the bithorax complex using in vivo formaldehyde crosslinked chromatin. Cell 75, 1187–1198.

Aparicio, O.M., and Gottschling, D.E. (1994). Overcoming telomeric silencing: a trans-activator competes to establish gene expression in a cell-cycle dependent way. Genes Dev. 8, 1133–1146.

Orlando, V., and Paro, R. (1995). Chromatin multiprotein complexes involved in the maintenance of transcription programs. Curr. Opin. Genet. Dev. 5, 174–179.

Belyaeva, E.S., and Zhimulev, I.F. (1991). Cytogenetic and molecular aspects of position effect variegation in Drosophila III. Continuous and discontinuous compaction of chromosomal material as a result of position effect variegation. Chromosoma 100, 453–466.

Pirrotta, V., Steller, H., and Bozzetti, M.P. (1985). Multiple upstream regulatory elements control the expression of the Drosophila white gene. EMBO J. 4, 3501–3508.

Boivin, A., and Dura, J.-M. (1998). In vivo chromatin accessibility correlates with gene silencing in Drosophila. Genetics 150, 1539– 1549. Bonifer, C. (2000). Developmental regulation of eukaryotic gene loci. Trends Genet. 16, 310–315. Brand, A.H., and Perrimon, N. (1993). Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development 118, 401–415. Cavalli, G., and Paro, R. (1999). Epigenetic inheritance of active chromatin after removal of the main transactivator. Science 286, 955–958.

Polach, K.J., and Widom, J. (1995). Mechanism of protein access to specific DNA sequences in chromatin: a dynamic equilibrium model for gene regulation. J. Mol. Biol. 254, 130–149. Polach, K.J., Lowary, P.T., and Widom, J. (2000). Effects of core histone tail domains on the equilibrium constants for dynamic DNA site accessibility in nucleosomes. J. Mol. Biol. 298, 211–223. Ready, D.F., Hanson, T.E., and Benzer, S. (1976). Development of the Drosophila retina, a neurocrystalline lattice. Dev. Biol. 53, 217–240. Renauld, H., Aparicio, O.M., Zierath, P.D., Billington, B.L., Chablani, S.K., and Gottschling, D.E. (1993). Silent domains are assembled continuously from the telomere and are defined by promoter distance and strength, and by SIR3 dosage. Genes Dev. 7, 1133–1145.

GAL4 Activator Counteracts Variegation 847

Riggleman, B., Wieschaus, E., and Schedl, P. (1989). Molecular analysis of the armadillo locus: uniformly distributed transcripts and a protein with novel internal repeats are associated with a Drosophila segment polarity gene. Genes Dev. 3, 96–113. Robertson, H.M., Preston, C.R., Phillis, R.W., Johnson-Schlitz, D.M., Benz, W.K., and Engels, W.R. (1988). A stable genomic source of P element transposase in Drosophila melanogaster. Genetics 118, 461–470. Rørth, P. (1996). A modular misexpression screen in Drosophila detecting tissue-specific phenotypes. Proc. Natl. Acad. Sci. USA 93, 12418–12422. Roseman, R.R., Johnson, E.A., Rodesch, C.K., Bjerke, M., Nagoshi, R.N., and Geyer, P.K. (1995). A P element containing suppressor of Hairy-wing binding regions has novel properties for mutagenesis of Drosophila melanogaster. Genetics 141, 1061–1074. Roseman, R.R., Pirrotta, V., and Geyer, P.K. (1993). The su(Hw) protein insulates expression of the Drosophila melanogaster white gene from chromosomal position-effects. EMBO J 12, 435–442. Sabl, J.F., and Henikoff, S. (1996). Copy-number and orientation determine the susceptibility of a gene to silencing by nearby heterochromatin in Drosophila. Genetics 142, 447–458. Sanson, B., White, P., and Vincent, J.P. (1996). Uncoupling cadherinbased adhesion from wingless signalling in Drosophila. Nature 383, 627–630. Sass, G.L., and Henikoff, S. (1999). Pairing-dependent mislocalization of a Drosophila brown gene reporter to a heterochromatic environment. Genetics 152, 595–604. Spofford, J.B. (1976). Position-effect variegation in Drosophila. In The Genetics and Biology of Drosophila, M. Ashburner and E. Novitski, eds. (New York: Academic Press). Steger, D.J., and Workman, J.L. (1996). Remodeling chromatin structures for transcription: what happens to the histones? Bioessays 18, 875–884. Strutt, H., Cavalli, G., and Paro, R. (1997). Co-localization of Polycomb protein and GAGA factor on regulatory elements responsible for the maintenance of homeotic gene expression. EMBO J. 16, 3621–3632. Talbert, P.B., and Henikoff, S. (2000). A re-examination of spreading of position-effect variegation in the white-roughest region of Drosophila melanogaster. Genetics 154, 259–272. Tartof, K.D., Bishop, C., Jones, M., Hobbs, C.A., and Locke, J. (1989). Towards an understanding of position effect variegation. Dev. Genet. 10, 162–176. Wade, P.A., Pruss, D., and Wolffe, A.P. (1997). Histone acetylation: chromatin in action. Trends Biochem. Sci. 22, 128–132. Wallrath, L.L., and Elgin, S.C. (1995). Position effect variegation in Drosophila is associated with an altered chromatin structure. Genes Dev. 9, 1263–1277. Walters, M.C., Fiering, S., Eidemiller, J., Magis, W., Groudine, M., and Martin, D.I. (1995). Enhancers increase the probability but not the level of gene expression. Proc. Natl. Acad. Sci. USA 92, 7125– 7129. Walters, M.C., Magis, W., Fiering, S., Eidemiller, J., Scalzo, D., Groudine, M., and Martin, D.I. (1996). Transcriptional enhancers act in cis to suppress position-effect variegation. Genes Dev. 10, 185–195. Widom, J. (1999). Equilibrium and dynamic nucleosome stability. Methods Mol. Biol. 119, 61–77. Wilder, E.L., and Perrimon, N. (1995). Dual functions of wingless in the Drosophila leg imaginal disc. Development 121, 477–488. Wines, D.R., Talbert, P.B., Clark, D.V., and Henikoff, S. (1996). Introduction of a DNA methyltransferase into Drosophila to probe chromatin structure in vivo. Chromosoma 104, 332–340. Zhimulev, I.F., Belyaeva, E.S., Fomina, O.V., Protopopov, M.O., and Bolshakov, V.N. (1986). Cytogenetic and molecular aspects of position effect variegation in Drosophila melanogaster. I. Morphology and genetic activity of the 2AB region in chromosome rearrangement T(1;2)dorvar7. Chromosoma 94, 492–504. Zhimulev, I.F., Belyaeva, E.S., Bgatov, A.V., Baricheva, E.M., and Vlassova, I.E. (1988). Cytogenetic and molecular aspects of position

effect variegation in Drosophila melanogaster. II. Peculiarities of morphology and genetic activity of the 2B region in the T(1;2)dorvar7 chromosome in males. Chromosoma 96, 255–261.